Skip to content

Rendering ​

A render makes an MP4 and an .srt subtitle file for each format, under the project's renders/ folder.

From the editor ​

Press Render at the top of the editor:

  1. Pick a Driver. Each one shows ready, or what it needs installed. Set Frame rate to 15 for a quick draft.
  2. Press Render. It renders every format checked under Project → Formats to render.
  3. Progress shows as it goes. Finished videos appear underneath, ready to play.

Renders run one at a time. More wait their turn as queued.

Stopping or restarting the server stops any render in progress and fails the ones waiting, so start them again afterwards. Under pnpm dev, a hot reload from saving or pulling a file in src/ counts as a restart.

From the command line ​

bash
pnpm kick formercial render my-product --format all

See Command line for every option.

What happens ​

  1. Missing or changed voice lines are generated (when edge-tts is installed).
  2. Each scene's start and length are worked out from its blocks and voice.
  3. Headless Chromium opens the render page at the format's size and steps through every frame, taking a screenshot of each.
  4. ffmpeg encodes the frames, places each voice clip at its scene's start, and mixes in the music.

Rendering takes roughly 2–3 times the length of the spot at 30 fps. Use --fps 24 or --seconds 10 for quick checks. Chromium and ffmpeg run at low priority, so other work on the machine goes first. With an NVIDIA GPU, the chromium-nvenc driver uses about half the CPU.

Output ​

projects/my-product/renders/
  16x9-20261030-134300.mp4
  9x16-20261030-135512.mp4
  srt/
    16x9-20261030-134300.srt
    9x16-20261030-135512.srt
  • Video: H.264, from x264 or, with chromium-nvenc, an NVIDIA GPU.
  • Audio: AAC at 192 kbps.
  • Captions: when on, they're drawn into the picture. The .srt then goes in renders/srt/, ready to upload. It isn't beside the video because players like VLC and mpv load a same-named .srt automatically, which would show the captions twice.
  • Captions follow the voice: edge-tts reports when each word is said, so each piece of a caption shows while its words are spoken, in the picture and in the .srt. Turn on Light each word as it's said ("captionStyle": "words") to light the word being spoken too. Clips voiced before this have no word times: change the line, or delete voice/.made.json, to voice them again.
  • Subtitles without burned-in captions: with captions off, the .srt sits beside the MP4, so players do load it. Each cue runs while its line is spoken.

Git ignores renders, and everything else in the default projects/ folder except the example project. To keep your commercials in version control, point FORMERCIAL_PROJECTS at a folder in your own repository.

Released under the Apache License 2.0.