Create 480p video with text, frame control, and multimodal references.
MiniMax H3 Fast supports text-to-video, first/last-frame animation, and image, video, or audio references through one asynchronous API. Generate 5–15 second clips at 480p for 11.68 credits per output second, 92% of the H3 Max 480p rate rounded to two decimal places. All input materials are free.
Model capabilities
Use MiniMax H3 Fast from the same platform surface as the rest of your video and image stack: API keys, credits, logs, webhooks, and docs stay consistent across providers.
Describe the scene and omit media inputs. Choose 21:9, 16:9, 4:3, 1:1, 3:4, or 9:16; text-only requests default to 16:9.
Use first_and_last_frames with one or two image URLs to control the opening and optional ending frame. Output composition follows the frame images.
The default omni_reference mode accepts up to nine images, three videos, and three audio files. Audio references require at least one image or video.
Every input image, video, and audio reference is free. The estimate depends only on the requested output seconds and your account multiplier.
Use a prompt of up to 7,000 characters and public HTTP(S) media URLs. Choose first_and_last_frames for frame control or omni_reference for optional multimodal inputs.
480p is the only resolution. The supported duration range is 5–15 seconds, with a default of 5. Text mode uses a fixed aspect ratio; reference mode may use adaptive.
Estimate with POST /v1/estimate, then submit to POST /v1/videos/generations. Poll the returned task ID or receive a callback, and collect the completed video from output.urls.
AI Video APIcurl -X POST https://api.aivideoapi.ai/v1/videos/generations \
-H "Authorization: Bearer sk-your-api-key" \
-H "Content-Type: application/json" \
-d '{
"model": "minimax-h3-fast",
"callback_url": "https://your-server.com/webhook",
"input": {
"prompt": "Keep the product identity while following the reference camera motion",
"generation_type": "omni_reference",
"image_urls": ["https://example.com/product.png"],
"video_urls": ["https://example.com/camera-motion.mp4"],
"resolution": "480p",
"duration": 5,
"aspect_ratio": "adaptive"
}
}'Frequently asked questions
Fast supports 480p output and adds omni_reference image, video, and audio inputs alongside text-to-video and first/last frames. Its output rate is 11.68 credits per second, 92% of the H3 Max 480p rate rounded to two decimal places. All Fast input materials are free.
The supported range and Playground controls are 5–15 seconds, with a default of 5. The API forwards any positive integer duration unchanged, including 4 or 16; acceptance outside the supported range depends on the generation service. Estimation uses the same requested duration.
Omni-reference accepts up to nine images, three videos, and three audio files. Video and audio clips must each be 2–15 seconds, with a 15-second combined limit per media type. Audio requires an image or video reference. Frame mode accepts one or two images and cannot include video or audio.
Multiply the requested output seconds by 11.68 and round the full base charge to two decimal places, apply your account multiplier, then round again. Input usage never adds fees. A confirmed failed task is refunded once; a successful generation awaiting output delivery keeps its pre-charge while delivery retries.
Pricing and usage
92% of H3 Max 480p, rounded to two decimal places
Before account-specific multipliers; 5 seconds is the default
All supported references, including images 6–9, carry no input charge
Create one API key, use one credit balance, and switch between video and image models without provider-specific plumbing.