Choose an input mode
Choose first frame, first and last frames, reference images, or an existing video, and match the request to live model capabilities.
On this page
Choose by what you already have
| Input | Choose | What it controls |
|---|---|---|
| One picture | First frame | The starting appearance and composition |
| A start and an end pose | First + last frame | Two endpoints for the model to connect |
| Several identity or appearance references | Reference images | Visual guidance, not an ordered animation timeline |
| An existing clip | Import video | Reuse motion and continue to editing and export |
Open Studio, select a workspace, then the input mode. The model selector only lists models compatible with that mode. Import video does not require choosing an AI video model.
Follow the current model, not a remembered preset
After changing a mode, model, or resolution, check duration, ratio, prompt requirements, and your quote again. The interface can normalize unsupported options. Switching away from first + last frame clears its end-image selection; switching away from reference images clears the additional references. Keep your original files.
A model may support only some modes, a fixed reference-image duration, or different durations at different resolutions. Watermark, negative-prompt, and seed options are not universal. See models and prompts.
Mapping for API and MCP callers
The shared public request uses input.first_frame, optional input.last_frame, and optional input.reference_images. The first image is still required in reference-image mode; the reference array contains the additional images. Do not send both an end frame and reference images to simulate a new mode.
Discover modes, supports_last_frame, supports_reference_images, max_reference_images, reference_image_duration_seconds, and the supported video settings before quoting. Read exact field constraints in the API schemas. Internal Studio requests and public requests do not use identical field names.
The public one-call image-to-animation API and its MCP tools do not expose Studio's video-import or Advanced Editor recipe workflow. Use Studio for those tasks instead of inventing public endpoints.
Keep transparency and cost separate
Input mode describes the images sent to a video model; Alpha Key describes the transparency workflow. Every image used in Alpha Key generation must contain meaningful Alpha. Adding a white-background reference to an otherwise transparent set can invalidate the request.
Changing the input or settings means a new intended request. Obtain a fresh quote and approval before generation. Reusing an existing video for another export does not require a new AI video generation, but export processing and format entitlement still apply.
Continue with first and last frames, reference images, input canvas, or video import.
Was this page helpful?
No search queries, code, or free text are collected. This switch controls documentation interactions only; general site analytics follow the privacy policy. Privacy policy