product-videolisted

# Product video as code Build product videos as **code**: an HTML deck driven by a deterministic timeline, an AI voiceover, real screen recordings of the product, and an ffmpeg audio master. Everything is reproducible, diffable and re-renderable, so revisions cost minutes instead of a rebuild. This is not a video editor. It is a pipeline. The output is an MP4. ## The one idea that matters **Audio first, then timing, then picture.** Generate the narration before deciding how long anything is on screen. Measure each clip, derive scene durations from those measurements, and only then render frames. The alternatives - guessing durations, or writing narration to fit a fixed storyboard - produce visuals that drift out of sync with the voice and force a re-render every time a line is reworded. ``` script.json ──► TTS ──► measure ──► timings.json ──► render ──► composite ──► mux ──► MP4 ▲ ▲ ▲ assemble.py render.py composite.py ``` For a video with no narration (a title sting, a silent loop, a muted social clip), skip stage 1 and write `timings.json` by hand. Everything downstream is unchanged. ## Pipeline | Stage | Script | Produces | |---|---|---| | 1. Narration | your TTS tool of choice | `audio/vo_NN.mp3`, one per scene | | 2. Timing + voice track | `scripts/assemble.py` | `timings.json`, `out/voice.wav` | | 3. Picture | `scripts/render.py` | `out/silent.mp4` | | 4