Home Creators Posts Import Register Favorites Logout

Downloads

Content

Friends, this one is for you. 💛

Minimax H3 is the king of local AI video right now. Everything below is what you need to download and where to put it — free, public, no tier gate. No one should be locked out of the fun.

Official Minimax Prompt Guide:https://huggingface.co/MiniMaxAI/MiniMax-H3/blob/main/docs/VIDEO_PROMPT_WRITING_GUIDE_base_en.md

Take your time with it. Grab a coffee. I'll walk beside you.

If you are a paid supporter you can use my installer Software that will handle downloading of models, workflows and custom nodes for you. It even installs Sage Attention. Here is the Link to the latest Version:

https://www.patreon.com/AIKnowledgeCentral/posts/minimax-h3-and-2-167237656

⚡ Performance Extensions

⚠ Important: Triton and Sage Attention require specific installation steps. Don't rush these two — they're the only fiddly part, and I've made you a guide so you never have to guess.

🏎 Sage Attention Install Guide: https://youtu.be/mMwyKNh3TqY

✔ Triton — Required for GPU kernel optimization

✔ Sage Attention — Optimized attention mechanism

📦 Git Repositories You Need

Main Repository (ComfyUI):

https://github.com/comfyanonymous/ComfyUI

Additional Repositories

ComfyUI-VideoHelperSuite — https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite

ComfyUI-IntoTheLatent-Utils — https://github.com/Into-The-Latent/ComfyUI-IntoTheLatent-Utils

ComfyUI-Crystools — https://github.com/crystian/ComfyUI-Crystools

ComfyUI-Custom-Scripts — https://github.com/pythongosssss/ComfyUI-Custom-Scripts

ComfyUI-KJNodes — https://github.com/kijai/ComfyUI-KJNodes

ComfyUI_essentials — https://github.com/cubiq/ComfyUI_essentials

ComfyUI-Manager — https://github.com/Comfy-Org/ComfyUI-Manager

rgthree-comfy — https://github.com/rgthree/rgthree-comfy

cg-use-everywhere — https://github.com/chrisgoringe/cg-use-everywhere

gguf — https://github.com/calcuis/gguf

🎨 Models You Need

🖼 Diffusion Models

(Tip: Place in your models\diffusion_models folder)

📥 MiniMax-H3-Ref2VA-Pruned-GGUF

Pick the one that fits your VRAM — and please don't feel small if you're on the 12GB line. A modest card and a curious mind beat an expensive card every single time.

12GB https://huggingface.co/Abiray/MiniMax-H3-Pruned-GGUF/resolve/main/MiniMax-H3-Ref2VA-Pruned-Q3_K_M.gguf?download=true

16GB https://huggingface.co/Abiray/MiniMax-H3-Pruned-GGUF/resolve/main/MiniMax-H3-Ref2VA-Pruned-Q4_K_M.gguf?download=true

24GB https://huggingface.co/Abiray/MiniMax-H3-Pruned-GGUF/resolve/main/MiniMax-H3-Ref2VA-Pruned-Q6_K.gguf?download=true

32GB https://huggingface.co/Abiray/MiniMax-H3-Pruned-GGUF/resolve/main/MiniMax-H3-Ref2VA-Pruned-Q8_0.gguf?download=true

📥 MiniMax-H3-FL2VA-Pruned-GGUF

Same idea here — one file, the one that matches your VRAM:

12GB https://huggingface.co/Abiray/MiniMax-H3-Pruned-GGUF/resolve/main/MiniMax-H3-FL2VA-Pruned-Q3_K_M.gguf?download=true

16GB https://huggingface.co/Abiray/MiniMax-H3-Pruned-GGUF/resolve/main/MiniMax-H3-FL2VA-Pruned-Q5_K_M.gguf?download=true

24GB https://huggingface.co/Abiray/MiniMax-H3-Pruned-GGUF/resolve/main/MiniMax-H3-FL2VA-Pruned-Q6_K.gguf?download=true

32GB https://huggingface.co/Abiray/MiniMax-H3-Pruned-GGUF/resolve/main/MiniMax-H3-FL2VA-Pruned-Q8_0.gguf?download=true

📝 Text Encoders

(Tip: Place in your models\text_encoders folder)

📥 qwen3vl_32b_minimax_h3_nvfp4_awq

https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/text_encoders/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors

🎭 VAEs

(Tip: Place in your models\vae folder)

📥 minimax_h3_video_vae_fp16

https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_video_vae_fp16.safetensors

📥 minimax_h3_audio_vae_fp32

https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_audio_vae_fp32.safetensors

🔧 LoRAs

(Tip: Place in your models\loras folder)

📥 Minimax-h3-Turbo-8-Step

https://huggingface.co/lightx2v/Minimax-h3-Turbo/resolve/main/minimax_h3_fl2v_turbo_8step_v1.0_comfyui_bf16.safetensors

That's the whole shopping list. Two extensions, a handful of repos, and the models above — and then your own machine can make video with voices, sound and up to 9 reference inputs. Sit with that for a second. A few years ago that was science fiction, and now it lives on your desk.

Thank you for being here. You're the reason these guides stay free and public, so that the person who just installed ComfyUI last week can follow along too. Go make something you love — I can't wait to see it. 🙏

Best

Chris

Files

Run the King of Local AI Video Free on Your Own PC (Full ComfyUI Setup)

Minimax H3 is the king of local AI video: up to 15 seconds per clip, native speech and sound, knowledge of a lot of TV shows, and up to 9 reference inputs. It runs for free on your own machine, and this video shows you exactly how. We walk through the ComfyUI workflows step by step, starting with text to video — the model loader and why the FL2VA model matters, the speed LoRA, Sage Attention, the video length and resolution settings, and how to prompt using the structure the Minimax creators recommend, including multi-speaker dialogue in different languages. From there we move to image to video and to first frame / last frame transitions, with a look at how the image size node keeps your aspect ratio intact instead of cropping or morphing. You also get a quick peek at multi-reference generation with the Ref2VA model and up to eight reference images, plus a bonus workflow that doubles your framerate through frame interpolation for smoother shots. This is not a beginner video, so if you have never used ComfyUI, start with the beginner video linked in the video first. The free setup guide and all the workflows are on my Patreon. https://huggingface.co/MiniMaxAI/MiniMax-H3/blob/main/docs/VIDEO_PROMPT_WRITING_GUIDE_base_en.md 📩 Stay up to date with Local AI & Open Source AI https://intothelatent.com/newsletter 🎁 Free Downloads & Resources https://intothelatent.com ❤️ Support the channel & get exclusive workflows, guides and early access https://patreon.com/intothelatent 📅 Need help getting started? Book a personal 2-hour 1-on-1 AI coaching session (paid offer) where we'll solve your specific workflow, ComfyUI, Stable Diffusion, or local AI questions together. https://koalendar.com/e/1hr-1-on-1-stable-diffusion Chapters 00:00 Intro 01:00 Text 2 Video 08:30 Image 2 Video 11:35 First Frame Last Frame 12:55 Frame Interpolation 13:45 Reference 2 Video

Comments

Oobawooba

im really stupid on comfyui and have been tinkering around making my own stuff with workflows from others but i really like your shopping list and workflows. im really into taking drawn photos to realist then to video and so far thats my limits on making my own stuff. anyways thanks for making it easy.