Reference-to-video takes subjects as references, not as the opening picture. The stack is those inputs together: three angles of a bottle, a face, a garment, a location still, sometimes a clip. Multi-reference is the ceiling of that mode — how many the row will hold. Extra files past the limit are dropped, not blended. The stack is the set you actually sent. Still contract is a different job: image-to-video must obey one input still as frame one. Mixing a pack shot into a reference slot is how the label becomes a suggestion and the camera invents a new angle.
A reference is identity without locking the frame. The model may re-angle, re-light, re-size. That is the point of a stack: describe the product in the round so a new shot can still be the same SKU. Duplicates teach nothing. A red sofa in the background of a 'product' still teaches the sofa. Clean plates and cutouts leak less. Clips in a stack, where a row accepts video input, are motion or identity over time, not a driving clip unless the row is motion control.
The prompt still owns the action. The stack owns who and what. Seed-lock once the identities hold, then change one clause of motion. Do not re-describe the bottle in the prompt if the bottle is already in the stack — that fight is how the SKU drifts.
In practice
- Count the row's reference limit before you add the fifth still. The extras will not be used.
- Disagree on purpose: different angles of one subject beat three near-identical crops.
- Keep pack-shot type on an I2V still when the label must hold. A reference stack will not lock frame one.
Reference-to-video models
Catalog entries that carry a subject from reference stills into motion. 23 of the 351 models in the Versely catalog qualify.
| Model | Provider | Type |
|---|---|---|
| Seedance 2.0 | ByteDance | Video |
| Gemini Omni Video | Video | |
| Happy Horse 1.0 Reference to Video | Alibaba | Video |
| Wan 2.7 Reference to Video | Wan | Video |
| Seedance 2.0 Fast Reference to Video | ByteDance | Video |
| Kling O3 Standard Reference to Video | Kling | Video |
| VEO 3.1 Reference to Video | Video |
The mistake to avoid
Dumping a moodboard into the reference slots and calling it a stack. A stack is a counted set of identities the row can actually hold.
Where you will run into it
- AI Video Generator — Text-to-video, image-to-video, and story chaining in one studio.
- AI UGC Video Generator — UGC ads at the speed and price of a prompt.
- Seedance 2.0 — Seedance 2.0 generates Hollywood-grade cinematic videos from text prompts with native audio-visual synchronization, director-level camera and lighting control, and exceptional motion stability
Related terms
Reference-to-video
Reference-to-video means the model builds a clip from separate images of a subject it can place anywhere. Those images are an identity, not the first frame.
Multi-reference
Multi-reference generation conditions a clip on several stills at once — a stack of identities or products, not one picture and a prompt.
Still contract
A still contract is the rule that an image-to-video clip must obey its input still — identity, product, framing — rather than drifting into a new picture.
Text-to-video
Text-to-video is generation from a written prompt alone — you describe a shot, the model invents every frame of it, and no image or footage goes in.
Image-to-video
Image-to-video animates a still you supply: the picture becomes the opening frame, and the prompt describes only what happens next.
The all-in-one AI studio for creators. 60+ models for video, image, voice, music and lipsync, in your browser or on your phone.
Free on the web and on iPhone, one account.