Inspiration
Editorial window light
IMAGE · VIDEO · AUDIO REFERENCES
MiniMax H3 Reference to Video
Use images, video, and audio to guide different parts of a new shot. Assign each reference a role for appearance, movement, camera behavior, or sound, then generate with Standard H3 or H3 Max.
Model
Example videos
Get inspired
Inspiration
Game cinematic
Inspiration
Performance cut
Inspiration
Coastal aerial
If you're new to the generator, start with how to use MiniMax H3.
Creative direction from source media
When to Use Reference to Video
Use MiniMax H3 Reference to Video when you need to combine a subject's appearance from an image with movement from a separate video, or add audio guidance to a visually defined scene. It suits character scenes, product concepts, and shots that require several types of source material.
Unlike animating a start frame, this workflow uses source files to guide a new composition. Begin with the files needed for that shot; you do not need to fill every upload slot. If you want to animate one specific opening image, use Image to Video.
Reference roles
Image, Video, and Audio Reference Types
Separate the decisions each source should guide before uploading the files.
Images: Appearance and Visual Direction
Use clear images for a character, product, material, environment, or color palette. If two images show different people or product designs, explain which one defines the subject. Keep style references separate from appearance references in your brief.
Video: Movement and Timing
Choose a clip that clearly shows the action or camera behavior you need. Say whether to follow the person's movement, the camera path, or the pacing. Trim irrelevant lead-in and ending footage so the useful motion is easy to identify.
Audio: Sound Direction
Use an audio reference when words alone do not describe the ambience or rhythm you want. Explain its role and whether new dialogue or effects should accompany it. Listen to the generated result; supplying a track does not guarantee an unchanged copy.
Current workspace limits
Reference File Limits
Standard and Max use different upload limits; Turbo does not support this workflow.
| Control | H3 Standard | H3 Max |
|---|---|---|
| Reference count | Up to 9 images, 3 videos, and 3 audio files | Up to 9 images; 12 files combined |
| Visual source | At least one image or video | At least one image or video |
| File size | Images: 30 MB; video/audio: 100 MB | Images: 30 MB; video/audio: 100 MB |
| Output resolution | 768p or up to 2K | 480p or 768p |
With H3 Max, keep all reference videos to 15 seconds total. Standard H3 uses a separate reference setup; the Max duration limit should not be assumed to apply to Standard.
Credit estimate
How Reference Inputs Affect Credits
The estimate can include both the generated clip and the media used to guide it.
Standard H3: A 768p Example
A 5-second output uses 15 credits at 3 credits per second. An 8-second reference video adds 24 credits at the same rate. With one image reference, the estimate is 39 credits total.
In the current Standard calculation, the first five image references do not add an image charge; each additional image adds 2 credits. Reference video duration is rounded up to a whole second. Max uses different rates, so check the live estimate after choosing the model, files, and output settings.
Quick start
How to Generate a Video with References
Upload your sources first, assign their roles, then describe the shot they should help create.
- 01
Choose the Files That Matter
Start with the visual anchor: a character, product, setting, or a video showing the movement you need. Include at least one image or video. Add audio only when it has a specific purpose. Sign in when prompted to upload, and check the previews before writing the brief.
- 02
Give Each Reference a Role
Identify the files by their preview labels and say what to take from each one. For example, use Image 1 for the jacket design and Video 1 for the walking pace. State which background, camera, or sound details should be ignored if they are unrelated to the new shot.
- 03
Describe the New Shot
Explain the action, camera direction, setting, and audio you want in the output. References provide guidance, but the prompt still needs to define the scene. Keep the instructions compatible: a gentle walking reference and a request for an abrupt sprint describe different movements.
- 04
Review the Sources and Estimate
Choose the output settings and check the total credit estimate, including reference inputs. Click Generate. Review whether each source influenced the intended part of the result; remove an irrelevant file or revise its role before changing the whole reference set.
Role-based briefs
Reference to Video Prompt Examples
Upload your own matching files before trying these prompts, and adjust the file labels to match the previews.
Upload: Image 1 — character photo; Video 1 — walking clip.
Character Appearance + Walking Motion
Use Image 1 for the character's facial appearance, hairstyle, and jacket. Use Video 1 only for the relaxed walking pace and body movement, not its location or camera movement. The character walks through a quiet concrete courtyard. The camera tracks smoothly beside them at eye level. Keep their appearance consistent with Image 1. Soft daylight, footsteps, and light outdoor ambience. End with the character still fully in frame.
Reference to Video · H3 Standard · 16:9 · 768p · 5s
Upload: Image 1 — product image; Image 2 — lighting or background reference.
Product Appearance + Scene Mood
Use Image 1 for the product's shape, color, and material. Use Image 2 only for the lighting mood and background palette, not its objects. Place the product on a clean stone pedestal. The camera makes one slow push-in while the product stays stationary. Preserve the product proportions and keep the surrounding space uncluttered. Quiet studio ambience. End on a stable close-up. Do not add text or additional products.
Reference to Video · H3 Standard · 16:9 · 768p · 5s
Upload: Image 1 — character photo; Audio 1 — rain ambience.
Character Image + Audio Ambience
Use Image 1 for the character's appearance and clothing. Use Audio 1 as a reference for the quiet rain ambience, without adding dialogue or music. Place the character beside an apartment window at dusk. They slowly look outside while the camera remains still. Keep their face and outfit consistent with Image 1. Gentle rain is heard throughout, with a soft fabric rustle as they turn. End with the character looking through the window.
Reference to Video · H3 Standard · 16:9 · 768p · 5s
Resolve mixed signals
Common Reference to Video Problems
Use the result to identify which source or instruction is pulling the shot away from your brief.
The Wrong Background Carries Over
Clarify that the image supplies the character or product only, then describe the new setting. Remove additional environment images if they contradict that setting. Keep one visual source responsible for each major decision.
The Camera Copies the Motion Clip
If you wanted only the subject's movement, state that distinction: “Use Video 1 for the walking action; keep the output camera locked.” Avoid adding a second camera reference until the first role is clear.
The Result Combines Too Many Looks
Reduce the set to the subject source and one supporting reference. Add the others back one at a time, with a stated purpose. This makes it easier to see whether a new source improves the shot or introduces competing visual cues.
Questions
MiniMax H3 Reference to Video FAQ
Answers about reference roles, upload limits, model support, and input costs.
What is MiniMax H3 Reference to Video?
It generates a new shot using a prompt and supporting image, video, or audio files. Each reference can guide a different creative decision, such as a character's appearance, a movement pattern, a setting, or sound direction.
How many reference files can I upload?
Standard H3 currently supports up to 9 images, 3 videos, and 3 audio files. Images must be 30 MB or smaller; video and audio files can be up to 100 MB each. H3 Max supports up to 9 images within a combined limit of 12 reference files and limits reference videos to 15 seconds total. The selected model's upload area shows its current limits.
Can I generate from an audio reference alone?
No. Include at least one image or video reference to supply visual guidance. Audio can accompany those sources. Describe the intended role of the audio, such as ambience or rhythmic direction, rather than assuming it will be reproduced unchanged.
How should I refer to uploaded files in the prompt?
Use the preview labels, such as Image 1 or Video 1, and describe each file's job in plain language. Keep the relationship explicit: “Use Image 1 for the character and Video 1 for the movement.” Recheck the labels after replacing or removing a file. This is a writing convention, not a guarantee that every detail will transfer.
How is this different from Image to Video?
With Image to Video, an uploaded frame guides the opening or ending composition. A reference file supplies creative guidance for a new shot and does not necessarily become its first frame. For a planned beginning and ending, use the first and last frame workflow.
Can I use H3 Max Turbo for reference generation?
No. This workflow is available with Standard H3 and H3 Max. Turbo supports other input modes but has no Reference to Video option. Compare the faster workflows on the H3 Max page.
Do reference files change the credit cost?
They can. The current Standard estimate includes output duration, the total length of reference videos, and additional image charges beyond five image references. Model and resolution also affect cost. Review the live total before generating; a prompt-only estimate does not describe every reference setup. See credit pricing for available packs.
What if the references give conflicting directions?
Decide which file controls each part of the shot, then remove anything that competes with that role. If one image supplies the product and another supplies lighting, say so. Try fewer references before adding more; inspect appearance, motion, and audio separately to see which instruction needs refinement.
Generate a Video with References
Choose the visual source, add only the supporting media you need, and tell the generator what each file should guide.
Create with References →Sign in to upload and generate · Include an image or video