MIXED REFERENCES
Direct MiniMax H3 with image, video, and audio references
A focused MiniMax H3 reference workflow covering mixed-media order, duration budgets, file limits, and prompt strategy.
References mode is designed for direction that cannot be expressed reliably with text alone. Images can establish identity and art direction, video can communicate movement or camera language, and audio can provide rhythm or sonic intent. A strong reference set is selective: every item has a clear job.
Choose References mode for a reference-led brief
Use Standard mode for text-to-video or first/last-frame animation. Use References mode when you want a mixed ordered set of image, video, and audio cues. References mode requires at least one image or video; audio by itself is not a valid visual foundation. Start with the smallest set that communicates the concept, then add only assets that resolve a specific ambiguity.
Order references intentionally
Minimax Motion preserves the order you set. Put the most important identity or composition reference first, then supporting style, motion, and sound cues. Use the move controls to reorder items before generation. In the prompt, refer to each asset by its role rather than assuming that a large pile of references will explain itself.
- check_circleIdentity or hero subject
- check_circleEnvironment and composition
- check_circleMotion or camera behavior
- check_circleRhythm, ambience, or sound direction
Stay inside the media budget
A single H3 request can include up to 9 images, 3 videos, and 3 audio files. Each video or audio reference must be 2–15 seconds, with separate 15-second total budgets for video and audio. Supported workspace formats include JPEG, PNG, WebP, HEIC/HEIF, MP4, MOV, WAV, and MP3. The interface shows per-item and total durations so an over-budget set can be corrected before generation.
Let assets carry appearance and the prompt carry intent
Do not transcribe every visible detail from the references. State what each source should influence and describe the new shot’s action, camera, timing, and desired relationship to the sources. If two references disagree, identify which one controls. This makes the brief easier to interpret and makes later iterations easier to diagnose.
Reference workflow checklist
- checkInclude at least one image or video
- checkGive every reference a distinct purpose
- checkArrange references in deliberate order
- checkKeep video and audio within separate 15-second totals
- checkUse the prompt to explain action and influence
Put the workflow into motion
Choose Standard or References, set the duration and resolution, and see the exact credit cost before you generate.
Create with H3 arrow_forward