Omni Reference AI Video Generator
Combine image, video, and audio references in one generation. Choose ByteDance's Seedance 2.5 or MiniMax H3, give every reference a role, and generate consistent video.
Two Models, One Platform
Seedance 2.5 — Long Takes, Deep Reference Control
ByteDance's model. 4–30s at 480P or 720P, with up to 30 images, 10 video clips, and 10 audio files — 50 references in one generation.
MiniMax H3 — Higher Resolution, Focused Shots
MiniMax's model. 4–15s at 768P up to 2K, with up to 9 images, 3 video clips, and 3 audio files. Suited for short storyboarded scenes.
One Reference Library, Both Models
Switch models without reuploading. Compare both against the same reference pack and keep the take that fits the shot.
Omni Reference in 3 Simple Steps
- STEP 1
Upload Your References
Drag, paste, or select images, video clips, and audio from your device or the Assets panel, where anything you made in AI Image Editing already sits. - STEP 2
Assign Reference Roles
Name them in the prompt: Image 1 the character, Image 2 the scene, Video 1 the motion, Audio 1 the dialogue. One job per file. - STEP 3
Configure and Generate
Pick your model, then set duration, aspect ratio, and resolution. Check the credit cost on the Generate button before you submit — it varies.
What Happens When Every Reference Has a Job
Produce AI Drama with a Consistent Cast
Character consistency across shots and across episodes is the most expensive problem in AI drama production. Omni Reference locks cast identity to image references that carry from scene to scene.
One Cast, Every Shot
Upload character references and assign each an identity role. The same face, hair, and outfit carry through every shot.
Continuity Across Episodes
Reuse the same reference pack across scenes and episodes. Cast, sets, and props stay consistent because the references carry the constraints.
Assign Dialogue by Speaker
Attach audio references carrying your dialogue, then assign each line to a named character in the prompt itself.
Matching Scene Transitions
Set first and last frames to control where a scene opens and lands. Both endpoints stay reference-driven.
Music Video Production with Audio References
An AI music video lives or dies on whether the performer looks the same in every cut. Upload your track as an audio reference and performer images as identity references, then generate.
Audio-Driven Generation
With Seedance 2.5, an audio reference guides soundtrack and dialogue timing. Pair it with image references.
Plan Your Beats First
Structure the shot as timed beats in the prompt: one action and one camera intent per beat.
Performer Stays On-Model
Lock the performer with image references. Change set, lighting, and angle while the face stays fixed.
Cross-Cut Continuity
Generate a wide, a close-up, and a profile from one reference pack — same performer, wardrobe, location.
Built for All Creators
Anime & Webtoon Creators
Lock a character's face, hair, outfit, and proportions across a whole series. Upload character sheets as image references and generate anime character video that keeps the same cast episode after episode.
Short-Form Filmmakers
Cast, location, and wardrobe stay fixed across cuts. Upload identity, set, and costume references once, then generate any shot in the sequence from the same pack.
Brand & Ad Teams
Lock the product, spokesperson, and setting in every shot. Image references control silhouette, color, and packaging. Add audio for dialogue or soundtrack input.
Music Video Producers
Upload a track as an audio reference and performer images as identity references. Generate scenes where the performer stays on-model across every cut.