MiniMax H3 Fast Reference-to-Video API by MiniMax

MiniMax H3 Fast Reference-to-Video

MiniMax H3 Fast Reference-to-Video keeps the subjects from your reference materials consistent throughout a clip, at the lowest cost and quickest turnaround in the H3 family. Instead of animating a fixed frame, it uses the references as identity anchors — ideal for placing a character, product, or style into new scenes and motions. References can be any mix of images, videos, and an audio track (e.g. a music beat) to sync the motion to. Output is a complete 24fps 480P clip with audio.

Why Choose This?

• Fastest, most economical tier Speed-optimized 480P generation — the quickest and cheapest way to keep subjects consistent with H3 quality.

• Subject consistency Preserve one or several people, characters, or objects' identities across the whole clip.

• Mix reference types Combine reference images, reference videos, and reference audio in a single request.

• Audio sync (optional) Provide a reference audio track and sync the action to the beat.

• Prompt-driven scenes Put the reference subjects into any scene or action you describe.

• Selectable duration Produce 5–15s clips.

Parameters

Reference asset limits

refers takes a mix of images, videos and audio. Each type has its own limits:

• Total reference assets: up to 12 across all three types combined (the prompt does not count).
• Audio alone is rejected — every request needs at least one reference image or video.
• Video and audio durations are capped independently — 15s of video and 15s of audio can coexist.
• Beyond 5 reference images, each additional image adds to the cost; reference video is billed by its metered input duration.

Pass public HTTPS URLs. Every entry in refers must be a URL that returns the media file directly — not an HTML page. Base64 data URIs are not processed for this field, and signed URLs need a validity window long enough to survive asynchronous processing. Assets are validated after the task is accepted, so a successful submit does not guarantee a successful generation.

Task Status Callbacks (callback_url)

When you supply a callback_url, the MiniMax server notifies it on every task status change, so you don't have to poll for the result.

1. Verification handshake — right after the task is created, the server first sends a verification request to your URL containing a challenge field. Your endpoint must return the challenge value unchanged, within 3 seconds, to complete verification.
2. Status pushes — once verification succeeds, the server sends a POST to your URL every time the task status changes. The push body has the same structure as the Query Task (task status) response.

Callback status values: queued, running, succeeded, failed, cancelled.

The callback_url must be a publicly reachable HTTPS endpoint.

refers example

"refers": [
  { "url": "https://.../subject1.png", "type": "image" },
  { "url": "https://.../subject2.png" },
  { "url": "https://.../clip.mp4", "type": "video" },
  { "url": "https://.../beat.mp3", "type": "audio" }
]

How to Use

1. Add your references — any mix of images and/or videos (at least one), optionally plus an audio track.
2. Write your prompt — describe the scene, action, and camera movement for the subjects.
3. Set duration — pick any length from 5s to 15s.
4. Choose an aspect ratio — match your target platform.
5. Run — submit and download your video.

Best Use Cases

• Character Consistency — Keep the same character across multiple shots.
• Product Placement — Feature a specific product in generated scenes.
• Branded Content — Maintain a consistent mascot or spokesperson.
• Rapid Iteration — Explore reference-driven scenes quickly and cheaply.

Pro Tips

• Use clean, well-lit reference images with each subject clearly visible.
• Give one image per subject when combining several subjects in one video.
• Describe the new scene and action, not the reference materials themselves.
• Supply a reference audio track to drive the timing/beat of the motion.
• Combine with different ratios and durations to repurpose the same subjects across platforms.

Notes

• prompt and at least one reference image or video are required; a request with only audio is rejected. The prompt itself is capped at 7000 characters.
• Reference-to-video and image-to-video are mutually exclusive: this model uses reference materials, not first/last frames — do not mix the two.
• Ensure all reference URLs (image / video / audio) are publicly accessible and return the file directly.
• Reference assets must meet the size, dimension and format limits above — an oversized or wrongly encoded file fails after the task is accepted, not at submit time.
• Supported durations between 5s and 15s (whole seconds), at 24fps with audio.
• Resolution is limited to 480P — 768P and 2K are not available on the Fast model.
• Generation is asynchronous — submit, then poll for the finished video.

Related Models

• MiniMax H3 Fast Text-to-Video — Generate video directly from a text prompt at the same speed.
• MiniMax H3 Fast Image-to-Video — Animate a first-frame image (optionally with a last frame).
• MiniMax H3 Reference-to-Video — The standard H3 model, with 2K output.