cocktail peanut's banner
cocktail peanut's profile picture

cocktail peanut

@cocktailpeanut34,316 subscribers

Creator https://t.co/Wgwpl7lJ4I, run AI apps on your local PC, for free.

Shorts

Run MiniMax H3 Video Generation on 5GB VRAM PC! The great wizard deepbeepmeep has dropped low VRAM support for MiniMax H3 on WanGP. Super optimized as usual! - 5-6GB of VRAM only for 5s (124 frames) - 8-9GB of VRAM for 15s at 832x480 1-click launch on pinokio!

Run MiniMax H3 Video Generation on 5GB VRAM PC! The great wizard deepbeepmeep has dropped low VRAM support for MiniMax H3 on WanGP. Super optimized as usual! - 5-6GB of VRAM only for 5s (124 frames) - 8-9GB of VRAM for 15s at 832x480 1-click launch on pinokio!

127,944 views

WanGP's latest MiniMax H3 update is huge: - Way faster Generation - 2X faster Ref2VA - Lower-RAM & new accelerators - Long videos with smoother visual/audio transitions. - Huge speed gains with Sol-Attention, First Block Cache, etc. And Pinokio launcher is optimized for it 👇

WanGP's latest MiniMax H3 update is huge: - Way faster Generation - 2X faster Ref2VA - Lower-RAM & new accelerators - Long videos with smoother visual/audio transitions. - Huge speed gains with Sol-Attention, First Block Cache, etc. And Pinokio launcher is optimized for it 👇

58,539 views

Prompt adherence in MiniMax H3 is crazy, missed no detail and even the timing is perfect. I take back my earlier comment about text-to-video being shitty. This was pure text-to-video with no start image. "Inside Kowloon Walled City, late 1980s, ultra-realistic handheld one-take. A restaurant worker hurries through impossibly narrow, crowded corridors carrying a huge bowl of boiling soup. Residents press themselves against the damp walls to let him pass. He ducks beneath dripping pipes, squeezes past hanging laundry, and leaps over children playing mahjong on the floor. The maze feels claustrophobic, noisy, and alive. He reaches a cramped apartment doorway and hands over the soup just as the lights suddenly cut out, plunging everything into darkness. Oppressive realism, fluorescent lighting, wet concrete, exposed wires, Cantonese shouting, clattering dishes, echoing footsteps, tense cinematic pacing. No cyberpunk, no stylization, documentary-like realism."

Prompt adherence in MiniMax H3 is crazy, missed no detail and even the timing is perfect. I take back my earlier comment about text-to-video being shitty. This was pure text-to-video with no start image. "Inside Kowloon Walled City, late 1980s, ultra-realistic handheld one-take. A restaurant worker hurries through impossibly narrow, crowded corridors carrying a huge bowl of boiling soup. Residents press themselves against the damp walls to let him pass. He ducks beneath dripping pipes, squeezes past hanging laundry, and leaps over children playing mahjong on the floor. The maze feels claustrophobic, noisy, and alive. He reaches a cramped apartment doorway and hands over the soup just as the lights suddenly cut out, plunging everything into darkness. Oppressive realism, fluorescent lighting, wet concrete, exposed wires, Cantonese shouting, clattering dishes, echoing footsteps, tense cinematic pacing. No cyberpunk, no stylization, documentary-like realism."

31,141 views

One gripe with MiniMax H3: unless I specify exactly what each character says, the dialogue sometimes devolves into gibberish. Always write out the speech. Prompt: "A gritty, photorealistic nighttime crime scene outside a convenience store, shot like a serious detective TV drama. Yellow police tape, flashing patrol lights, forensic markers, and two stern live-action detectives carefully examining evidence. Assisting them are Bugs Bunny and Daffy Duck, fully recognizable in classic Looney Tunes style - bright, clean 2D cartoon characters composited into the real world. Bugs is unusually calm and professional, kneeling beside a clue with a cartoon magnifying glass, while Daffy nervously points at random evidence and overreacts. The detectives remain completely serious and treat them like normal forensic partners. At the climax, Bugs holds the magnifying glass over a surface and it reveals a photorealistic fingerprint, while Daffy dramatically gasps and points. Strong contrast between cartoon and live-action, believable shadows and interaction, cinematic camera movement, tense crime-drama atmosphere played completely straight."

One gripe with MiniMax H3: unless I specify exactly what each character says, the dialogue sometimes devolves into gibberish. Always write out the speech. Prompt: "A gritty, photorealistic nighttime crime scene outside a convenience store, shot like a serious detective TV drama. Yellow police tape, flashing patrol lights, forensic markers, and two stern live-action detectives carefully examining evidence. Assisting them are Bugs Bunny and Daffy Duck, fully recognizable in classic Looney Tunes style - bright, clean 2D cartoon characters composited into the real world. Bugs is unusually calm and professional, kneeling beside a clue with a cartoon magnifying glass, while Daffy nervously points at random evidence and overreacts. The detectives remain completely serious and treat them like normal forensic partners. At the climax, Bugs holds the magnifying glass over a surface and it reveals a photorealistic fingerprint, while Daffy dramatically gasps and points. Strong contrast between cartoon and live-action, believable shadows and interaction, cinematic camera movement, tense crime-drama atmosphere played completely straight."

20,232 views

Actually scratch that----You just need 6GB VRAM!!!!!

Actually scratch that----You just need 6GB VRAM!!!!!

125,909 views

6-step MiniMax H3 Turbo Generation with Maestro! Blaine Brown has implemented MiniMax H3 Turbo LoRA into Maestro, and now you can generate videos faster! You will need to use the full model (not pruned), select the turbo mode, and finally, if you have low VRAM (not 24GB+), recommend switching to 480p instead of the default 768p

6-step MiniMax H3 Turbo Generation with Maestro! Blaine Brown has implemented MiniMax H3 Turbo LoRA into Maestro, and now you can generate videos faster! You will need to use the full model (not pruned), select the turbo mode, and finally, if you have low VRAM (not 24GB+), recommend switching to 480p instead of the default 768p

18,088 views

LTX-2 is THE open source video AI moment I've been waiting for--the first to generates high quality video AND is FAST!! - 4090: Generate 20s at 720p in 2 minutes - A4500 (3070): Generate 10s at 480p in 3 minutes 1-click install on Pinokio, do it now!

LTX-2 is THE open source video AI moment I've been waiting for--the first to generates high quality video AND is FAST!! - 4090: Generate 20s at 720p in 2 minutes - A4500 (3070): Generate 10s at 480p in 3 minutes 1-click install on Pinokio, do it now!

116,072 views

Generate FULLY CONTROLLABLE 3D assets from a SINGLE image, locally on your PC. Made a 1-click launcher for the official Anigen Gradio app, and a dedicated viewer. Crazy this is now possible. What you're seeing here came from one image. Requires: NVIDIA GPU 6GB VRAM

Generate FULLY CONTROLLABLE 3D assets from a SINGLE image, locally on your PC. Made a 1-click launcher for the official Anigen Gradio app, and a dedicated viewer. Crazy this is now possible. What you're seeing here came from one image. Requires: NVIDIA GPU 6GB VRAM

68,200 views

FramePack: Generate Video Forever [NVIDIA ONLY] The creator of ControlNet just dropped a new approach to generating videos in a PROGRESSIVE way--you can generate loooong videos with LOW VRAM (just 6GB) SUP3RMASS1VE wrote a 1-click launcher to run the Gradio app on your PC!

FramePack: Generate Video Forever [NVIDIA ONLY] The creator of ControlNet just dropped a new approach to generating videos in a PROGRESSIVE way--you can generate loooong videos with LOW VRAM (just 6GB) SUP3RMASS1VE wrote a 1-click launcher to run the Gradio app on your PC!

202,368 views

Elevenlabs-level TTS on your laptop. I'm always skeptical about new AI models hyped up sounding too good to be true. But this...is crazy good. And now, you can run the gradio app by tyler on your laptop with 1 click. Meet Elon Musk, from Silicon Valley.

Elevenlabs-level TTS on your laptop. I'm always skeptical about new AI models hyped up sounding too good to be true. But this...is crazy good. And now, you can run the gradio app by tyler on your laptop with 1 click. Meet Elon Musk, from Silicon Valley.

225,020 views

Even faster, even lower-VRAM MiniMax H3 on Maestro! 1. H3 Turbo now works with the pruned 20B model, fits into low VRAM (Recommended) 2. Turbo mode now starts at 6 steps 3. First Block Cache optimization option (Even more optimization) Still high quality. Example:

Even faster, even lower-VRAM MiniMax H3 on Maestro! 1. H3 Turbo now works with the pruned 20B model, fits into low VRAM (Recommended) 2. Turbo mode now starts at 6 steps 3. First Block Cache optimization option (Even more optimization) Still high quality. Example:

12,227 views

Generate Long Videos with Wan even on Low VRAM machines! [NVIDIA ONLY] With this super-optimized gradio app from deepbeepmeep you can generate up to 12 second videos with Wan. And even generate on low VRAM machines! - Generate up to 12 sec videos - Run on 5GB~ VRAM (Read the full thread)

Generate Long Videos with Wan even on Low VRAM machines! [NVIDIA ONLY] With this super-optimized gradio app from deepbeepmeep you can generate up to 12 second videos with Wan. And even generate on low VRAM machines! - Generate up to 12 sec videos - Run on 5GB~ VRAM (Read the full thread)

135,961 views

Run Ideogram 4 with 6GB VRAM Locally WanGP now supports Ideogram 4, just need around 6GB VRAM to run. The JSON based prompting lets you describe the layout of the image PRECISELY, and you can use the built-in prompt helper to build the prompt.

Run Ideogram 4 with 6GB VRAM Locally WanGP now supports Ideogram 4, just need around 6GB VRAM to run. The JSON based prompting lets you describe the layout of the image PRECISELY, and you can use the built-in prompt helper to build the prompt.

24,794 views

Qwen3-TTS Web UI on Mac! Qwen3-TTS web UI now works on Macs too, thanks to contribution from Pierre Bruno, who is also our Pinokio community mod. Thank you! Just install fresh and it should work, I've confirmed this working on my Macbook M1 Max here

Qwen3-TTS Web UI on Mac! Qwen3-TTS web UI now works on Macs too, thanks to contribution from Pierre Bruno, who is also our Pinokio community mod. Thank you! Just install fresh and it should work, I've confirmed this working on my Macbook M1 Max here

50,663 views

Wow Qwen3-TTS is crazy good and so interesting. Listen to this guy go through the whole emotional journey -- starting confused, having an epiphany, and getting cocky at the end.

Wow Qwen3-TTS is crazy good and so interesting. Listen to this guy go through the whole emotional journey -- starting confused, having an epiphany, and getting cocky at the end.

45,835 views

Clarity Refiners UI The best open source image upscaler/enhancer i've tried, powered by Finegrain's Refiners project, which includes an implementation of Clarity upscaler by philz1337x - runs on Gradio. FAST on ALL platforms (Windows, Mac, Linux) Only 8GB VRAM!

Clarity Refiners UI The best open source image upscaler/enhancer i've tried, powered by Finegrain's Refiners project, which includes an implementation of Clarity upscaler by philz1337x - runs on Gradio. FAST on ALL platforms (Windows, Mac, Linux) Only 8GB VRAM!

83,655 views

Run Hunyuan3D TURBO Locally (NVIDIA GPU) These AI models are super fast indeed! Depending on machine (details in thread) it takes: - Mesh Generation: just 1~5sec - Mesh + Texture Generation: 30~100sec And now, the Hunyuan3D-GP gradio app supports them. Run on your PC!

Run Hunyuan3D TURBO Locally (NVIDIA GPU) These AI models are super fast indeed! Depending on machine (details in thread) it takes: - Mesh Generation: just 1~5sec - Mesh + Texture Generation: 30~100sec And now, the Hunyuan3D-GP gradio app supports them. Run on your PC!

61,163 views

[Audio On] This cafe does not exist. Generated 100% with AI. - Video: Framepack - Soundscape: MMAudio

[Audio On] This cafe does not exist. Generated 100% with AI. - Video: Framepack - Soundscape: MMAudio

45,084 views

His name is Sam. He is a 100% synthetic human I created from the Kangaroo.

His name is Sam. He is a 100% synthetic human I created from the Kangaroo.

21,036 views

Videos

cocktailpeanut's profile picture

POV: Claude users

cocktail peanut

1,766,821 views • 1 month ago

cocktailpeanut's profile picture

Introducing Pinokio 8.

cocktail peanut

70,456 views • 1 month ago