All local H3 guides
Start here / Local H3

How to run MiniMax H3 locally

A practical MiniMax H3 local deployment guide covering checkpoints, frameworks, hardware paths, 768p output, 2K limits, and validation.

Reviewed
2026-08-09
Reading time
9 min
Evidence
5 official / 0 community
Short answer

Run H3-Base-FL2VA for text and first/last-frame generation, or H3-Base-Ref2VA for image, video, and audio references. ComfyUI is the most accessible path; SGLang, vLLM, and Diffusers are better suited to code-driven or multi-GPU serving. The released base weights produce 768p video with stereo audio. The official Context-IR and Regenerate-2K stages remain hosted services.

01

What local deployment includes

The open release is H3-Base, the video-and-audio generator at the center of the complete H3 system. It accepts a prompt and optional visual or audio conditions, then produces 24 fps video with a 32 kHz stereo soundtrack. The native output recipe uses a 768-pixel short edge, such as 1344 x 768 for 16:9.

Two surrounding modules are different. H3-Context-IR interprets complex multimodal instructions before generation, and H3-Regenerate-2K regenerates a selected 768p result at higher resolution using the original context. MiniMax says both modules are hosted and are not included in the initial open release.

ModuleLocal weightsJob
H3-Context-IRNoHosted instruction processing for complex multimodal requests
H3-BaseYesGenerates 768p video and stereo audio
H3-Regenerate-2KNoRegenerates a selected base result at 2K
02

Choose the checkpoint before the framework

FL2VA and Ref2VA are separate task-specific checkpoint families. Downloading both wastes storage when a machine serves only one workflow.

  • FL2VA: text-to-video, first-frame-to-video, last-frame-to-video, and first-and-last-frame generation.
  • Ref2VA: reference-driven generation with up to 9 images, 3 videos, and 3 audio clips within the documented combined limits.
  • Video-to-video is a Ref2VA use case, not a separate public checkpoint or task value.
03

Pick a deployment path

ComfyUI 0.30.0 and later includes native H3 nodes and templates. SGLang publishes measured recipes for NVIDIA datacenter cards, dual RTX 5090, and AMD Instinct hardware. Diffusers can fetch only the components required by its modular pipeline. Avoid mixing commands, model layouts, or quantization files from different frameworks unless the relevant guide explicitly supports that combination.

PathBest forTrade-off
ComfyUIWorkstations and visual iterationFastest setup; hardware-specific repacks and offloading vary
DiffusersPython experiments and custom pipelinesDirect control; you own memory placement and serving
SGLangMeasured multi-GPU servingStrong official recipes; production hardware is expensive
vLLMAPI-oriented serving teamsFamiliar service workflow; follow the current H3 recipe exactly
04

Validate with a small reproducible run

Do not start with a 15-second reference-heavy job. First prove that model loading, conditioning, video decode, audio decode, and file export all work on the target machine.

  • Use the 768-pixel-short-edge recipe only after a smaller preview completes.
  • Fix the prompt, seed, dimensions, duration, step count, and model revision.
  • Check the MP4 for video, stereo audio, duration, frame size, and a conditioning input that visibly affects the result.
  • Record peak VRAM, peak system RAM, model load time, denoise time, decode time, and end-to-end time.
05

The practical recommendation

Use ComfyUI when the goal is local creative iteration on a single workstation. Use SGLang or vLLM when several users need a stable endpoint and the deployment has multiple large GPUs. Use the local base model to test prompts and motion at 768p, then send only selected clips through the hosted 2K stage when the project permits an API boundary.

Frequently asked questions
Can MiniMax H3 run completely offline?

The released H3-Base checkpoints can generate 768p video and stereo audio locally after their files are downloaded. The official Context-IR preprocessing system and Regenerate-2K stage are not part of the open release.

Which MiniMax H3 checkpoint should I download?

Choose FL2VA for text-to-video and first-frame, last-frame, or first-and-last-frame generation. Choose Ref2VA when the request uses reference images, videos, or audio.

Does local MiniMax H3 generate 2K video?

The open H3-Base release is the 768p generation stage. Official 2K output uses H3-Regenerate-2K, which MiniMax has not open-sourced as of the review date.

What is the easiest way to run MiniMax H3 locally?

For a workstation, start with the native MiniMax H3 templates in ComfyUI 0.30.0 or later. For a production endpoint, use an official SGLang, vLLM, or Diffusers recipe.

Sources and evidence

Read the original material

Community figures describe the named machine and workflow only. They are not MiniMax performance guarantees.

  1. Official documentationMiniMax H3 model card and local deployment
  2. Official documentationComfyUI native H3 workflows
  3. Official documentationSGLang H3 deployment cookbook
  4. Official documentationHugging Face Diffusers H3 pipeline
  5. Official documentationMiniMax H3 Community License