/ldg/ - Local Diffusion General
Anonymous 01/13/25(Mon)22:27:41 | 320 comments | 76 images | 🔒 Locked
Discussion of Free and Open-Source Diffusion models.
Pretending to be Retarded Edition
Last Bread: >>103880366
Newfrien Guide: https://rentry.org/ldg-lazy-getting-started-guide
>UI
Metastable:https://metastable.studio
SwarmUI:https://github.com/mcmonkeyprojects/SwarmUI
Forge:https://github.com/lllyasviel/stable-diffusion-webui-forge
reForge:https://github.com/Panchovix/stable-diffusion-webui-reForge
ComfyUI:https://github.com/comfyanonymous/ComfyUI
InvokeAI:https://github.com/invoke-ai/InvokeAI
>Models, LoRAs, & Upscalers
https://civitai.com
https://tensor.art/
https://openmodeldb.info
>Training
https://github.com/Nerogar/OneTrainer
https://github.com/kohya-ss/sd-scripts/tree/sd3
https://github.com/derrian-distro/LoRA_Easy_Training_Scripts
>HunyuanVideo
Comfy:https://github.com/kijai/ComfyUI-HunyuanVideoWrapper/
Windows:https://rentry.org/crhcqq54
Training:https://github.com/tdrussell/diffusion-pipe|https://github.com/kohya-ss/musubi-tuner
>Flux
Forge:https://github.com/lllyasviel/stable-diffusion-webui-forge/discussions/1050
ComfyUI:https://comfyanonymous.github.io/ComfyUI_examples/flux
DeDistilled:https://huggingface.co/TheYuriLover/flux-dev-de-distill-GGUF/tree/main
Pixelwave:https://civitai.com/models/141592/pixelwave
>Misc
Share Metadata:https://catbox.moe|https://litterbox.catbox.moe/
Img2Prompt:https://huggingface.co/spaces/fancyfeast/joy-caption-alpha-two
Archive:https://rentry.org/sdg-link
Samplers:https://stable-diffusion-art.com/samplers/
Digital Art Software:https://krita.org/en/
Txt2Img Plugin:https://kritaaidiffusion.com/
Collagebaker:https://www.befunky.com/create/collage/
Video Collagebaker:https://kdenlive.org/en/
>Related
>>>/aco/csdg
>>>/b/degen
>>>/c/kdg
>>>/d/ddg
>>>/e/edg
>>>/h/hdg
>>>/trash/slop
>>>/vt/vtai
>Local Sexting
>>>/g/lmg
>Maintain thread quality
https://rentry.org/debo
Pretending to be Retarded Edition
Last Bread: >>103880366
Newfrien Guide: https://rentry.org/ldg-lazy-getting
>UI
Metastable:https://metastable.studi
SwarmUI:https://github.com/mcmonkey
Forge:https://github.com/lllyasviel
reForge:https://github.com/Panchovi
ComfyUI:https://github.com/comfyano
InvokeAI:https://github.com/invoke-
>Models, LoRAs, & Upscalers
https://civitai.com
https://tensor.art/
https://openmodeldb.info
>Training
https://github.com/Nerogar/OneTrain
https://github.com/kohya-ss/sd-scri
https://github.com/derrian-distro/L
>HunyuanVideo
Comfy:https://github.com/kijai/Comf
Windows:https://rentry.org/crhcqq54
Training:https://github.com/tdrusse
>Flux
Forge:https://github.com/lllyasviel
ComfyUI:https://comfyanonymous.gith
DeDistilled:https://huggingface.co/
Pixelwave:https://civitai.com/model
>Misc
Share Metadata:https://catbox.moe|https:/
Img2Prompt:https://huggingface.co/s
Archive:https://rentry.org/sdg-link
Samplers:https://stable-diffusion-a
Digital Art Software:https://krita.org/en/
Txt2Img Plugin:https://kritaaidiffusion.com
Collagebaker:https://www.befunky.co
Video Collagebaker:https://kdenlive.org/e
>Related
>>>/aco/csdg
>>>/b/degen
>>>/c/kdg
>>>/d/ddg
>>>/e/edg
>>>/h/hdg
>>>/trash/slop
>>>/vt/vtai
>Local Sexting
>>>/g/lmg
>Maintain thread quality
https://rentry.org/debo
Anonymous 01/13/25(Mon)22:28:35 No.103886331
>my gen nowhere to be found in the collage
Anonymous 01/13/25(Mon)22:29:06 No.103886338
total expert base model prompter death
Anonymous 01/13/25(Mon)22:29:46 No.103886343
Blessed thread of frenship
Anonymous 01/13/25(Mon)22:30:28 No.103886350
>hunyuan TWITTER hasn't even posted in a week
They pulled the H800 plug
They pulled the H800 plug
Anonymous 01/13/25(Mon)22:31:39 No.103886360
>>103886350
I was about to doom aswell but they updated their github, so they're not giving up on this project I guess
I was about to doom aswell but they updated their github, so they're not giving up on this project I guess
Anonymous 01/13/25(Mon)22:32:11 No.103886366
im nostalgic for that time when the pixart guys let us gen on their laptop
Anonymous 01/13/25(Mon)22:33:10 No.103886378
show me your pics you want to come to life, what's in the tank
Anonymous 01/13/25(Mon)22:34:06 No.103886392
>>103885833
Evangelion vibes from this for some reason
Evangelion vibes from this for some reason
Anonymous 01/13/25(Mon)22:34:17 No.103886395
>>103886378
blue board
blue board
Anonymous 01/13/25(Mon)22:34:26 No.103886398
>>103886338
You're jealous.
You're jealous.
Anonymous 01/13/25(Mon)22:35:42 No.103886409
>>103886378
I don't want to be permab&
I don't want to be permab&
Anonymous 01/13/25(Mon)22:37:22 No.103886426
>>103886328
>>103886337
That's why I said you choose the right frame, just like a director.
You don't continue from someone running, you continue after someone has finished running for example.
All the negatives you tsalk abotu are thing that can be solved and in ways we don't know about yet, because guess what new techniques get created all the time.
>>103886337
That's why I said you choose the right frame, just like a director.
You don't continue from someone running, you continue after someone has finished running for example.
All the negatives you tsalk abotu are thing that can be solved and in ways we don't know about yet, because guess what new techniques get created all the time.
Anonymous 01/13/25(Mon)22:38:10 No.103886436
You know /g/ is dead when this thread is the most active on the board lol
Anonymous 01/13/25(Mon)22:38:57 No.103886442
>>103886436
>You know /g/ is dead
why would it be dead? it's not like people aren't interested in technology anymore right?
>You know /g/ is dead
why would it be dead? it's not like people aren't interested in technology anymore right?
Anonymous 01/13/25(Mon)22:41:25 No.103886455
>>103886426
As I said. Good luck.
As I said. Good luck.
Anonymous 01/13/25(Mon)22:42:29 No.103886463
Anonymous 01/13/25(Mon)22:44:14 No.103886481
>>103886463
The most innocuous video I've seen in a while here.
The most innocuous video I've seen in a while here.
Anonymous 01/13/25(Mon)22:44:22 No.103886482
>>103886455
Thanks, I understand where you are coming from. But we shall see
Thanks, I understand where you are coming from. But we shall see
Anonymous 01/13/25(Mon)22:45:10 No.103886491
Anonymous 01/13/25(Mon)22:45:15 No.103886492
>>103886481
Sorry
Sorry
Anonymous 01/13/25(Mon)22:48:07 No.103886509
>>103886463
it should have exploded or something
it should have exploded or something
Anonymous 01/13/25(Mon)22:48:53 No.103886517
>>103886492
My penis remembers
My penis remembers
Anonymous 01/13/25(Mon)22:48:56 No.103886520
Anonymous 01/13/25(Mon)22:50:16 No.103886529
Anonymous 01/13/25(Mon)22:51:51 No.103886541
Despite earlier claims, Hunyuan can't do cunny.
Anonymous 01/13/25(Mon)23:00:45 No.103886598
>>103886360
I mean the penguin bench? Hyvid even has official comfy support and they won't tick it off.
I mean the penguin bench? Hyvid even has official comfy support and they won't tick it off.
Anonymous 01/13/25(Mon)23:01:46 No.103886607
>>103886598
yeah I know, but it shows they're still working on it, it's not like it's over for them, at least we got signs they're still alive
yeah I know, but it shows they're still working on it, it's not like it's over for them, at least we got signs they're still alive
Anonymous 01/13/25(Mon)23:02:43 No.103886612
isn't tencent a military company? do they produce munitions or mustard gas or something
Anonymous 01/13/25(Mon)23:02:51 No.103886614
>>103886541
how stringent was your testing process?
how stringent was your testing process?
Anonymous 01/13/25(Mon)23:03:12 No.103886615
>>103886612
100 pound coom bombs mostly
100 pound coom bombs mostly
Anonymous 01/13/25(Mon)23:03:31 No.103886618
>>103886612
no it's a big video game vendor
no it's a big video game vendor
Anonymous 01/13/25(Mon)23:05:08 No.103886628
im a genner
i gen
i gen
Anonymous 01/13/25(Mon)23:06:03 No.103886635
>>103886612
Nah, no more than any other tech company.
The west is in a weird period of self destructive paranoia right now so things are getting weird.
Nah, no more than any other tech company.
The west is in a weird period of self destructive paranoia right now so things are getting weird.
Anonymous 01/13/25(Mon)23:07:50 No.103886647
Anonymous 01/13/25(Mon)23:08:47 No.103886652
Anonymous 01/13/25(Mon)23:08:48 No.103886653
>>103886635
>The west is in a weird period of self destructive paranoia right now
I really pity the zoomers who've only seen the "self destructive paranoia" era of the west, I'm old enough to have lived through the 90's when it was still fine, I wonder if they find this normal or they're looking at old ass series/movies and they're crying like litte bitches because they knew how much better it was back then
>The west is in a weird period of self destructive paranoia right now
I really pity the zoomers who've only seen the "self destructive paranoia" era of the west, I'm old enough to have lived through the 90's when it was still fine, I wonder if they find this normal or they're looking at old ass series/movies and they're crying like litte bitches because they knew how much better it was back then
Anonymous 01/13/25(Mon)23:09:29 No.103886657
>>103886653
it normal desu
it normal desu
Anonymous 01/13/25(Mon)23:10:50 No.103886669
>>103886653
I remember the 90s, when we thought about the future we kept thinking wow things are gonna keep getting cooler. Good vibes, yeah bad things still happened, but there was an overall bigger atmosphere of hope and people were more chill in general without social media etc
Now it's more like doom and gloom when we think ahead
I remember the 90s, when we thought about the future we kept thinking wow things are gonna keep getting cooler. Good vibes, yeah bad things still happened, but there was an overall bigger atmosphere of hope and people were more chill in general without social media etc
Now it's more like doom and gloom when we think ahead
Anonymous 01/13/25(Mon)23:10:54 No.103886670
>>103886635
it's america, and they're changing administration soon
tencent are commies but I don't care if their product is good, that's meritocracy
it's america, and they're changing administration soon
tencent are commies but I don't care if their product is good, that's meritocracy
Anonymous 01/13/25(Mon)23:11:16 No.103886675
Local Diffusion?
Anonymous 01/13/25(Mon)23:11:37 No.103886677
>>103886657
I'll never find this normal
https://www.youtube.com/watch?v=ox2O6QiNwv8
https://youtu.be/1JkPHdlJc20?t=106
I'll never find this normal
https://www.youtube.com/watch?v=ox2
https://youtu.be/1JkPHdlJc20?t=106
Anonymous 01/13/25(Mon)23:12:28 No.103886682
Anonymous 01/13/25(Mon)23:12:58 No.103886687
>>103886669
>Now it's more like doom and gloom when we think ahead
Nah, we're at the verge of i2v. If it works as we think it will (ie: at all) the future will be so bright we'll need sunglasses.
>Now it's more like doom and gloom when we think ahead
Nah, we're at the verge of i2v. If it works as we think it will (ie: at all) the future will be so bright we'll need sunglasses.
Anonymous 01/13/25(Mon)23:13:43 No.103886694
>>103886687
I'm also the i2v guy, I guess being a 90s kid makes me automatically be more hopeful in general
I'm also the i2v guy, I guess being a 90s kid makes me automatically be more hopeful in general
Anonymous 01/13/25(Mon)23:13:59 No.103886696
>>103886669
>>103886687
true, at this point AI is the only good thing that happened to the west the last 15 years, and they want to kill it and let China eat all the cake, this is insanity
>>103886687
true, at this point AI is the only good thing that happened to the west the last 15 years, and they want to kill it and let China eat all the cake, this is insanity
Anonymous 01/13/25(Mon)23:14:22 No.103886702
>>103886687
hailuo is a pretty advanced model, it can produce good outputs from time to time but its i2v sucks
maybe loras are enough to make up for it, but we're starting from square 1 with those when the weights drop
hailuo is a pretty advanced model, it can produce good outputs from time to time but its i2v sucks
maybe loras are enough to make up for it, but we're starting from square 1 with those when the weights drop
Anonymous 01/13/25(Mon)23:15:22 No.103886715
>>103886677
i dont know the first one but from the thumbnail looks after my time. more for kids born 2010 and after
the second one.. the last AAA title i played was six years ago
>>103886694
hope is not generational desu
i dont know the first one but from the thumbnail looks after my time. more for kids born 2010 and after
the second one.. the last AAA title i played was six years ago
>>103886694
hope is not generational desu
Anonymous 01/13/25(Mon)23:15:31 No.103886717
>>103886702
>hailuo is a pretty advanced model, it can produce good outputs from time to time but its i2v sucks
what about kling 1.6? I've heard it's currently the best one
>hailuo is a pretty advanced model, it can produce good outputs from time to time but its i2v sucks
what about kling 1.6? I've heard it's currently the best one
Anonymous 01/13/25(Mon)23:16:10 No.103886723
>>103886717
if you gotta pay then I haven't used it
if you gotta pay then I haven't used it
Anonymous 01/13/25(Mon)23:16:38 No.103886725
>>103886715
>the last AAA title i played was six years ago
that's the issue with this new era, back then AAA games never missed, we had fucking ocarina of time, tekken 3, half life, gta san andreas, they were the kings of the world, now everyone assosiate AAA with woke garbage, as it should, but it really lost its prestige
>the last AAA title i played was six years ago
that's the issue with this new era, back then AAA games never missed, we had fucking ocarina of time, tekken 3, half life, gta san andreas, they were the kings of the world, now everyone assosiate AAA with woke garbage, as it should, but it really lost its prestige
Anonymous 01/13/25(Mon)23:17:26 No.103886730
>>103886682
Prompt?
Prompt?
Anonymous 01/13/25(Mon)23:17:41 No.103886735
>>103886715
>more for kids born 2010 and after
which is kinda predatory, teaching kids mentally ill stuff like this shit should be a crime
>more for kids born 2010 and after
which is kinda predatory, teaching kids mentally ill stuff like this shit should be a crime
Anonymous 01/13/25(Mon)23:17:42 No.103886736
Anonymous 01/13/25(Mon)23:19:53 No.103886752
>>103886730
>High quality cinematic closeup video of glass beaker filled with small samples of black charcoal. A dropper is seen dripping small amounts of a clear fluid into the beaker. The charcoal samples transform into rough deep blue sapphire. The video is photorealistic, with natural diffuse lighting.
>High quality cinematic closeup video of glass beaker filled with small samples of black charcoal. A dropper is seen dripping small amounts of a clear fluid into the beaker. The charcoal samples transform into rough deep blue sapphire. The video is photorealistic, with natural diffuse lighting.
Anonymous 01/13/25(Mon)23:20:46 No.103886759
>>103886628
Writing/tattoos on the body like that will always wow me
Writing/tattoos on the body like that will always wow me
Anonymous 01/13/25(Mon)23:20:51 No.103886761
>>103886378
tubgirl
tubgirl
Anonymous 01/13/25(Mon)23:22:31 No.103886771
Anonymous 01/13/25(Mon)23:31:05 No.103886829
Anonymous 01/13/25(Mon)23:37:50 No.103886882
Is there a good, recent, brainlet-tier guide to LoRA training? Ideally one that comes with an example dataset so I can make sure the scripts work and understand the proper formats for everything before I start assembling my actual data
Anonymous 01/13/25(Mon)23:39:56 No.103886904
>>103886882
Maybe another anon can link something recent but if you have any specific questions you can try asking chatgpt or claude.
Maybe another anon can link something recent but if you have any specific questions you can try asking chatgpt or claude.
Anonymous 01/13/25(Mon)23:43:17 No.103886929
>>103886331
i still love you anon
i still love you anon
Anonymous 01/13/25(Mon)23:50:11 No.103886975
>a five second video of a steak being cut took 48 minutes to render
We will never get good photorealist oneeloli like this.
>>103886715
>born 2010 and after
Aired 2021-2023, aimed at early school/preschool kids. They're really fucking up the next generation.
We will never get good photorealist oneeloli like this.
>>103886715
>born 2010 and after
Aired 2021-2023, aimed at early school/preschool kids. They're really fucking up the next generation.
Anonymous 01/13/25(Mon)23:59:35 No.103887044
Anonymous 01/14/25(Tue)00:00:20 No.103887051
>>103886975
>They're really fucking up the next generation.
that's the reason Trump got elected a second time, to stop this insanity, as it should
>They're really fucking up the next generation.
that's the reason Trump got elected a second time, to stop this insanity, as it should
Anonymous 01/14/25(Tue)00:16:00 No.103887144
>>103887110
>many closed online only services like klingai, pixverse, hailuoai, (and they all have image to video) but nothing local
yeah it's a bit weird so many services are popping up yet no more local competition exists, and zero local for i2v
>many closed online only services like klingai, pixverse, hailuoai, (and they all have image to video) but nothing local
yeah it's a bit weird so many services are popping up yet no more local competition exists, and zero local for i2v
Anonymous 01/14/25(Tue)00:18:24 No.103887154
>>103886702
>it can produce good outputs from time to time but its i2v sucks
It's fine, basically kling > hailuo > pixverse > anything else for "interesting" anatomy in motion
no idea where sora is but I have no desire to touch that and be banned because the rock I genned looked too horny
>it can produce good outputs from time to time but its i2v sucks
It's fine, basically kling > hailuo > pixverse > anything else for "interesting" anatomy in motion
no idea where sora is but I have no desire to touch that and be banned because the rock I genned looked too horny
Anonymous 01/14/25(Tue)00:18:54 No.103887157
>>103887144
>it's a bit weird so many services are popping up yet no more local competition exists, and zero local for i2v
it's perfectly "normal", they don't want the goyims to have access to powerful local models, it's too "dangerous" for the goyims and should only be for our trusted big company overloards, got it goyim?
>it's a bit weird so many services are popping up yet no more local competition exists, and zero local for i2v
it's perfectly "normal", they don't want the goyims to have access to powerful local models, it's too "dangerous" for the goyims and should only be for our trusted big company overloards, got it goyim?
Anonymous 01/14/25(Tue)00:19:56 No.103887163
>>103887154
sora sucks anyway, it was way better before but then they cucked it and distilled it to make a turbo version, the chink API models are way better than this turd
https://www.reddit.com/r/StableDiffusion/comments/1i0tqst/sora_img2vid_is_overated_it_does_too_much_and_yet/
sora sucks anyway, it was way better before but then they cucked it and distilled it to make a turbo version, the chink API models are way better than this turd
https://www.reddit.com/r/StableDiff
Anonymous 01/14/25(Tue)00:20:43 No.103887170
Anonymous 01/14/25(Tue)00:22:22 No.103887177
Just figured out i have been using vpred models incorrectly. Its been interesting learning what they can actually do
Anonymous 01/14/25(Tue)00:23:27 No.103887180
103887157
>when I stumble my toe in the morning, I blame the jews
>>103887163
holy shit it looks so pitiful
my guess is that they just lost too many brains and it shows on their latest sota models
+ probably not trained on a lot of copyrighted material and porn (even softcore), which chinese models don't give a shit about
>when I stumble my toe in the morning, I blame the jews
>>103887163
holy shit it looks so pitiful
my guess is that they just lost too many brains and it shows on their latest sota models
+ probably not trained on a lot of copyrighted material and porn (even softcore), which chinese models don't give a shit about
Anonymous 01/14/25(Tue)00:24:12 No.103887182
>>103887177
How were you using them before?
How were you using them before?
Anonymous 01/14/25(Tue)00:24:49 No.103887184
>>103887180
>>when I stumble my toe in the morning, I blame the jews
So you're trying to pretend that Sam the jew hasn't seen the US senate to kill the local ecosystem? lol, lmao even?
>>when I stumble my toe in the morning, I blame the jews
So you're trying to pretend that Sam the jew hasn't seen the US senate to kill the local ecosystem? lol, lmao even?
Anonymous 01/14/25(Tue)00:25:49 No.103887192
>>103887163
It's crazy it looks so shit, reminds on how everyone is complaining that DALLE looked worse and worse.
It's crazy it looks so shit, reminds on how everyone is complaining that DALLE looked worse and worse.
Anonymous 01/14/25(Tue)00:27:28 No.103887201
>>103887180
>probably not trained on a lot of copyrighted material and porn (even softcore), which chinese models don't give a shit about
which is ironic because China has porn banned and the US celebrate porn degenerancy
>probably not trained on a lot of copyrighted material and porn (even softcore), which chinese models don't give a shit about
which is ironic because China has porn banned and the US celebrate porn degenerancy
Anonymous 01/14/25(Tue)00:28:26 No.103887207
>Is there any competition locally that is trained to compete with Hunyuan Video?
what about mochi?
what about mochi?
Anonymous 01/14/25(Tue)00:29:28 No.103887213
>>103887207
I don't like mochi, it always give those weird blurry frames as if you were drunk or something
I don't like mochi, it always give those weird blurry frames as if you were drunk or something
Anonymous 01/14/25(Tue)00:32:33 No.103887225
>>103887182
i did not have
>Noise Schedule: Zero Terminal SNR
basically negates all the advantages of a vpred model
i did not have
>Noise Schedule: Zero Terminal SNR
basically negates all the advantages of a vpred model
Anonymous 01/14/25(Tue)00:32:53 No.103887227
>>103887207
did they ever do anything once they released their model, like promising i2v?
did they ever do anything once they released their model, like promising i2v?
Anonymous 01/14/25(Tue)00:35:08 No.103887238
>>103887227
they promised the VAE encoder, the i2v model and mochiHD before 2025, and nothing ever happened
they promised the VAE encoder, the i2v model and mochiHD before 2025, and nothing ever happened
Anonymous 01/14/25(Tue)00:37:43 No.103887251
>/sdg/ is literally all off topic talk with grown men posting as their avatars
>/ldgl/ is littered with hunyuan and flux shills harassing each other
Pick your poison
>/ldgl/ is littered with hunyuan and flux shills harassing each other
Pick your poison
Anonymous 01/14/25(Tue)00:39:00 No.103887260
>>103887251
>harassing each other
what are you talking about? so far the discussion is cozy and is about what happened to mochi or when will hunyuan i2v will release
>harassing each other
what are you talking about? so far the discussion is cozy and is about what happened to mochi or when will hunyuan i2v will release
Anonymous 01/14/25(Tue)00:39:19 No.103887262
>>103887225
Was this though an extension or? Can you link me?
Was this though an extension or? Can you link me?
Anonymous 01/14/25(Tue)00:40:14 No.103887266
>>103880735
was this a gen? how
was this a gen? how
Anonymous 01/14/25(Tue)00:40:22 No.103887268
>>103887238
ok so yeah they're functionally dead
ok so yeah they're functionally dead
Anonymous 01/14/25(Tue)00:40:24 No.103887269
Anonymous 01/14/25(Tue)00:41:14 No.103887273
>>103887268
I think they just gave up, they can't compete with hunyuan and they know it
I think they just gave up, they can't compete with hunyuan and they know it
Anonymous 01/14/25(Tue)00:42:02 No.103887280
>>103887266
he took a selfie and img2img'd it.
he took a selfie and img2img'd it.
Anonymous 01/14/25(Tue)00:45:58 No.103887302
>>103887262
It's in seeing if you are using a web gui. For comfy no clue. Just search noise in settings and you'll find it
It's in seeing if you are using a web gui. For comfy no clue. Just search noise in settings and you'll find it
Anonymous 01/14/25(Tue)00:47:12 No.103887306
>>103887251
>hunyuan and flux shills
Just gen with what you want and show us how good of a model it is
>hunyuan and flux shills
Just gen with what you want and show us how good of a model it is
Anonymous 01/14/25(Tue)00:49:00 No.103887323
>>103887251
>when people enjoy hunyuan and flux's output they are shills
>BUT, when people enjoy SDXL's output they are genuine fans
curious how that works huh
>when people enjoy hunyuan and flux's output they are shills
>BUT, when people enjoy SDXL's output they are genuine fans
curious how that works huh
Anonymous 01/14/25(Tue)00:51:49 No.103887334
https://civitai.com/user/razzz/models
that razzz guy isn't afraid of death for real, his only loras are about female celebrities, there's no way he won't get in trouble by one of those roasties
that razzz guy isn't afraid of death for real, his only loras are about female celebrities, there's no way he won't get in trouble by one of those roasties
Anonymous 01/14/25(Tue)00:53:12 No.103887339
>>103887334
he has quite a few download them now
he has quite a few download them now
Anonymous 01/14/25(Tue)00:54:05 No.103887349
>>103887334
The quality seems good too bad his taste in women is poor
The quality seems good too bad his taste in women is poor
Anonymous 01/14/25(Tue)00:56:27 No.103887365
fuck this shit, i'm going back to sd 1.5
Anonymous 01/14/25(Tue)00:57:28 No.103887372
>>103887365
you are a free man anon, do whatever pleases you the most
you are a free man anon, do whatever pleases you the most
Anonymous 01/14/25(Tue)01:01:01 No.103887390
hmmmm
Anonymous 01/14/25(Tue)01:02:18 No.103887396
Anonymous 01/14/25(Tue)01:02:29 No.103887398
>>103887154
>kling > hailuo > pixverse > anything else
surprised none of them leaked, with how jank some of the websites are
>kling > hailuo > pixverse > anything else
surprised none of them leaked, with how jank some of the websites are
Anonymous 01/14/25(Tue)01:03:50 No.103887403
Anonymous 01/14/25(Tue)01:04:01 No.103887405
>>103887398
what about veo2? I know it's not publicly API released but I've seen of its i2v gens and its quite good
what about veo2? I know it's not publicly API released but I've seen of its i2v gens and its quite good
Anonymous 01/14/25(Tue)01:05:01 No.103887412
>>103887403
ooch, in a human analogy that must hurt for the guy
ooch, in a human analogy that must hurt for the guy
Anonymous 01/14/25(Tue)01:20:29 No.103887501
>>103887405
didn't try it so dunno
didn't try it so dunno
Anonymous 01/14/25(Tue)01:22:07 No.103887514
>>103887501
NTA but veo 2 shits on basically everything and it's not even close.
NTA but veo 2 shits on basically everything and it's not even close.
Anonymous 01/14/25(Tue)01:22:08 No.103887515
I hope you guys show respect for your 1girls by taking them out on a nice date after jackhammering and impregnating them.
Anonymous 01/14/25(Tue)01:23:38 No.103887528
>>103887515
Screencapping this so I have a point of reference on what a post that belongs in /sdg/ looks like.
Screencapping this so I have a point of reference on what a post that belongs in /sdg/ looks like.
Anonymous 01/14/25(Tue)01:24:30 No.103887536
>>103887514
this, veo2 is next level, like wtf happened to google to be this good???
this, veo2 is next level, like wtf happened to google to be this good???
Anonymous 01/14/25(Tue)01:27:05 No.103887552
Anonymous 01/14/25(Tue)01:31:16 No.103887568
Anonymous 01/14/25(Tue)01:34:16 No.103887591
>>103887568
*slurp smack* oh, rei *slobber slurp* i love you so much honey *smack smack mwah* don't pay attention to the people staring at us, they're just jealous *slurp slurp slurrrrrp*
*slurp smack* oh, rei *slobber slurp* i love you so much honey *smack smack mwah* don't pay attention to the people staring at us, they're just jealous *slurp slurp slurrrrrp*
Anonymous 01/14/25(Tue)01:39:59 No.103887617
Anonymous 01/14/25(Tue)01:48:45 No.103887676
Anonymous 01/14/25(Tue)02:02:36 No.103887746
>>103887617
that's a pretty cool gen, bro, honestly a lot better than the 1girl slop that i make
i'm curious to know the artist name(s)
that's a pretty cool gen, bro, honestly a lot better than the 1girl slop that i make
i'm curious to know the artist name(s)
Anonymous 01/14/25(Tue)02:09:28 No.103887785
I wish my 1girl were realidada
Anonymous 01/14/25(Tue)02:20:33 No.103887831
>>103887746
sabu \(sabudenego\), incase, yd_\(orange_maru\)
Those are just for anatomy and composition, I use loras
sabu \(sabudenego\), incase, yd_\(orange_maru\)
Those are just for anatomy and composition, I use loras
Anonymous 01/14/25(Tue)02:52:53 No.103887986
>can get NTRmix to make decent images in my regional prompting setup easily
>trying to use noob pred v or even the recently linked cat tower mix based on pred v generates much rougher images overall
>using pred v out of my regional prompt workflow gens better results
Anyone using pred v in their regional prompt workflows? I have reCFG and the proper pred v stuff configured but I think something in my setup is hated by pred v
>trying to use noob pred v or even the recently linked cat tower mix based on pred v generates much rougher images overall
>using pred v out of my regional prompt workflow gens better results
Anyone using pred v in their regional prompt workflows? I have reCFG and the proper pred v stuff configured but I think something in my setup is hated by pred v
Anonymous 01/14/25(Tue)02:52:59 No.103887987
>>103887617
Soul
Soul
Anonymous 01/14/25(Tue)03:01:37 No.103888033
I liked this hobby better when the art theft wasn't as blatant and using the artists online handle as a prompt.
Anonymous 01/14/25(Tue)03:04:27 No.103888057
>>103888033
(you).
(you).
Anonymous 01/14/25(Tue)03:04:38 No.103888059
>>103888033
>I liked this hobby better when the art theft wasn't as blatant
the art theft was at its peak when SD1.5 was released and had a shit ton of artist names inside of the base model, you have no idea what you're talking about don't you?
>I liked this hobby better when the art theft wasn't as blatant
the art theft was at its peak when SD1.5 was released and had a shit ton of artist names inside of the base model, you have no idea what you're talking about don't you?
Anonymous 01/14/25(Tue)03:05:22 No.103888064
>>103886319
Moar of top right, please
Moar of top right, please
Anonymous 01/14/25(Tue)03:06:23 No.103888069
>>103888059
Nah everything was Greg Rutkowski and a few other literal whos. This is way worse. If your art was on danbooru or any other dataset source your handle is now a means to steal your work.
Nah everything was Greg Rutkowski and a few other literal whos. This is way worse. If your art was on danbooru or any other dataset source your handle is now a means to steal your work.
Anonymous 01/14/25(Tue)03:07:55 No.103888084
>>103888069
>Nah everything was Greg Rutkowski and a few other literal whos. This is way worse.
what is he talking about? the modern models we have don't know any artists compared to before because they are now trained on """ethical""" datasets, Flux doesn't know shit, SD3.5 neither, SD1.5 had way more artists in it, is this a bait or something?
>Nah everything was Greg Rutkowski and a few other literal whos. This is way worse.
what is he talking about? the modern models we have don't know any artists compared to before because they are now trained on """ethical""" datasets, Flux doesn't know shit, SD3.5 neither, SD1.5 had way more artists in it, is this a bait or something?
Anonymous 01/14/25(Tue)03:11:14 No.103888108
Anonymous 01/14/25(Tue)03:11:52 No.103888112
>>103888108
they are, but like I said they don't have any artists tags in it, Flux can't even make Picasso
they are, but like I said they don't have any artists tags in it, Flux can't even make Picasso
Anonymous 01/14/25(Tue)03:15:04 No.103888131
>>103887365
Based
Based
Anonymous 01/14/25(Tue)03:17:23 No.103888144
Illustrioushitters, why is there always a medal/star/shiny/trophy in my images?
I tried putting them into the goddamn negatives, but it is still there.
I changed to another Illustrious checkpoint, it is still there.
I am getting fucking schizophrenic about this stupid problem and it won't fucking stop no matter how much I try to get it away.
Like this shit doesn't exist when I gen shit with Pony checkpoints, but when I fucking use Illustrious it is always there.
I'm at my wits end because of that goddamn problem.
I tried putting them into the goddamn negatives, but it is still there.
I changed to another Illustrious checkpoint, it is still there.
I am getting fucking schizophrenic about this stupid problem and it won't fucking stop no matter how much I try to get it away.
Like this shit doesn't exist when I gen shit with Pony checkpoints, but when I fucking use Illustrious it is always there.
I'm at my wits end because of that goddamn problem.
Anonymous 01/14/25(Tue)03:17:44 No.103888146
Anonymous 01/14/25(Tue)03:18:02 No.103888150
>>103888069
>>103888084
>>103888108
>entire prompt no negatives: brutalist ruins, acid rain, plant overgrowth, oil painting by albert bierstadt
>Stable Diffusion 1.4
It was a different time
>>103888084
>>103888108
>entire prompt no negatives: brutalist ruins, acid rain, plant overgrowth, oil painting by albert bierstadt
>Stable Diffusion 1.4
It was a different time
Anonymous 01/14/25(Tue)03:18:11 No.103888151
>>103888144
You're going to have to tell us the prompt.
You're going to have to tell us the prompt.
Anonymous 01/14/25(Tue)03:21:12 No.103888173
>>103888150
>corrugated metal hatchback car, leaking ((pearly opalescent glowing fluid)) white yogurt slime, art by yoshiyuki tomino, zdzisław beksiński, albert bierstadt, abandoned space hangar interior, fine exquisite details, copper and chartreuse tones, sharp lighting
>corrugated metal hatchback car, leaking ((pearly opalescent glowing fluid)) white yogurt slime, art by yoshiyuki tomino, zdzisław beksiński, albert bierstadt, abandoned space hangar interior, fine exquisite details, copper and chartreuse tones, sharp lighting
Anonymous 01/14/25(Tue)03:22:18 No.103888179
>>103888144
Post examples and catbox some gens so we can see your workflows, prompts, and checkpoints.
I wanna see these
Post examples and catbox some gens so we can see your workflows, prompts, and checkpoints.
I wanna see these
Anonymous 01/14/25(Tue)03:23:48 No.103888190
>>103888150
>>103888173
imagine if you combined the artist tag sovl from sd1.4 to the refined outputs of Flux, would be the best model ever
>>103888173
imagine if you combined the artist tag sovl from sd1.4 to the refined outputs of Flux, would be the best model ever
Anonymous 01/14/25(Tue)03:29:25 No.103888223
>>103888190
We had it so good. 1.4 had a lot more compositional diversity for whatever reason as well. Flux and DALL-E 3 seem to have limited perspectives presumably to deal with earlier SD's inability to maintain a consistent perspective.
We had it so good. 1.4 had a lot more compositional diversity for whatever reason as well. Flux and DALL-E 3 seem to have limited perspectives presumably to deal with earlier SD's inability to maintain a consistent perspective.
Anonymous 01/14/25(Tue)03:29:49 No.103888227
>>103888179
>>103888151
Wait, I actually discovered the source.
It was from the Style Selector node that I used while I was still using Pony, literally turned sai-enhance off and it solved it.
I am retarded.
>>103888151
Wait, I actually discovered the source.
It was from the Style Selector node that I used while I was still using Pony, literally turned sai-enhance off and it solved it.
I am retarded.
Anonymous 01/14/25(Tue)03:30:00 No.103888229
>>103873054
I tried very hastily
If you're still around someone recently made many yugioh loras
https://civitai.com/models/626302?modelVersionId=1257362
I tried very hastily
If you're still around someone recently made many yugioh loras
https://civitai.com/models/626302?m
Anonymous 01/14/25(Tue)03:43:13 No.103888294
catchup summary on AI inference and training from nothing 1
Anonymous 01/14/25(Tue)03:43:38 No.103888298
>>103887514
>veo 2 shits on basically everything and it's not even close.
understatement of the year
https://xcancel.com/kimmonismus/status/1879068296638824958#m
>veo 2 shits on basically everything and it's not even close.
understatement of the year
https://xcancel.com/kimmonismus/sta
Anonymous 01/14/25(Tue)03:46:46 No.103888320
Anonymous 01/14/25(Tue)03:49:02 No.103888340
>>103888298
I think the woman in the purple shirts arm kinda warps at the end there but you'd have to be LOOKING for it.
I think the woman in the purple shirts arm kinda warps at the end there but you'd have to be LOOKING for it.
Anonymous 01/14/25(Tue)03:52:44 No.103888372
>>103888084
They still do know but the association is really weak and are only retained for the dead. So you can still prompt Renoir and Picasso and even as recent as Pollock but anyone contemporary and still alive isn't in because of objections.
They still do know but the association is really weak and are only retained for the dead. So you can still prompt Renoir and Picasso and even as recent as Pollock but anyone contemporary and still alive isn't in because of objections.
Anonymous 01/14/25(Tue)03:55:21 No.103888389
>>103888298
Without reading what you were talking about I thought the video was real, since it wasn't the usual recognizable hunyuanslop.
Without reading what you were talking about I thought the video was real, since it wasn't the usual recognizable hunyuanslop.
Anonymous 01/14/25(Tue)03:57:22 No.103888405
>>103888389
to be fair this is a 1280x720 res of veo2, at 720p hunyuan looks really good, we just can't run it on our vramlet computer, thanks nvdia
to be fair this is a 1280x720 res of veo2, at 720p hunyuan looks really good, we just can't run it on our vramlet computer, thanks nvdia
Anonymous 01/14/25(Tue)04:01:54 No.103888439
Anonymous 01/14/25(Tue)04:05:46 No.103888456
>>103888405
Veo at its worst still runs rings around Hyvid at its best. They clearly did something novel in their training.
Veo at its worst still runs rings around Hyvid at its best. They clearly did something novel in their training.
Anonymous 01/14/25(Tue)04:06:18 No.103888460
Anonymous 01/14/25(Tue)04:06:52 No.103888465
>>103888456
>They clearly did something novel in their training.
google owns youtube, it means they can train on the whole library if they want, and they probably have the deleted videos on their servers aswell, you can't compete with that
>They clearly did something novel in their training.
google owns youtube, it means they can train on the whole library if they want, and they probably have the deleted videos on their servers aswell, you can't compete with that
Anonymous 01/14/25(Tue)04:09:24 No.103888479
I can't post in this thread, are certain words like cuda and pytorch banned or something?
Anonymous 01/14/25(Tue)04:10:21 No.103888493
>>103888479
there are no banned words on 4chan, not that I'm aware of
there are no banned words on 4chan, not that I'm aware of
Anonymous 01/14/25(Tue)04:13:55 No.103888513
>>103888493
I just wrote the same post twice in a row and got the green post successful message, but it's just not here, I'll just include it as a screenshot so I don't get cucked for a 3rd time, seriously I feel schizophrenic, it's not really that important anyway, just a general question
I just wrote the same post twice in a row and got the green post successful message, but it's just not here, I'll just include it as a screenshot so I don't get cucked for a 3rd time, seriously I feel schizophrenic, it's not really that important anyway, just a general question
Anonymous 01/14/25(Tue)04:15:05 No.103888525
>>103888513
dunno why you can't post that, seems like a normal post, weird
dunno why you can't post that, seems like a normal post, weird
Anonymous 01/14/25(Tue)04:15:23 No.103888528
>>103888150
SD1.4/1.5 is still my go-to model for kino outputs
SD1.4/1.5 is still my go-to model for kino outputs
Anonymous 01/14/25(Tue)04:15:53 No.103888532
>>103888513
I hate docker though. I've rarely had to deal with it before, but ever time I do it's not pleasant. Might just be a wsl thing.
I hate docker though. I've rarely had to deal with it before, but ever time I do it's not pleasant. Might just be a wsl thing.
Anonymous 01/14/25(Tue)04:19:33 No.103888550
ComfyUI-Zluda mentions ZLUDA versions 3.8.4 and 3.8.5
Where and when are those two versions from?
Where and when are those two versions from?
Anonymous 01/14/25(Tue)04:20:55 No.103888562
>>103888525
could test some more but also I just don't care, I know some sql queries and stuff will get blocked but it fails to post. But for this, I was getting a successful post with nothing afterward which I've never seen before ever.
>>103888532
I have lots of experience with it so I don't mind it, but not much experience using cuda / gpu stuff with it. I haven't yet tested it on a linux box so it could just be wsl, but yeah beware that it will use your windows-host driver. which kind of defeats the purpose of docker making things little reproducible modules and for different projects that have different pytorch versions you might have to juggle drivers which seems really dumb
could test some more but also I just don't care, I know some sql queries and stuff will get blocked but it fails to post. But for this, I was getting a successful post with nothing afterward which I've never seen before ever.
>>103888532
I have lots of experience with it so I don't mind it, but not much experience using cuda / gpu stuff with it. I haven't yet tested it on a linux box so it could just be wsl, but yeah beware that it will use your windows-host driver. which kind of defeats the purpose of docker making things little reproducible modules and for different projects that have different pytorch versions you might have to juggle drivers which seems really dumb
Anonymous 01/14/25(Tue)04:24:45 No.103888581
>>103888550
Bet you wish you had a Nvidia gpu right now, huh?
Bet you wish you had a Nvidia gpu right now, huh?
Anonymous 01/14/25(Tue)04:26:12 No.103888591
Anonymous 01/14/25(Tue)04:26:47 No.103888596
Anonymous 01/14/25(Tue)04:29:43 No.103888616
>>103888550
>ComfyUI-Zluda
Your reading comprehension...
https://github.com/lshqqytiger/ZLUDA/releases
You need to use the fork pointed out by the repo I linked above and it is best you probably use the latest one anyways.
This is the kind of deep diving and etc. people aren't ready to face and can't admit they are too out of their depth when they need to do this.
I have an Intel card which is even worse with no compatibility layer. Practically everything needs to manually be hacked but I make it work because it isn't my job yet and I can use the learning experience.
>ComfyUI-Zluda
Your reading comprehension...
https://github.com/lshqqytiger/ZLUD
You need to use the fork pointed out by the repo I linked above and it is best you probably use the latest one anyways.
This is the kind of deep diving and etc. people aren't ready to face and can't admit they are too out of their depth when they need to do this.
I have an Intel card which is even worse with no compatibility layer. Practically everything needs to manually be hacked but I make it work because it isn't my job yet and I can use the learning experience.
Anonymous 01/14/25(Tue)04:30:21 No.103888625
>>103888550
https://github.com/lshqqytiger/ZLUDA
I think they forked it from the last version that included AMD-funded code (before AMD said it actually couldn't be shared, no idea if there was any legal basis for that or the creator just didn't want to fight a multi billion dollar company, instead of being the new "general" version with an unknown funder
https://github.com/lshqqytiger/ZLUD
I think they forked it from the last version that included AMD-funded code (before AMD said it actually couldn't be shared, no idea if there was any legal basis for that or the creator just didn't want to fight a multi billion dollar company, instead of being the new "general" version with an unknown funder
Anonymous 01/14/25(Tue)04:31:38 No.103888632
Anonymous 01/14/25(Tue)04:35:42 No.103888652
>>103888632
The fork is a "community" thing from someone's backup of the code with AMD's funded code. Nvidia could technically still sue that but it woulds be bad PR. The main new version of ZLUDA restarted from scratch and has no basis for Nvidia to sue AMD so is legally clear. So yeah, confusing.
The fork is a "community" thing from someone's backup of the code with AMD's funded code. Nvidia could technically still sue that but it woulds be bad PR. The main new version of ZLUDA restarted from scratch and has no basis for Nvidia to sue AMD so is legally clear. So yeah, confusing.
Anonymous 01/14/25(Tue)04:42:57 No.103888687
>>103888632
AMD agreed by email that the project would be open sourced if they stopped funding it. AMD stopped funding it and then the project was open sourced as expected. Months later, AMD said "actually, that code can't be shared" and the dev reverted it
AFAIK he was literally the only dev. There are a few notable "translation layer" projects like ZLUDA, dxvk, gl4es, box86 that are mostly a one man operation.
>>103888652
NVIDIA would have to prove that laws were broken when their code(/hardware) was reverse-engineered, or that patents were infringed. I'm not sure of how different this is from the x86-64 to ARM translators that both Apple and Microsoft built into the operating systems that are being used by millions of commercial devices.
AMD agreed by email that the project would be open sourced if they stopped funding it. AMD stopped funding it and then the project was open sourced as expected. Months later, AMD said "actually, that code can't be shared" and the dev reverted it
AFAIK he was literally the only dev. There are a few notable "translation layer" projects like ZLUDA, dxvk, gl4es, box86 that are mostly a one man operation.
>>103888652
NVIDIA would have to prove that laws were broken when their code(/hardware) was reverse-engineered, or that patents were infringed. I'm not sure of how different this is from the x86-64 to ARM translators that both Apple and Microsoft built into the operating systems that are being used by millions of commercial devices.
Anonymous 01/14/25(Tue)04:44:49 No.103888696
>>103888687
>AFAIK he was literally the only dev.
Imagine hiring only one guy to destroy the monopoly of your biggest rival, totally not a controlled opposition at all...
>AFAIK he was literally the only dev.
Imagine hiring only one guy to destroy the monopoly of your biggest rival, totally not a controlled opposition at all...
Anonymous 01/14/25(Tue)04:52:06 No.103888747
>>103888696
I'm not sure why AMD never said "fuck it" and implemented CUDA straight into their drivers
I'm not sure why AMD never said "fuck it" and implemented CUDA straight into their drivers
Anonymous 01/14/25(Tue)05:02:12 No.103888820
https://github.com/dogehhh/ReCLIP
>ReCLIP++: Learn to Rectify the Bias of CLIP for Unsupervised Semantic Segmentation
I think they improved Clip or something, could this be used on flux or hunyuan?
>ReCLIP++: Learn to Rectify the Bias of CLIP for Unsupervised Semantic Segmentation
I think they improved Clip or something, could this be used on flux or hunyuan?
Anonymous 01/14/25(Tue)05:03:04 No.103888823
I've been away from generation from some time. I see a bunch of new samplers and schedulers. What benefit do they have over a simple ass euler/normal?
I've tried DEIS, but it just seems to produce slightly worse results.
I've tried DEIS, but it just seems to produce slightly worse results.
Anonymous 01/14/25(Tue)05:05:16 No.103888842
>>103888823
if you use hunyuan you have to try res_multistep, that makes the output less blury compared to euler that's really cool
if you use hunyuan you have to try res_multistep, that makes the output less blury compared to euler that's really cool
Anonymous 01/14/25(Tue)05:05:36 No.103888846
>>103888687
x86 had patents and that is massive different in treament than CUDA which has a license and is not open at all. No one is interested in testing this in court like Oracle and Good did over the JDK to deem Google's use of it fair use. A case with CUDA would be its own case and the custom license is a primary blocker because here, there is no OpenJDK equivalent.
x86 had patents and that is massive different in treament than CUDA which has a license and is not open at all. No one is interested in testing this in court like Oracle and Good did over the JDK to deem Google's use of it fair use. A case with CUDA would be its own case and the custom license is a primary blocker because here, there is no OpenJDK equivalent.
Anonymous 01/14/25(Tue)05:25:10 No.103888984
Anonymous 01/14/25(Tue)06:16:38 No.103889284
>>103888465
Do you realize how much automatically-generated slop Youtube is filled with? Even before the advent of AI there was still a giant mound of essentially garbage, so they must've had to do lots of filtering.
Regardless, local will probably be at Veo2 level by this time next year, and paid services will probably be indistinguishable from reality. Oh well, truth was fun while it lasted.
Do you realize how much automatically-generated slop Youtube is filled with? Even before the advent of AI there was still a giant mound of essentially garbage, so they must've had to do lots of filtering.
Regardless, local will probably be at Veo2 level by this time next year, and paid services will probably be indistinguishable from reality. Oh well, truth was fun while it lasted.
Anonymous 01/14/25(Tue)06:18:30 No.103889299
>>103889284
>local will probably be at Veo2 level by this time next year, and paid services will probably be indistinguishable from reality.
Sure an Hatsune Miku is going to reach out of the screen and give me a handjob.
>local will probably be at Veo2 level by this time next year, and paid services will probably be indistinguishable from reality.
Sure an Hatsune Miku is going to reach out of the screen and give me a handjob.
Anonymous 01/14/25(Tue)06:20:05 No.103889304
Anonymous 01/14/25(Tue)06:27:56 No.103889355
>>103889299
You don't believe?
You don't believe?
Anonymous 01/14/25(Tue)07:19:58 No.103889723
https://www.youtube.com/watch?v=-XM29eQGNbw
>Tfw it's Google who killed Holywood at the end
never expected that
>Tfw it's Google who killed Holywood at the end
never expected that
Anonymous 01/14/25(Tue)07:27:21 No.103889772
>>103889723
Veo is like Street Fighter, no one remembered the first iteration, but the 2nd one revolutionized everything kek.
Veo is like Street Fighter, no one remembered the first iteration, but the 2nd one revolutionized everything kek.
Anonymous 01/14/25(Tue)07:31:53 No.103889802
>>103889723
Looks impressive. I still wonder how long a continuous coherent scene you can do with it. It seems like AI generated video tends to a lot of short clips spliced together if that Coke commercial is any indication.
Looks impressive. I still wonder how long a continuous coherent scene you can do with it. It seems like AI generated video tends to a lot of short clips spliced together if that Coke commercial is any indication.
Anonymous 01/14/25(Tue)07:34:00 No.103889817
Cosmos image to video is good.
https://github.com/comfyanonymous/ComfyUI/commit/3aaabb12d422eb35cd0314a09582c0a47d505a37
https://github.com/comfyanonymous/C
Anonymous 01/14/25(Tue)07:34:25 No.103889820
Anonymous 01/14/25(Tue)07:35:53 No.103889831
>>103889817
"CosmosImageToVideoLatent"
can you do the same thing for hunyuan, a sort of "HunyuanImageToVideoLatent", I know we don't have the i2v model yet but maybe the t2v version works ok with it
"CosmosImageToVideoLatent"
can you do the same thing for hunyuan, a sort of "HunyuanImageToVideoLatent", I know we don't have the i2v model yet but maybe the t2v version works ok with it
Anonymous 01/14/25(Tue)07:37:43 No.103889837
>>103889817
your webm example is on the 14b fp8 right?
your webm example is on the 14b fp8 right?
Anonymous 01/14/25(Tue)07:38:51 No.103889844
>>103889802
>It seems like AI generated video tends to a lot of short clips spliced together if that Coke commercial is any indication.
yeah, if you want to make long ass videos you need a shit ton of frames to be calculated at the same time, that means a shit ton of vram aswell
>It seems like AI generated video tends to a lot of short clips spliced together if that Coke commercial is any indication.
yeah, if you want to make long ass videos you need a shit ton of frames to be calculated at the same time, that means a shit ton of vram aswell
Anonymous 01/14/25(Tue)07:41:00 No.103889854
>>103889817
care to share a workflow with it?
care to share a workflow with it?
Anonymous 01/14/25(Tue)07:44:13 No.103889877
>>103889817
>Cosmos image to video is good.
let's hope that'll motivate tencent to release a better i2v model than the cosmos one
>Cosmos image to video is good.
let's hope that'll motivate tencent to release a better i2v model than the cosmos one
Anonymous 01/14/25(Tue)07:44:44 No.103889881
When using cosine annealing warm restarts (cawr) how do I figure out the optimal amount of cycles? is it tied to dataset size?
Anonymous 01/14/25(Tue)08:01:18 No.103889964
>>103889837
14B bf16
>>103889854
https://gist.github.com/comfyanonymous/2f57adabe5a22b36a21ae024306daddb
>>103889831
Once they actually release their model I will.
14B bf16
>>103889854
https://gist.github.com/comfyanonym
>>103889831
Once they actually release their model I will.
Anonymous 01/14/25(Tue)08:02:39 No.103889969
>>103889964
>Once they actually release their model I will.
you don't have the curiosity about how the t2v model can handle i2v process? maybe it's working well like that already, who knows?
>Once they actually release their model I will.
you don't have the curiosity about how the t2v model can handle i2v process? maybe it's working well like that already, who knows?
Anonymous 01/14/25(Tue)08:03:41 No.103889976
Anonymous 01/14/25(Tue)08:11:52 No.103890031
>>103889964
all the video models have this issue of making objects dissapear like nothing, I wonder when that'll be fixed, like having good hands I feel it'll take a while though
all the video models have this issue of making objects dissapear like nothing, I wonder when that'll be fixed, like having good hands I feel it'll take a while though
Anonymous 01/14/25(Tue)08:12:32 No.103890032
is 10-ish images enough to train a good character lora?
Anonymous 01/14/25(Tue)08:15:21 No.103890053
>>103889976
In this computer I have an ada 6000 and a W7900.
>>103889969
The CosmosImageToVideoLatent might work on hunyuan.
What it does is set up the inpainting for Cosmos in a way that it inpaints the frames based on the initial image or sequence of images.
>>103890031
I think the trick with this video model is to include those objects in the prompt. The objects that disappeared in my video above were not in the prompt.
In this computer I have an ada 6000 and a W7900.
>>103889969
The CosmosImageToVideoLatent might work on hunyuan.
What it does is set up the inpainting for Cosmos in a way that it inpaints the frames based on the initial image or sequence of images.
>>103890031
I think the trick with this video model is to include those objects in the prompt. The objects that disappeared in my video above were not in the prompt.
Anonymous 01/14/25(Tue)08:18:14 No.103890069
>>103890053
>I think the trick with this video model is to include those objects in the prompt. The objects that disappeared in my video above were not in the prompt.
Is that why the official repo strongly recommends you use the mistral captioning model when doing img2vid?
>I think the trick with this video model is to include those objects in the prompt. The objects that disappeared in my video above were not in the prompt.
Is that why the official repo strongly recommends you use the mistral captioning model when doing img2vid?
Anonymous 01/14/25(Tue)08:26:08 No.103890141
>>103890053
>The CosmosImageToVideoLatent might work on hunyuan.
yeah it's working I guess but the result is really dissapointing, it acted like the first frame never existed lol
>The CosmosImageToVideoLatent might work on hunyuan.
yeah it's working I guess but the result is really dissapointing, it acted like the first frame never existed lol
Anonymous 01/14/25(Tue)08:27:56 No.103890154
>>103890053
>>103890141
I don't get it though, I put 121 frames and I only got a 2 sec video, should be 5
>>103890141
I don't get it though, I put 121 frames and I only got a 2 sec video, should be 5
Anonymous 01/14/25(Tue)08:34:19 No.103890208
>>103890069
Yeah the model likes long prompts.
>>103890154
That's because hunyuan only has 4x temporal compression vs 8x for cosmos.
>>103890141
That's what happens when the model isn't trained for "inpainting".
Yeah the model likes long prompts.
>>103890154
That's because hunyuan only has 4x temporal compression vs 8x for cosmos.
>>103890141
That's what happens when the model isn't trained for "inpainting".
Anonymous 01/14/25(Tue)08:42:10 No.103890262
>>103889723
Seems too good to be true. I wonder if models with these capabilities will ever be allowed out. Seems the governments around the world simultaneously are afraid of and craving this tech.
Even if it would be slow computing, buying 2 of those new digits mini computers and linking them for 256GB of unified memory will be an option in a few months. Sure it would be $6000 for 2 of them but that is doable for a large amount of people. Meaning full local control outside of any cloud service that could block "bad actors".
I just have a premonition this tech will keep progressing and an 'incident' will occur and it all gets locked down for regular citizens.
Seems too good to be true. I wonder if models with these capabilities will ever be allowed out. Seems the governments around the world simultaneously are afraid of and craving this tech.
Even if it would be slow computing, buying 2 of those new digits mini computers and linking them for 256GB of unified memory will be an option in a few months. Sure it would be $6000 for 2 of them but that is doable for a large amount of people. Meaning full local control outside of any cloud service that could block "bad actors".
I just have a premonition this tech will keep progressing and an 'incident' will occur and it all gets locked down for regular citizens.
Anonymous 01/14/25(Tue)08:43:40 No.103890269
wasnt hunyuan i2v supposed to be released before Dec 31 2024?
Am i mistaken or have they got some VC money and are now knee deep in hookers and cocaine?
Am i mistaken or have they got some VC money and are now knee deep in hookers and cocaine?
Anonymous 01/14/25(Tue)08:44:12 No.103890272
>>103890269
no, they said they'll release it on january
no, they said they'll release it on january
Anonymous 01/14/25(Tue)08:45:04 No.103890278
>>103890272
Thanks anon, my memory is generally at capacity 99.99% of the time
Thanks anon, my memory is generally at capacity 99.99% of the time
Anonymous 01/14/25(Tue)08:46:27 No.103890287
>>103890262
you can't stop progress, if the US are too scared, China will do it instead, at some point those chinks will get how Google made veo2 and they'll make their own chink-veo2 shit, it's just a matter of time at this point
https://www.youtube.com/watch?v=w_1abtGULtg
you can't stop progress, if the US are too scared, China will do it instead, at some point those chinks will get how Google made veo2 and they'll make their own chink-veo2 shit, it's just a matter of time at this point
https://www.youtube.com/watch?v=w_1
Anonymous 01/14/25(Tue)08:51:35 No.103890321
>>103890287
I'm really impressed how fast we progressed though, the transformers architecture is barely 7 years old, do you know how computers looked like 7 years after its creation??
I'm really impressed how fast we progressed though, the transformers architecture is barely 7 years old, do you know how computers looked like 7 years after its creation??
Anonymous 01/14/25(Tue)08:51:42 No.103890323
>>103890287
Sure but you can criminalize its use. The first amendment is on rocky ground, the TikTok ban is a prime example of that. So I wouldn't count on saying using this tech is free speech.
I am not saying progress will stop, quite the opposite. It's all the progress will be walled off to regular citizens. Megacorps and governments will be the sole beneficiaries.
Anything that escapes the garden
Sure but you can criminalize its use. The first amendment is on rocky ground, the TikTok ban is a prime example of that. So I wouldn't count on saying using this tech is free speech.
I am not saying progress will stop, quite the opposite. It's all the progress will be walled off to regular citizens. Megacorps and governments will be the sole beneficiaries.
Anything that escapes the garden
Anonymous 01/14/25(Tue)08:53:39 No.103890332
>>103890323
>Sure but you can criminalize its use. The first amendment is on rocky ground
the 2nd amendment is allowing citizens to have guns, I have to say that a gun is much more dangerous than a video model desu
>Sure but you can criminalize its use. The first amendment is on rocky ground
the 2nd amendment is allowing citizens to have guns, I have to say that a gun is much more dangerous than a video model desu
Anonymous 01/14/25(Tue)08:54:32 No.103890339
>>103890332
Then you lack imagination
Then you lack imagination
Anonymous 01/14/25(Tue)08:56:52 No.103890353
>>103890339
desu we should remove cameras, people can mise-use it, we should also remove photoshop, people can mise use it, we should also remove the internet, people do bad stuff in there, how about removing the computer from the citizens, some use it to make crimes, I think we'll all be safe this way, like North Korea!
desu we should remove cameras, people can mise-use it, we should also remove photoshop, people can mise use it, we should also remove the internet, people do bad stuff in there, how about removing the computer from the citizens, some use it to make crimes, I think we'll all be safe this way, like North Korea!
Anonymous 01/14/25(Tue)09:03:24 No.103890403
>>103890353
Just watch man. There will be some sort of 'incident' that destabilizes and threatens corporate and rich peoples money. It could even be a manufactured false flag incident.
It will empower Congress to pass draconian patriot act like legislation crippling the use of this tech by regular citizens. All in the name of public safety. It will actually just give corporations and oligarchs more security to protect their money and interests.
I give this happening within the next 2 presidencies.
Just watch man. There will be some sort of 'incident' that destabilizes and threatens corporate and rich peoples money. It could even be a manufactured false flag incident.
It will empower Congress to pass draconian patriot act like legislation crippling the use of this tech by regular citizens. All in the name of public safety. It will actually just give corporations and oligarchs more security to protect their money and interests.
I give this happening within the next 2 presidencies.
Anonymous 01/14/25(Tue)09:05:04 No.103890422
>>103890403
>I give this happening within the next 2 presidencies.
I doubt trump will do something like that, he and Elon are big fans of AI, that gives us 4 years to get our veo2 local model, I hope we'll make it for real
>I give this happening within the next 2 presidencies.
I doubt trump will do something like that, he and Elon are big fans of AI, that gives us 4 years to get our veo2 local model, I hope we'll make it for real
Anonymous 01/14/25(Tue)09:08:16 No.103890443
>>103890353
There was talk of restricting the purchase of GPU hardware "capable of AI" over a year ago.
There was talk of restricting the purchase of GPU hardware "capable of AI" over a year ago.
Anonymous 01/14/25(Tue)09:09:27 No.103890451
>>103890443
the us senate love to talk, they also talked about forbidding violent games like doom and gta 3 decades ago, and nothing happened lol
the us senate love to talk, they also talked about forbidding violent games like doom and gta 3 decades ago, and nothing happened lol
Anonymous 01/14/25(Tue)09:22:04 No.103890548
>>103890443
nta, but it just happened yesterday, but at a country level.
any country in the yellow zone is limited in the amount of gpus they can purchase, anything as powerful as a 4090 is going to be impossible to get as a normal consumer
https://public-inspection.federalregister.gov/2025-00636.pdf
https://www.whitehouse.gov/briefing-room/statements-releases/2025/01/13/fact-sheet-ensuring-u-s-security-and-economic-strength-in-the-age-of-artificial-intelligence/
nta, but it just happened yesterday, but at a country level.
any country in the yellow zone is limited in the amount of gpus they can purchase, anything as powerful as a 4090 is going to be impossible to get as a normal consumer
https://public-inspection.federalre
https://www.whitehouse.gov/briefing
Anonymous 01/14/25(Tue)09:23:06 No.103890558
>>103890548
sorry picture for ants
sorry picture for ants
Anonymous 01/14/25(Tue)09:24:00 No.103890564
>>103890558
>only shit countries that never contributed anything to AI (except china :'( )got restrictions on GPUs
lawwwwwwwl
>only shit countries that never contributed anything to AI (except china :'( )got restrictions on GPUs
lawwwwwwwl
Anonymous 01/14/25(Tue)09:26:14 No.103890587
>>103890548
>>103890558
you know the US is still the king of the world when they're the one who decide who's worthy of GPUs lol
>>103890558
you know the US is still the king of the world when they're the one who decide who's worthy of GPUs lol
Anonymous 01/14/25(Tue)09:28:01 No.103890602
>>103889723
https://www.youtube.com/watch?v=7LfvVkI_X0w
>high quality shitpost
that's why I love video models
https://www.youtube.com/watch?v=7Lf
>high quality shitpost
that's why I love video models
Anonymous 01/14/25(Tue)09:28:08 No.103890605
>>103890564
i don't think that's the point. how can they control what the masses have access to if someone else on the other side of the world releases a less restrictive and more powerful model? they can't have something like huyan mogging openai's own video model
i don't think that's the point. how can they control what the masses have access to if someone else on the other side of the world releases a less restrictive and more powerful model? they can't have something like huyan mogging openai's own video model
Anonymous 01/14/25(Tue)09:29:04 No.103890616
>>103890605
>how can they control what the masses have access to if someone else on the other side of the world releases a less restrictive and more powerful model?
what can I say? they're just fucking retarded
>how can they control what the masses have access to if someone else on the other side of the world releases a less restrictive and more powerful model?
what can I say? they're just fucking retarded
Anonymous 01/14/25(Tue)09:31:24 No.103890639
>>103886319
forget the dramma, merge
forget the dramma, merge
Anonymous 01/14/25(Tue)09:33:06 No.103890650
Anonymous 01/14/25(Tue)09:34:52 No.103890665
>>103890605
The point is to make the "bad countries" unable to train larger models than they currently have while "good countries" can
The point is to make the "bad countries" unable to train larger models than they currently have while "good countries" can
Anonymous 01/14/25(Tue)09:35:03 No.103890668
Anonymous 01/14/25(Tue)09:39:01 No.103890710
>>103890665
oh yeah, if Algeria doesn't make AI models it's definitely because of the restrictions, and not because the average IQ in there is in the 70's or something lol
oh yeah, if Algeria doesn't make AI models it's definitely because of the restrictions, and not because the average IQ in there is in the 70's or something lol
Anonymous 01/14/25(Tue)09:40:29 No.103890724
>>103890665
and gamers get fucked in the process since in a few GPU generations it might be impossible to buy any modern gpu
and gamers get fucked in the process since in a few GPU generations it might be impossible to buy any modern gpu
Anonymous 01/14/25(Tue)09:46:33 No.103890776
for reference, since this was announced, all the 4090s are gone from all online retailers in my country. anything as powerful as a 4090 counts as an advanced gpu
Anonymous 01/14/25(Tue)09:46:56 No.103890782
>>103890710
It's for preventing Chinese companies from training AI models using their subsidiaries in "bad countries"
It's for preventing Chinese companies from training AI models using their subsidiaries in "bad countries"
Anonymous 01/14/25(Tue)09:50:16 No.103890804
>>103890724
there is already a lobotomized 4090 designed for the chinese market
now this will apply to all tier2/3 countries, the future is now
there is already a lobotomized 4090 designed for the chinese market
now this will apply to all tier2/3 countries, the future is now
Anonymous 01/14/25(Tue)09:50:27 No.103890807
>>103890782
i have a couple bridges to sell you if you are interested
i have a couple bridges to sell you if you are interested
Anonymous 01/14/25(Tue)09:51:04 No.103890816
Anonymous 01/14/25(Tue)09:54:33 No.103890852
>>103890804
so are GPUs just the new nukes? because the countries getting affected by this either start producing their own chips, or they are effectively a technological slave of a country that can produce chips
so are GPUs just the new nukes? because the countries getting affected by this either start producing their own chips, or they are effectively a technological slave of a country that can produce chips
Anonymous 01/14/25(Tue)09:55:07 No.103890858
>>103890650
discussion is better with ai pictures in an image forum
>how did you manage the 5 fingers?
discussion is better with ai pictures in an image forum
>how did you manage the 5 fingers?
Anonymous 01/14/25(Tue)09:57:00 No.103890882
>>103890858
>discussion is better with ai pictures in an image forum
what's preventing you to go back to /sdg/, they do exactly what you love
>discussion is better with ai pictures in an image forum
what's preventing you to go back to /sdg/, they do exactly what you love
Anonymous 01/14/25(Tue)09:57:38 No.103890890
>have to move to another country to game in 2035
grim
grim
Anonymous 01/14/25(Tue)09:57:51 No.103890893
>>103889964
go back to your containment thread, you fucking piece of shit
go back to your containment thread, you fucking piece of shit
Anonymous 01/14/25(Tue)09:59:54 No.103890909
>>103890852
>or they are effectively a technological slave of a country that can produce chips
yes, chips are the new nukes and usa is the chip overlord and fighting to maintain its position
usa is already deciding who get the most advanced chip making machines and who don't, so no country can produce fast chips unless usa allows them
>or they are effectively a technological slave of a country that can produce chips
yes, chips are the new nukes and usa is the chip overlord and fighting to maintain its position
usa is already deciding who get the most advanced chip making machines and who don't, so no country can produce fast chips unless usa allows them
Anonymous 01/14/25(Tue)10:04:36 No.103890969
>>103890558
yeah yeah, that won't happen
yeah yeah, that won't happen
Anonymous 01/14/25(Tue)10:08:49 No.103891021
>>103890893
where is this vitrol when other avatars start batch spamming shit in here? he is actually posting about tech, unlike the usual slop batch spammers.
i think you are the schizo, since he has shown this pattern before, of being extremely aggressive when one of his "friends" posts on this thread instead of the other one.
where is this vitrol when other avatars start batch spamming shit in here? he is actually posting about tech, unlike the usual slop batch spammers.
i think you are the schizo, since he has shown this pattern before, of being extremely aggressive when one of his "friends" posts on this thread instead of the other one.
Anonymous 01/14/25(Tue)10:11:38 No.103891045
>>103890969
it was enough to cause all the 4090s to go out of stock on my country, so something happened.
it was enough to cause all the 4090s to go out of stock on my country, so something happened.
Anonymous 01/14/25(Tue)10:14:39 No.103891072
i just wanted virtual waifus. is the next immigration crisis really going to be gamers and gooners?
Anonymous 01/14/25(Tue)10:16:52 No.103891096
>>103891045
fear mongering
fear mongering
Anonymous 01/14/25(Tue)10:20:32 No.103891143
>>103891021
troonfy is low level garbage just like trani
troonfy is low level garbage just like trani
Anonymous 01/14/25(Tue)10:21:27 No.103891150
>>103891021
>where is this vitrol when other avatars start batch spamming shit in here? he is actually posting about tech, unlike the usual slop batch spammers.
this, only the slop spammers deserve to be insulted this way
>where is this vitrol when other avatars start batch spamming shit in here? he is actually posting about tech, unlike the usual slop batch spammers.
this, only the slop spammers deserve to be insulted this way
Anonymous 01/14/25(Tue)10:25:10 No.103891186
>>103886319
orig link to upper right plz
orig link to upper right plz
Anonymous 01/14/25(Tue)10:35:09 No.103891285
>>103890558
would intel gpus count? if they released a b580 equivalent for example
would intel gpus count? if they released a b580 equivalent for example
Anonymous 01/14/25(Tue)10:36:36 No.103891296
>>103891021
>start ui war
>harass other devs with trani
>had sex with ani
>only dropped his faggot superiority complex after getting tard wranglers
>turns out to be a low functioning autist
>start ui war
>harass other devs with trani
>had sex with ani
>only dropped his faggot superiority complex after getting tard wranglers
>turns out to be a low functioning autist
Anonymous 01/14/25(Tue)10:37:49 No.103891305
Anonymous 01/14/25(Tue)10:41:21 No.103891344
Anonymous 01/14/25(Tue)10:43:40 No.103891373
>>103891296
What is this phenotype?
What is this phenotype?
Anonymous 01/14/25(Tue)10:45:32 No.103891396
>>103888984
Inpaint the sword anon
Inpaint the sword anon
Anonymous 01/14/25(Tue)10:49:45 No.103891435
>>103891396
I actually like the weird, uneven design. It was "normal" before the schizo upscale.
I actually like the weird, uneven design. It was "normal" before the schizo upscale.
Anonymous 01/14/25(Tue)10:54:10 No.103891495
Anonymous 01/14/25(Tue)10:56:19 No.103891530
>>103891296
>start ui war
I don't think you can start a war by simply existing. Competition is competition.
>harass other devs with trani
I would like to know what "harass" means here but I'll give you the benefit of the doubt here and give that to you since I know we'll argue semantics all day.
>had sex with ani
This is pure slandar and has no proof up until today
>only dropped his faggot superiority complex after getting tard wranglers
>turns out to be a low functioning autist
Those two go hand in hand though for autists that need to perform? This isn't a fault here.
You don't have to worship the guy but you don't have to veer the goddamn thread off track and complain every single time he appears. He clearly has faults but you shouldn't need to make up half of the grievances against him.
>start ui war
I don't think you can start a war by simply existing. Competition is competition.
>harass other devs with trani
I would like to know what "harass" means here but I'll give you the benefit of the doubt here and give that to you since I know we'll argue semantics all day.
>had sex with ani
This is pure slandar and has no proof up until today
>only dropped his faggot superiority complex after getting tard wranglers
>turns out to be a low functioning autist
Those two go hand in hand though for autists that need to perform? This isn't a fault here.
You don't have to worship the guy but you don't have to veer the goddamn thread off track and complain every single time he appears. He clearly has faults but you shouldn't need to make up half of the grievances against him.
Anonymous 01/14/25(Tue)10:58:30 No.103891561
>>103890882
I too hate avatarfagging, but, just, idk, don't make anyone lonely, have more frens to discuss AI with
I too hate avatarfagging, but, just, idk, don't make anyone lonely, have more frens to discuss AI with
Anonymous 01/14/25(Tue)10:58:35 No.103891564
>>103891396
go back to your containment thread, you fucking piece of shit
go back to your containment thread, you fucking piece of shit
Anonymous 01/14/25(Tue)11:01:46 No.103891603
>>103891435
The uneven handles were cool, I'll give you that
>>103891564
I think you have me confused with someone else
The uneven handles were cool, I'll give you that
>>103891564
I think you have me confused with someone else
Anonymous 01/14/25(Tue)11:02:15 No.103891609
>>103891530
>meets up IRL and goes traveling with a schizo gooner
>that post images of their avatars having sex
Cope is strong with this one
>meets up IRL and goes traveling with a schizo gooner
>that post images of their avatars having sex
Cope is strong with this one
Anonymous 01/14/25(Tue)11:09:46 No.103891686
>>103891530
he insulted auto and auto's ui every single thread on sdg when he started the troonfyui, you might be a newfag, I'd suggest you to lurk moar
he insulted auto and auto's ui every single thread on sdg when he started the troonfyui, you might be a newfag, I'd suggest you to lurk moar
Anonymous 01/14/25(Tue)11:12:48 No.103891716
Anonymous 01/14/25(Tue)11:13:46 No.103891732
>>103891686
Then he attacked the forge Dev and trani tried taking the repo down
Then he attacked the forge Dev and trani tried taking the repo down
Anonymous 01/14/25(Tue)11:14:33 No.103891743
i2v in
>two
>more
>weeks
>two
>more
>weeks
Anonymous 01/14/25(Tue)11:15:05 No.103891747
Idk what anon is on about I cut my dick off and everythings chill
Anonymous 01/14/25(Tue)11:17:45 No.103891775
Anonymous 01/14/25(Tue)11:18:02 No.103891777
Anonymous 01/14/25(Tue)11:19:32 No.103891795
Have there been any breakthroughs in terms of 2D animations or consistent character sprite sheet creation? I remember when the best way was to screenshot a bunch of blender models doing various poses and inpaint them but this sounds like a giant pain in the ass.
Anonymous 01/14/25(Tue)11:19:55 No.103891796
Anonymous 01/14/25(Tue)11:22:01 No.103891820
>>103891296
>comfy is cute when he's sleeping
>looks like that
30 year old man sucking sperg dick for a job.
Do better Julien!
>comfy is cute when he's sleeping
>looks like that
30 year old man sucking sperg dick for a job.
Do better Julien!
Anonymous 01/14/25(Tue)11:24:00 No.103891837
Anonymous 01/14/25(Tue)11:24:54 No.103891844
>>103891296
I'm just happy he's not a tranny. :D
I'm just happy he's not a tranny. :D
Anonymous 01/14/25(Tue)11:26:12 No.103891857
>hunyuan workflow blurry and blank fix
does anyone still have this workflow or guide? why the fuck, they deleted this guide
does anyone still have this workflow or guide? why the fuck, they deleted this guide
Anonymous 01/14/25(Tue)11:34:54 No.103891936
Anonymous 01/14/25(Tue)11:35:11 No.103891941
>>103891857
Yet
Yet
Anonymous 01/14/25(Tue)11:36:47 No.103891951
>>103891837
Thanks to that anon that inspired this theme, this is cool as fuck
Thanks to that anon that inspired this theme, this is cool as fuck
Anonymous 01/14/25(Tue)11:37:37 No.103891958
>>103886725
And we had FF7, FF8, FF9, FF10 almost one year after each other.
Then it went hill after that, all we had was 15 plus years of FF13 and Versus/15.....
And we had FF7, FF8, FF9, FF10 almost one year after each other.
Then it went hill after that, all we had was 15 plus years of FF13 and Versus/15.....
Anonymous 01/14/25(Tue)11:38:38 No.103891968
>>103891958
downhill*
downhill*
Anonymous 01/14/25(Tue)11:40:16 No.103891988
Anonymous 01/14/25(Tue)11:57:00 No.103892149
Anonymous 01/14/25(Tue)11:57:45 No.103892158
>schizo seething again about ani
Anonymous 01/14/25(Tue)11:58:48 No.103892170
>>103891795
>Have there been any breakthroughs in terms of 2D animations or consistent character sprite sheet creation?
I'm so sorry but no
>Have there been any breakthroughs in terms of 2D animations or consistent character sprite sheet creation?
I'm so sorry but no
Anonymous 01/14/25(Tue)12:01:14 No.103892203
Anonymous 01/14/25(Tue)12:04:35 No.103892238
Anonymous 01/14/25(Tue)12:04:42 No.103892240
>>103892158
Hi trani
Hi trani
Anonymous 01/14/25(Tue)12:15:22 No.103892335
Is there anything other than img2vid to be hopeful for or is it over
Anonymous 01/14/25(Tue)12:20:14 No.103892387
Turns out flux does better with a paragraph prompt instead of "property, property, property, property"
Anonymous 01/14/25(Tue)12:21:12 No.103892397
Anonymous 01/14/25(Tue)12:28:46 No.103892471
>>103888069
>If your art was on danbooru or any other dataset source your handle is now a means to steal your work.
Your fault for uplooding your work
>steal your work.
kek be serious now anon
>If your art was on danbooru or any other dataset source your handle is now a means to steal your work.
Your fault for uplooding your work
>steal your work.
kek be serious now anon
Anonymous 01/14/25(Tue)12:30:28 No.103892487
Anonymous 01/14/25(Tue)12:46:05 No.103892679
>>103892387
Not only paragraphs but purple prose. Make sure to include some bullshit about the "feeling" or "mood".
Not only paragraphs but purple prose. Make sure to include some bullshit about the "feeling" or "mood".
Anonymous 01/14/25(Tue)12:54:43 No.103892768
Remember this is who you're dealing with
https://www.youtube.com/watch?v=NyUqhbPD2tc
https://www.youtube.com/watch?v=NyU
Anonymous 01/14/25(Tue)12:56:25 No.103892789
>>103892768
Just because I experience voices in my head, doesn't make me schizophrenic.
Just because I experience voices in my head, doesn't make me schizophrenic.
Anonymous 01/14/25(Tue)12:57:43 No.103892810
>>103892789
Poison!
Poison!
Anonymous 01/14/25(Tue)12:58:41 No.103892823
Anonymous 01/14/25(Tue)13:01:36 No.103892867
>>103892823
We're not your other voices trani
We're not your other voices trani
Anonymous 01/14/25(Tue)13:07:49 No.103892964
>>103891373
Successful person.
Successful person.
Anonymous 01/14/25(Tue)13:22:35 No.103893162
>>103891686
Where would this even be posted because certainly not here? And I am not hanging out in a ERP dungeon Discord and backtracking to find out.
>>103891686
I wasn't there for that but since you told me I went back to the archives to search. In my opinion reading the sales pitches, he was selling the project and pushing it too strongly even if he is right about new features and code quality so at least with this, okay, you are right.
>>103891732
I followed that issue at the time. Mis-attributation and not using the correct licenses is a serious thing in open source and I can read code so Comfy's accusations about that isn't unfounded. It took several people who was not lllyasviel but who contributed to the repository to comment and clarify on the issue and getting the licenses correctly done for attribution before the issue was dropped. This was justified.
Where would this even be posted because certainly not here? And I am not hanging out in a ERP dungeon Discord and backtracking to find out.
>>103891686
I wasn't there for that but since you told me I went back to the archives to search. In my opinion reading the sales pitches, he was selling the project and pushing it too strongly even if he is right about new features and code quality so at least with this, okay, you are right.
>>103891732
I followed that issue at the time. Mis-attributation and not using the correct licenses is a serious thing in open source and I can read code so Comfy's accusations about that isn't unfounded. It took several people who was not lllyasviel but who contributed to the repository to comment and clarify on the issue and getting the licenses correctly done for attribution before the issue was dropped. This was justified.
Anonymous 01/14/25(Tue)13:26:38 No.103893218
Anonymous 01/14/25(Tue)13:28:10 No.103893239
>>103893162
Working overtime I see
Working overtime I see
Anonymous 01/14/25(Tue)13:28:46 No.103893248
>local drama general
Anonymous 01/14/25(Tue)13:30:54 No.103893278
>>103893248
Literally
Literally
Anonymous 01/14/25(Tue)13:31:22 No.103893288
>>103893248
Same old, same old. It's not going to change until we have a real drought in models and that isn't stopping until the AI bubble bursts.
Same old, same old. It's not going to change until we have a real drought in models and that isn't stopping until the AI bubble bursts.
Anonymous 01/14/25(Tue)13:31:52 No.103893295
>>103893278
shut up bitch you contribute to it too
shut up bitch you contribute to it too
Anonymous 01/14/25(Tue)13:33:35 No.103893315
>>103893295
No I don't
No I don't
Anonymous 01/14/25(Tue)13:33:58 No.103893321
Comfy and Trani are some sassy faggots
Anonymous 01/14/25(Tue)13:34:05 No.103893323
>>103893315
we all know you do but dont worry im already bored of you
we all know you do but dont worry im already bored of you
Anonymous 01/14/25(Tue)13:34:49 No.103893332
>>103893323
Keep telling yourself that
Keep telling yourself that
Anonymous 01/14/25(Tue)13:35:15 No.103893339
Anonymous 01/14/25(Tue)13:35:47 No.103893351
MiniMax is now open source
https://www.minimaxi.com/en/news/minimax-01-series-2
https://huggingface.co/MiniMaxAI/MiniMax-Text-01
https://huggingface.co/MiniMaxAI/MiniMax-VL-01
https://github.com/MiniMax-AI/MiniMax-01
https://filecdn.minimax.chat/_Arxiv_MiniMax_01_Report.pdf
https://www.hailuo.ai/
https://www.minimaxi.com/en/news/mi
https://huggingface.co/MiniMaxAI/Mi
https://huggingface.co/MiniMaxAI/Mi
https://github.com/MiniMax-AI/MiniM
https://filecdn.minimax.chat/_Arxiv
https://www.hailuo.ai/
Anonymous 01/14/25(Tue)13:37:12 No.103893375
>>103893339
Are you bored or something?
Are you bored or something?
Anonymous 01/14/25(Tue)13:37:20 No.103893378
>>103893351
Nice
Nice
Anonymous 01/14/25(Tue)13:39:10 No.103893393
>>103888150
>>103888173
>>103888223
1.4 is cool in the sense that it's different. You need to take off your nostalgia goggles.
>>103888173
>>103888223
1.4 is cool in the sense that it's different. You need to take off your nostalgia goggles.
Anonymous 01/14/25(Tue)13:40:13 No.103893408
>>103893351
You sure you got the right thread for this? There is no diffusion model in here for video or images. Should go in >>>/g/lmg/ but thanks for sharing
You sure you got the right thread for this? There is no diffusion model in here for video or images. Should go in >>>/g/lmg/ but thanks for sharing
Anonymous 01/14/25(Tue)13:40:35 No.103893413
>>103893351
nearly a goddamn terrabyte
nearly a goddamn terrabyte
Anonymous 01/14/25(Tue)13:40:58 No.103893414
>>103893413
dont tell me youre a poorfag
dont tell me youre a poorfag
Anonymous 01/14/25(Tue)13:41:36 No.103893421
>>103893414
sure looks that way
sure looks that way
Anonymous 01/14/25(Tue)13:42:47 No.103893427
>>103893413
wait for quants I guess
wait for quants I guess
Anonymous 01/14/25(Tue)13:43:04 No.103893430
>>103893278
adetailer anon here. maybe add a sliver of weight? like a pinch, y'know
adetailer anon here. maybe add a sliver of weight? like a pinch, y'know
Anonymous 01/14/25(Tue)13:43:40 No.103893437
>prompt non-functioning lights
>the lights function
>the lights function
Anonymous 01/14/25(Tue)13:45:51 No.103893454
Anonymous 01/14/25(Tue)13:46:03 No.103893456
>>103893408
I guess the VL model is of some interest for captioning and etc. but it's not really improving that much on InternVL2.5 or Qwen2 and it is gigantic. The only upside is being able to ingest resolutions of images up to 2016×2016 but god knows how long the inference time will be on that on anything that isn't a B100/200 GPU.
I guess the VL model is of some interest for captioning and etc. but it's not really improving that much on InternVL2.5 or Qwen2 and it is gigantic. The only upside is being able to ingest resolutions of images up to 2016×2016 but god knows how long the inference time will be on that on anything that isn't a B100/200 GPU.
Anonymous 01/14/25(Tue)13:48:02 No.103893477
>>103893454
I'm intrigued
I'm intrigued
Anonymous 01/14/25(Tue)13:50:47 No.103893500
>>103893430
Okay find I'll try it
Okay find I'll try it
Anonymous 01/14/25(Tue)13:51:04 No.103893503
Anonymous 01/14/25(Tue)16:28:15 No.103895323
>>103886442
its nothing to do with interest in tech, its 4chan in general. most boards are dying. no one wants to email verify or wait 15 minutes to post not to mention the irritating captchas
its nothing to do with interest in tech, its 4chan in general. most boards are dying. no one wants to email verify or wait 15 minutes to post not to mention the irritating captchas