I designed the pipeline in Claude Code, with human approval gates between stages, and integrated generative image, video, speech and music tools through CLI workflows. Cost estimation, batch spend ceilings, job logging and resumable state keep generation under control. Automated QA checks catch clipped dialogue and visual glitches, and normalise loudness.
Key decisions
- Cost estimation, batch spend ceilings, job logging and resumable state
- Automated QA checks for clipped dialogue, visual glitches and loudness normalisation