MiniMax开源视频生成模型H3及社区加速方案
Open weights. Shared progress. MiniMax H3 is moving fast.
H3开源且原生支持音视频，社区已涌现出针对B300、Apple Silicon等硬件的高效推理与蒸馏方案，对视频生成开发者极具参考价值。
Open weights. Shared progress. MiniMax H3 is moving fast.
开放权重。共享进展。MiniMax H3 正在快速推进。
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.
我们构建 H3 旨在实现原生立体音频和多模态参考控制的视频生成。开源社区正使这一能力变得更快、更易获取，且更易于在此基础上进行开发。
Recent highlights:
近期亮点：
• FastH3 — FastVideo, Nuva Lab and NVIDIA: 4-step distillation, now running on DGX Spark and Apple Silicon.
• FastH3 — FastVideo、Nuva Lab 和 NVIDIA：4 步蒸馏，现已在 DGX Spark 和 Apple Silicon 上运行。
• 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.*
• Sol-H3 — NVIDIA 的 SANA 团队：在 8×B300 上，于团队的预热推理基准测试中，6.6 秒内生成 768p 分辨率、时长 15 秒的视频加音频。*
• VDN — Haocheng Xi and the OpenVDN team: rethinking attention for faster H3 inference, with weights, training and inference code released.
• VDN — Haocheng Xi 与 OpenVDN 团队：重新思考注意力机制以实现更快的 H3 推理，并发布了权重、训练及推理代码。
• PDD — NVIDIA’s distillation method, brought to H3 by Alibaba PAI as 8-step Acc-LoRAs, now supported in ComfyUI.
• PDD — NVIDIA 的蒸馏方法，由阿里巴巴 PAI 引入 H3 作为 8 步 Acc-LoRAs，现已支持 ComfyUI。
• LightX2V — 4- and 8-step Turbo LoRAs, with workflows for text, image and reference-conditioned video + audio.
• LightX2V — 4 步和 8 步 Turbo LoRAs，提供文本、图像及参考条件控制下的视频加音频工作流。
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
每一次发布背后，都有人在训练、优化、量化、测试和分享。特别感谢： @haoailab @nuvalab @NVIDIAAI @xieenze_jr @HaochengXiUCB @ArashVahdat @julberner @LightX2V @ComfyUI 以及推动工作前进的个人贡献者： @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. 🚀
感谢您使用 H3 进行构建，并帮助其变得更快、更易获取、对社区更有用。当我们携手共建时，强大的模型才能发挥更大价值。继续推动 H3 的发展。期待看到接下来的成果。🚀
Explore the ecosystem: https://github.com/MiniMax-AI/awesome-minimax-h3-integration
探索生态系统：https://github.com/MiniMax-AI/awesome-minimax-h3-integration
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力