Text-to-Video repeat-production workflow

Text-to-Video repeat-production workflow

Buyer: Creative agents that turn one approved scene into the next usable clip.

Job: Generate a short video, keep the returned prompt, duration, aspect_ratio, and resolution, then make one deliberate variation without changing the delivery shape.

Start with a 5-second 16:9 draft. It keeps the first paid pass small and gives the next pass a stable brief.

First call

{
  "prompt": "A low tracking shot follows a red delivery bicycle through a rainy neon street at night, wet pavement reflections, 35mm film grain",
  "duration": "5s",
  "aspect_ratio": "16:9",
  "resolution": "720p"
}

Save the returned video_url for review. Keep the returned settings with the job record, along with the fal_request_id if you need to trace that generation.

Second call

Use the same duration, aspect ratio, and resolution. Change the instruction that was under review. Here, the camera move changes while the scene stays put.

{
  "prompt": "A low tracking shot follows a red delivery bicycle through a rainy neon street at night, wet pavement reflections, 35mm film grain, then the camera rises slowly to reveal the street ahead",
  "duration": "5s",
  "aspect_ratio": "16:9",
  "resolution": "720p"
}

text-to-video returns a hosted MP4 URL on each successful call. A dependency timeout or provider problem returns a labeled error response, so a production agent can retry later instead of treating a missing clip as an approved result.