1. Select your source and template
Read subscription benefits, templates, voices and avatars. Use returned IDs in subsequent requests. Create project accepts supported text, URL, document, script and audio origins. Its schema defines the source fields andscript_params; choose the fields appropriate to your origin.
Preserve the full project ID returned by creation or listing. Poll Get project with
query.project_id until the script is ready. To change the source and regenerate, use Regenerate script with prj_id.
2. Review and save the script settings
Get settings returns saved settings. Use Save project to save scenes, speakers, voices, avatars, layout and subtitle options before preparing audio. Keep the endpoint’s exact scene structure and IDs.The settings response can contain a shortened display ID. Retain the full project ID from creation or listing for future requests.
3. Prepare audio
data.task_id, then poll without an idempotency key:
3 means success and 4 means failure. This stage prepares the editable workspace as well as speech. For one changed scene, use Redo speech and Get speech task.
4. Edit the workspace
Read Get workspace, edit its supported timeline/layout fields, then Save workspace and Apply workspace when using a draft. Preserve existing media references so you can reuse prepared audio.5. Check credits and render
Call Render credit check withproject_id. Review the returned estimate against the active subscription.
Render accepts a tasks array of prepared workspace render requests. It does not accept only a project ID. Build each task from the project’s prepared workspace and the endpoint schema; don’t send an empty timeline or an empty task list.
Poll Get project for render state and the final video URL. Use Retry project only after inspecting a failed project’s state and existing retry rules.