Open Source Must Win We do open weight video model optimization, customization and deployment Years of deep collab with vLLM and FastVideo
Bay Area, CA
Joined June 2023
- Tweets41
- Following18
- Followers298
- Likes194
It's made with our open weight released FastH3-Omni and our creative agent, deployable on consumer GPU too.
Learn more in detail and talk to us if you're a player who need to scale in volume with your data customization ;)
nuvalab.ai/fasth3/
FastH3-Omni, the accelerated Minimax H3 Omni Ref is in preview TODAY with @nuvalab and @reactorworld , in close collaboration with FastVideo team from @haoailab
It's the first full version with omni ref support along the lines of FastH3 v1 and v2 preview that only covered text to video.
We've already shipped the omni preview with top ads agency, animation studio and entertainment industry leaders for dedicated deployment as well as self improvement based on their proprietary data. Check out nuvalab.ai to get in touch with us.
Final model will be fully open sourced on Oct 20th along with consumer GPU support, as our core mission of continuous support of open source AI ecosystem.
Check our website to fill out info if you want your own real "unlimited" video gen on your data ;)
Already rolling out to fitting users.
(1/13) Introducing FastVideo UniServe Engine, the fastest serving engine for @MiniMax_AI FastH3 🚀
- 🏭 Production ready: UniServe is serving production API traffic on @reactorworld and @nuvalab on day-0!
- ⚡️ Fastest performance: up to 1.8x lower latency and 1.45x higher throughput
- 💻 Production Hardware: Supports NVIDIA Blackwell (B200/B300/RTX 6k pro) and Hopper GPUs
Open weight is the way 🧑🍳🤌
Open weights. Shared progress. MiniMax H3 is moving fast.
We built H3 for video generation with native stereo audio and multimodal reference control. The open-source community is making that capability faster, more accessible, and easier to build on.
Recent highlights:
• FastH3 — FastVideo, Nuva Lab and NVIDIA: 4-step distillation, now running on DGX Spark and Apple Silicon.
• Sol-H3 — NVIDIA’s SANA team: 15 seconds of 768p video + audio in 6.6 seconds on 8×B300, in the team’s warm-inference benchmark.*
• VDN — Haocheng Xi and the OpenVDN team: rethinking attention for faster H3 inference, with weights, training and inference code released.
• PDD — NVIDIA’s distillation method, brought to H3 by Alibaba PAI as 8-step Acc-LoRAs, now supported in ComfyUI.
• LightX2V — 4- and 8-step Turbo LoRAs, with workflows for text, image and reference-conditioned video + audio.
Behind every release are people training, optimizing, quantizing, testing and sharing. Special thanks to:
@haoailab @nuvalab @NVIDIAAI @xieenze_jr @HaochengXiUCB @ArashVahdat @julberner @LightX2V @ComfyUI
And to the individual contributors pushing the work forward:
@haozhangml @cxlcl1 @lawrence_cjs @yitongli165665 @haopengl33 @songhan_mit @shanasaimoe
Thank you for building with H3 and helping make it faster, more accessible, and more useful for the community. Powerful models go further when we build together. Keep pushing H3. Excited to see what comes next. 🚀
Explore the ecosystem: github.com/MiniMax-AI/awesom…
Open, democratized, personalized video generation running on local device 🥳
(1/7) Last week FastVideo released FastH3. Today we bring it local: @NVIDIA DGX Spark 💚 and @Apple Silicon 🍎through MLX. 8x speedup!
FastH3 now supported on NVIDIA DGX Spark and Apple Silicon! over base @MiniMax_AI H3 for local generation!
This 15 seconds 1344×768 clip was generated on two NVIDIA DGX Sparks. Fully local. @NVIDIARTXSpark @NVIDIAAI
Nuva Lab retweeted
🎬Video generation faster than playback!
🚀MiniMax H3 on vLLM-Omni + FastVideo's FastH3: a complete 10.1s MP4 - video AND synchronized audio - rendered in 8.7s!⚡️
Thanks to @MiniMax_AI for the great Minimax H3 release, the FastVideo team @haoailab for open-sourcing FastH3 and helping on the serving integration, and @NVIDIAAI for the continued sponsorship and joint optimization efforts!
Open source fasth3 live video! we remember playing with vibe direction LTX-2 back in March and May 2026 to get a glimpse of the future haoailab.com/blogs/fastvideo… and open source model kept improving that push new boundaries on quality and efficiency 🥳
Amazing!! @reactorworld has developed a new infinite stream video using FastH3. The community is moving fast!! And it is fully open source
WhatIf you put #FastH3 into the hands of anime lover.. Any #luckystar fans? 😜
Based on @MiniMax_AI H3 and open sourced FastH3 preview from @haoailab
That was way better than what we've expected from the preview test with 1/10 of planned compute and data. WhatIf we scale it up properly 😜
45% preferred FastVideo FastH3 or are neutral compared with @MiniMax_AI H3.
Our open-weight FastH3 achieves 15s generation in 13s and 14x speedup. 🚀
Big shoutout to the folks at @fal and @isidentical for helping us run an eval against H3. As an open source project we always appreciate all the help we can get.
The entire team is quite excited by this result for a preview checkpoint trained on a fraction of our planned total compute and partial data. Excited for what our next versions can do!!
Weights: huggingface.co/FastVideo/Fas…
Blog: haoailab.com/blogs/fasth3-pr…
Thank you for your contribution to the open source H3 ecosystem @fal!
xitter.cf/isidentical/status/209…
A great start with many, many more to come! Omni ref, quality improvements, nvfp4, model customization and deployment, etc 🥳
(1/6) Open Weight @MiniMax_AI FastH3 v1: Generate 15s 768p video in 13s 🚀
- FastVideo collab w/ @nuvalab + FastGen - Up to 14x speedup on @NVIDIAAI Blackwell GPU
- Fully open so community can run and improve the acceleration recipe. The era of open weight video models just started, that calls for post training.
With great community effort such as Minimax H3 Max based on @MiniMax_AI , we have more to share about how post training can improve speed, quality and most importantly, how it’s done, with open weights and recipes.