Use case
Your generator returns a URL. Pass it straight in.
Video models hand you a hosted file. Getting captions onto it usually means downloading it somewhere, running a transcriber, styling the result and re-encoding. This replaces all of that with one call that never touches your infrastructure.
// the URL your generator just handed you
caption_video({ inputUrl, preset: "highlight" })
Why it fits
Nothing is downloaded to your side.
- The file is fetched once, from us, over a connection that refuses anything that is not a real video stream.
- Generated speech is clean speech, so word timings land well. Send the product and character names as a dictionary and they come out spelled right on the first burn.
- Output is capped at 1080p, so a 4K generation comes back at a size a phone will actually play.
- If the model garbled a word, your agent fixes it and re-burns for free, 3 times per job.