MiniMax H3
MiniMax's open-weight video model that turns text, images, video and audio into clips with stereo sound.
MiniMax H3 is MiniMax's open, general-purpose multimodal video model: it takes text, images, video and audio as input and generates video with native stereo sound, up to 15 seconds at 2K. On MiniMax's pay-as-you-go API, H3 costs $0.13 per second of 2K output and $0.08 per second at 768P.
About MiniMax H3
H3 generates video from a text prompt, from a first frame, a last frame or both, or from reference images, videos and audio that guide the character, motion, camera, style, voice or editing rhythm. A request takes up to 9 reference images, 3 video clips and 3 audio clips, 12 files in all, and a prompt of up to 7,000 characters. An optional H3-Context-IR task turns mixed inputs into an enhanced prompt without making a video.
The API picks a variant through its model field: H3 is the quality-first variant at 768P or 2K, H3 Highspeed balances quality and speed at 768P, H3 Lite costs less at 480P or 768P, and H3 Max, released jointly with fal.ai and post-trained by fal.ai on H3, is the fast variant at 480P or 768P. Clips run 4 to 15 seconds, or 5 to 15 on H3 Max. Generation is asynchronous: you create a task, poll its status and download the video.
MiniMax has published H3-Base, which generates 768p video with synchronized stereo audio, under the MiniMax H3 Community License Agreement, with weights in the MiniMaxAI/MiniMax-H3 repository. Its docs describe a local ComfyUI workflow with native T2V, I2V and R2V templates on converted Comfy-Org weights, and a self-hosted SGLang API service on NVIDIA data-center GPUs.
MiniMax says H3 is built for advertising, branding, e-commerce, product design, UI/UX and gaming, and that early testing shows it excels at instruction following, accurate text and brand rendering and V2V motion transfer. Besides the API, H3 runs in the MiniMax Design and MiniMax Code apps, and the launch post sends readers to Hailuo AI to try it.
Sources: platform.minimax.io/docs/pricing/overview, agent.minimax.io/about, /docs/guides/video-generation, /docs/api-reference/video-generation-v2-create, minimax.io/blog/minimax-h3, /docs/guides/local-deploy-h3 and /docs/guides/video-prompt, read 2026-10-27
MiniMax H3 pros and cons
Pros
- Audio input is free, and the first 5 images per task are free on the flagship H3 models and the first 2 on H3 Max.
- Every task type, from text-to-video to reference generation, costs the same per second for a given model and resolution.
- For 2K output, H3 regenerates its own lower-resolution result in context from the original inputs instead of using a separate super-resolution module, which MiniMax says recovers small text and fine detail.
- Prompts can say in plain words how the inputs relate, such as taking the camera movement from one clip and the vocals from an audio file; MiniMax says H3 handles that full-modality understanding on its own.
Cons
- The MiniMax H3 Community License lists the United States, the European Union, the United Kingdom and South Korea as Excluded Territories that require a separate license.
- The open release has H3-Base but not H3-Context-IR or the 2K regeneration step, so a self-hosted service does not reproduce MiniMax's platform workflow.
- MiniMax's prepaid Video Packages do not cover H3 yet, so H3 API use is pay as you go or a custom plan through sales.
- Self-hosting has no published minimum: the ComfyUI guide gives no GPU memory figure, and the SGLang path is a Preview whose reference setup is one node with 8 NVIDIA B200 GPUs.
Tagged
- Web-based
MiniMax H3 alternatives
The closest alternatives to MiniMax H3.
Related reading
Featured on Vibedonalds
Own MiniMax H3? Add this badge to your site to show you’re listed — and link back to your profile here.
<a href="https://vibedonalds.com/tools/minimax-h3" target="_blank" rel="noopener">
<img src="https://vibedonalds.com/badge/featured-on-vibedonalds.svg" alt="MiniMax H3 — Featured on Vibedonalds" width="240" height="60" loading="lazy" />
</a>Frequently asked questions
- What is MiniMax H3?
- MiniMax H3 — MiniMax's open-weight video model that turns text, images, video and audio into clips with stereo sound.
- What are the alternatives to MiniMax H3?
- The closest alternatives to MiniMax H3 are Gemini Omni 1.1 Flash, Kling 4.0, Fal.ai.
- What are the downsides of MiniMax H3?
- The MiniMax H3 Community License lists the United States, the European Union, the United Kingdom and South Korea as Excluded Territories that require a separate license. The open release has H3-Base but not H3-Context-IR or the 2K regeneration step, so a self-hosted service does not reproduce MiniMax's platform workflow. MiniMax's prepaid Video Packages do not cover H3 yet, so H3 API use is pay as you go or a custom plan through sales. Self-hosting has no published minimum: the ComfyUI guide gives no GPU memory figure, and the SGLang path is a Preview whose reference setup is one node with 8 NVIDIA B200 GPUs.
- What is MiniMax H3?
- MiniMax H3 is a general-purpose multimodal video model that MiniMax launched on July 31, 2026. It reads text, images, video and audio as one context and generates video with native stereo sound, up to 15 seconds at 2K, and MiniMax says it is built for advertising, branding, e-commerce, product design, UI/UX and gaming.
- How much does the MiniMax H3 API cost?
- It is pay as you go, in USD per second of output: H3 costs $0.13 at 2K and $0.08 at 768P, H3 Highspeed $0.06 at 768P, H3 Lite $0.04 at 768P and $0.025 at 480P, and H3 Max $0.08 at 768P and $0.05 at 480P. Input video is billed per second too, and audio input is free.
- Can I run MiniMax H3 locally?
- MiniMax's docs cover a ComfyUI workflow with native T2V, I2V and R2V templates and a self-hosted SGLang API whose reference setup is one node with 8 NVIDIA B200 GPUs, plus a community-maintained vLLM-Omni recipe. The license names the United States, European Union, United Kingdom and South Korea as Excluded Territories that require a separate license.
- Where can I try MiniMax H3?
- MiniMax's H3 docs list MiniMax Design, where you can claim desktop benefits and try H3 video generation 3 times for free, along with Hailuo AI and the Open Platform API. MiniMax Code also offers H3 video generation, from a prompt or reference material to the final cut.
- What is the difference between MiniMax H3 and H3 Max?
- H3 is the quality-first variant, with 768P or 2K output and clips of 4 to 15 seconds. H3 Max, jointly released by MiniMax and fal.ai and post-trained by fal.ai on H3, is the fast variant: it outputs 480P or 768P, generates faster than H3, and makes clips of 5 to 15 seconds.