Now on VeniceVideoPrivate

MiniMax H3 Max

MiniMax H3 Max is a high-speed, audio-native video model optimized for prompt adherence and fast iteration, supporting text, image, and reference-based generation up to 1080P with native audio.

For agents
curl https://api.venice.ai/api/v1/video/queue \
  -H "Authorization: Bearer $VENICE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "minimax-h3-max-text-to-video",
    "prompt": "Aerial drone shot over a misty mountain valley at golden hour"
  }'

# Use the returned queue_id with https://api.venice.ai/api/v1/video/retrieve.
# Call /video/complete after downloading if needed.
Model IDminimax-h3-max-text-to-video
Maker
MiniMax (base model), fal Research (post-trained derivative)
Modality
Video + audio
Max duration
15 seconds
Max resolution
1080P

Overview

What is MiniMax H3 Max

MiniMax H3 Max is a high-performance video generation model developed by fal Research on top of MiniMax's open H3 base. It supports text-to-video, image-to-video, and reference-to-video generation with native synchronized audio, offering fast turnaround (under 3 seconds for a 5s clip) and strong prompt adherence at resolutions up to 1080P.

Running it privately on Venice

On Venice, MiniMax H3 Max runs with full privacy — your prompts are never stored or profiled, ensuring zero retention. This uncensored, permissionless access lets creators and developers generate video content without surveillance, ideal for sensitive or commercial use where sovereignty over prompts and outputs matters.

PrivateNo prompt trainingTEE · hardware enclaveEnd-to-end encrypted

Specifications

Datasheet

Maker
MiniMax (base model), fal Research (post-trained derivative)
Modality
Text-to-video, image-to-video, reference-to-video
License
MiniMax H3 Community License
Clip lengths
5s – 15s
Resolutions
480P, 768P, 1080P
Mode
text-to-video
Audio
Yes
Prompt limit
10,000 chars
Released
August 27, 2026
Architecture
Post-trained derivative of MiniMax H3
Parameters
Not disclosed
Aspect ratios
16:9, 21:9, 4:3, 1:1, 3:4, 9:16
Privacy on Venice
Private — zero retention
Available on Venice since
Aug 2026

Assessment

Strengths and limitations

Strengths
  • Extremely fast generation: ~3 seconds for a 5-second 768P clip, enabling rapid iteration.
  • Native synchronized audio generation: sound effects, dialogue, and music are baked into the video during inference.
  • Strong prompt adherence and camera control: supports explicit directions like 'slow dolly-in' with accurate execution.
  • Reference-to-video capability: preserves specific people, products, or objects from reference images into new scenes.
  • Available in multiple aspect ratios including cinematic 21:9 and vertical 9:16, ideal for social and ad content.
Limitations
  • No open weights for H3 Max itself: only the base H3 model is open under a community license.
  • Maximum resolution capped at 1080P: 2K output is only available in the base MiniMax H3.
  • No seed control observed: makes exact reproducibility between runs difficult.
  • Reference inputs limited to images: video or audio references are not supported.

Use cases

What it is good for

  1. 01Rapid prototyping of ad creatives and social media clips with consistent branding.
  2. 02Animating product images or character designs into short promotional videos.
  3. 03Generating cinematic scenes with precise camera movements from descriptive prompts.
  4. 04Creating content with synchronized voice or ambient sound without post-production.
  5. 05Maintaining visual consistency of people or products across multiple clips using reference images.

Prompting

Getting better results

Use explicit camera direction language like 'slow zoom in' or 'over-the-shoulder shot' for precise motion.

Include detailed lighting and wardrobe descriptions to guide scene composition.

Upload reference images of people or products to ensure visual consistency in the output.

Keep prompts under 10,000 characters but be specific — the model excels at following detailed instructions.

Use 16:9 or 9:16 aspect ratios for platform-optimized content (YouTube, TikTok, Instagram).

For image-to-video, provide a clear first and last frame description to guide motion.

Samples

Sample outputs

Generated on Venice with our standard prompt suite — the same prompts we run through every model of this type, so you can judge it like-for-like.

Cinematic landscape

Slow aerial drone shot gliding over a misty mountain valley at golden hour, sunlight piercing clouds onto a winding river, ancient pine forests on either side, ultra-smooth motion, professional color grading, atmospheric haze, 4K

Seamless loop

Calm ocean waves rolling onto a black-sand beach at sunrise, golden light on wet sand, foam dissolving into the shore, a single silhouetted figure at the waterline, smooth continuous forward push, serene cinematic atmosphere

Urban cinematic

A slow tracking shot through a rain-soaked Tokyo street at night, neon reflecting in puddles, steam rising from a food cart, a person with a translucent umbrella, shallow depth of field, teal-and-orange grade, smooth steady camera

Compare every video model on these prompts

Alternatives

How it compares

ModelBest forMax durationMax resolutionNative audio
MiniMax H3 MaxFast, controllable short-form video15s1080PYes
Grok Imagine 1.5Integrated X ecosystem15s1080pYes
Wan 2.7 EnhancedOpen weights, high fidelity15s1080pYes
Vidu Q3Long coherent scenes16s1080pYes
Flux 3 First Last FramePrecise motion control20s1080pYes

MiniMax H3 Max is the best pick for creators who need fast, high-quality short-form video with strong prompt control, native audio, and reference-based consistency — ideal for ads, social content, and storyboarding.

API

Call it from your code

Venice exposes this model through the REST API. Queue a generation with the model id.

curl https://api.venice.ai/api/v1/video/queue \
  -H "Authorization: Bearer $VENICE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "minimax-h3-max-text-to-video",
    "prompt": "Aerial drone shot over a misty mountain valley at golden hour"
  }'

# Use the returned queue_id with https://api.venice.ai/api/v1/video/retrieve.
# Call /video/complete after downloading if needed.

Pricing

What it costs on Venice

Pay per clip on Venice — price scales with resolution and duration (5s–15s), from $0.15.

480P · 5s
$0.15
Per clip
768P · 5s
$0.24
Per clip
1080P · 5s
$0.48
Per clip

New Venice accounts include a free daily allowance and 500 welcome credits — no credit card required.

FAQ

Frequently asked questions

MiniMax H3 Max is a high-speed video generation model developed by fal Research on the open MiniMax H3 base. It supports text-to-video, image-to-video, and reference-to-video with native audio, offering fast turnaround and strong prompt adherence at up to 1080P resolution.

On Venice, pricing starts at $0.15 per 5-second 480P clip and scales with resolution and duration. A 1080P 5s clip costs $0.48. You pay per generation with no subscription required.

MiniMax H3 Max is not open source — it's a post-trained derivative of the open MiniMax H3 base, which is released under the MiniMax H3 Community License. The Max variant itself is closed and not available for self-hosting.

No, MiniMax H3 Max does not support tool use. It is a specialized video generation model and does not have access to external tools or APIs during inference.

MiniMax H3 Max supports 480P, 768P, and 1080P resolutions. The 768P output is rendered at 1344×768, not 1280×720, so the actual frame size is slightly wider.

Yes, MiniMax H3 Max supports image-to-video generation. You can animate a still image into motion by providing a starting frame, and optionally a last frame, to guide the animation.

Yes, MiniMax H3 Max supports reference-to-video. You can upload reference images of people, products, or objects to maintain visual consistency when generating new scenes.

Yes, MiniMax H3 Max generates native synchronized audio, including dialogue, ambient sound, and music, directly during video generation — no separate audio pass is needed.

No, MiniMax H3 Max does not support video references. It only accepts image references for subject consistency. Video or audio files cannot be used as input references.

Run MiniMax H3 Max privately

No prompt logging. No data used for training.