Image to Video starts from an image and a motion brief. Reference to Video uses a model-supported set of images, videos, or both to guide a new shot. Choose it when several views or a motion reference help explain your intention. Neither workflow guarantees exact preservation of every source detail.
Give each reference a clear role in the shot.
Reference to Video combines a written brief with authorized reference images, reference videos, or a supported combination. Use the inputs to explain appearance, setting, composition, or motion, then describe the new shot you want. Available models accept different input combinations. The portrait alongside this guide is an illustration, not a generated Reference to Video result.

Assign a purpose before adding another reference
A useful reference set gives the model a coherent direction. Choose only media that contributes to the shot, distinguish the role of each input in your prompt, and review the selected model's requirements before submitting. More references do not guarantee stronger consistency.
- 01
Choose authorized source material
Start with images or videos you created or have permission and informed consent to process. Use clearly adult subjects and remove redundant or conflicting references. Check whether your selected model accepts images, videos, or both.
- 02
Explain what each reference contributes
Separate appearance, environment, framing, and movement in your brief. State the intended action and camera direction, then identify the visual qualities that should remain important. References guide a new generation; they do not guarantee exact reproduction.
- 03
Review the live setup and estimate
Upload the selected files, wait for validation, and review the available output settings and Credit estimate. Changing the model can change the permitted input combination. Resolve any upload or limit message before generating.
Match the reference set to the model
Reference to Video limits vary by the selected live model. Use the controls above to review supported media, file size, input count, video duration, and output settings. Reference audio is not supported by this workflow. The live Credit estimate applies to the complete selected setup.
| Goal | Focus on | Why it helps |
|---|---|---|
| Guide appearance and composition | A small, compatible image set | Choose clear views with consistent visual direction. Additional images should resolve an uncertainty rather than introduce a competing subject or setting. |
| Guide movement | Reference video when the model accepts it | Use a concise clip and describe the motion that matters. Review both individual-file and combined-duration limits in the generator. |
| Combine image and video inputs | The model's media and total-count limits | A model may impose a shared limit across input types. Recheck the complete reference set after switching models and remove unsupported files. |
| Compare results with a clear budget | One changed variable and the current Credit estimate | Keep references and output settings stable while refining the prompt. Review the updated estimate before each new request. |
Resolve an unclear reference direction
If Reference to Video produces an inconsistent shot, simplify the relationship between the inputs and the requested action. Compare one revision at a time. Inspect the completed result before saving or reusing it, especially when appearance, movement, or composition matters.
- References compete with each other
- Remove redundant views and conflicting environments. Explain which reference guides appearance, which guides the setting, and which guides movement.
- A video cannot be added
- Confirm the selected model accepts reference video, then check the displayed format, file-size, count, and duration requirements. Choose a shorter compatible clip when needed.
- Switching models changes the input set
- Review the remaining uploads and any validation message. The new model may support fewer files or a different media combination; complete that setup before generating.
- The result does not preserve a detail
- Use a clearer reference and name the important quality in the prompt. Reduce competing instructions and review a new test; exact identity, motion, and detail preservation are not guaranteed.
Reference to Video questions
Practical answers about live settings, Credits, account access, and consent-first creation.
The selected model determines supported image and video formats, file sizes, quantities, and duration limits. Some models accept images only. Others accept reference video or a combination, sometimes with a shared total-count limit. Check the current generator before preparing your files. Reference audio is not supported.
Describe the new shot, the subject's action, the camera movement, and the role of the references. Explain which visual or motion qualities matter instead of repeating every detail in every file. Keep the first test focused on one coherent scene and remove instructions that conflict with the source material.
Sign in to review your generation in History. Reuse reopens Reference to Video with the prompt and available image references. Reference videos are not restored: add them again before generating. Recheck the selected model, all inputs, output settings, and current Credit estimate. Private inputs remain tied to their owning account.
You can inspect the generator and setup while signed out. Submission requires an authenticated account. Return to the same route after signing in.
The live Workspace shows estimated Credits for the selected model and settings before submission. Change either to see the current estimate. Catalog availability and Credit requirements can change.
Clear references. One intentional shot.
Choose authorized inputs, explain their roles, and review the live Reference to Video setup above before generating.