meta_transcribe_creative
Transcribe one library VIDEO now (videos are also transcribed automatically after upload, launch or history import; this also retries one the automatic pass gave up on): a model listens to the whole file once and the transcript — language, and what is said with the second each line starts — is stored on the upload and returned.
write meta_transcribe_creative
Transcribe one library VIDEO now (videos are also transcribed automatically after upload, launch or history import; this also retries one the automatic pass gave up on): a model listens to the whole file once and the transcript — language, and what is said with the second each line starts — is stored on the upload and returned. Written once, never overwritten: an upload that already has one returns it unchanged — except a stored transcript whose timings are impossible (lines faster than anyone speaks), which is hidden from reads and replaced here. Spoken words only; on-screen text is not transcribed. Refuses a static, an ad with no file in the library, and a file over the classifier's 18MB inline cap — it never guesses. A video with no speech is stored as hasSpeech false. Spends OUR model credits, not ad spend. Not money-moving; tagged as a write because it mutates stored data.
Parameters
| Parameter | Type | Required | Description |
|---|---|---|---|
assetId | string | Yes | The library upload's id. |