Skip to main content

​
Choose the source

Start with the original image, video, or audio you want to use. Confirm that you have access to the asset and that the chosen workflow accepts its media type. Use assets you have permission to provide. Check the source’s format, dimensions, orientation, duration, and audio track. An image-to-video workflow, a video edit, and a talking avatar have different input requirements. Select the source and references for the actual workflow.

​
Upload a local file

The MCP local upload flow separates transfer from verification:
1

Prepare the upload

Compute the file’s exact size, MIME type, and SHA-256 checksum. Request upload instructions through your connected client.
2

Transfer the bytes

Upload the file directly to the provided upload URL. Include every returned header exactly and complete the transfer before the URL expires.
3

Verify completion

Complete the upload verification using the existing upload identifier. Use the verified asset reference returned by completion for generation.
The MCP transfer URL is valid for five minutes. Completion verification can be retried with the same upload identifier for one hour. A prepared upload or a successful transfer alone is not a verified asset. If verification fails, check the size, MIME type, and checksum of the original file. Follow the completion flow for that upload before preparing another one.

​
Import a remote file

For a remote source, use the supported import flow with a publicly downloadable URL. A page that requires login or displays a preview is not necessarily a direct file download. Wait for the import to return a usable asset reference. Upload and import references belong to the connected account. A temporary upload URL is for transferring bytes; use the completed asset reference as the input to a later operation.

​
Check references together

Source footage, additional elements, end frames, and voices can share reference budgets. Check them together rather than validating each item in isolation. Limits vary by workflow and quality; choose the settings supported by your client. For dialogue, distinguish written speech from selected voices and independent audio references. See the generation workflow before submitting.