One Docker box that turns video noise into VTT and chapters

What it does Subvert is a self-contained PHP application packaged in a single Docker image. You feed it a video file; it extracts the audio with FFmpeg, sends it to OpenAI’s Whisper for transcription into VTT format, and optionally passes the resulting text through a ChatGPT model to generate timestamped chapters and a brief summary. The whole thing runs as a local web server, which means your video files leave your machine for OpenAI’s APIs.

The interesting bit The project is essentially a tidy orchestration layer around two OpenAI endpoints and FFmpeg, but the tidy part matters: it bundles a full audio pipeline, transcript formatting, and generative summarization into one container with no external dependencies beyond an OpenAI API key. For a tool that is mostly glue, it is remarkably self-contained.

Key highlights

• Ships as a single Docker image; the README notes it can deploy anywhere Docker runs, including AWS, Azure, GCP, or Fly.io.
• Generates standard VTT subtitle files, plus chapters and summaries derived from the transcript via ChatGPT.
• Exposes a web interface for uploading and processing videos locally.
• Configurable PHP upload limits and memory bounds via environment variables.
• Can also run from source on a machine with PHP 8.1+ and npm.

Caveats

• The project is explicitly marked as a work-in-progress.
• The Docker image currently exposes only the insecure HTTP port, not HTTPS.
• Requires an OpenAI API key; all audio transcription and text generation happen remotely, not locally.

Verdict Worth a look if you produce long videos and want a quick, containerized way to generate subtitles and chapter markers without touching ffmpeg or a text editor. Skip it if you need offline processing, local LLM inference, or enterprise-grade security out of the box.