Podcast Show Notes & Transcript Generator

About Podcast Show Notes & Transcript Generator

The Podcast Show Notes & Transcript Generator turns a finished episode file into everything its description needs — entirely on your device. Drop an audio or video file (up to 500 MB) and on-device Whisper produces a timestamped transcript with progress you can watch and cancel, plus word count, duration and words-per-minute. A local AI model (in-browser or your own Ollama / LM Studio server) then writes show notes with takeaways, a valid YouTube chapters block starting at 0:00, three pull quotes with timestamps, and five title ideas — and a built-in extractive summary works even with no AI model connected. Export the transcript as TXT or SRT and the whole package as one Markdown file. The recording never leaves your computer; only open-source model weights are downloaded.

How to use Podcast Show Notes & Transcript Generator

1. Enable on-device transcription (a one-time ~145 MB open Whisper model download, cached by your browser afterwards).
2. Drop your episode — MP3, M4A, WAV, MP4, MOV and more, up to 500 MB — and pick the spoken language for best accuracy.
3. Watch the transcript arrive with clickable timestamps; a progress bar tracks the model download and transcription, and you can cancel anytime and keep what's done.
4. Review the stats (duration, words, words/min) and the built-in key sentences, and download the transcript as TXT or SRT.
5. Connect a local text model (in-browser, or Ollama / LM Studio) and generate show notes, YouTube chapters, pull quotes and title ideas — each with one click.
6. Click “Download show notes (MD)” to get metadata, notes, chapters and quotes assembled into one Markdown file.

Frequently asked questions

Is my episode uploaded anywhere?

No. The audio is decoded, transcribed and summarized entirely on your device — the recording never leaves your computer. The only network traffic is the one-time download of open-source model weights from a public CDN, which are then cached in your browser.

How long does a one-hour episode take?

It depends on your hardware. On a desktop with WebGPU (Chrome or Edge) transcription typically runs at a small fraction of the episode's length; without WebGPU it falls back to a slower CPU path. The tool transcribes in one-minute windows, so a live progress bar shows exactly how far along it is and you can cancel and keep the partial transcript.

Which languages can it transcribe?

The multilingual Whisper model handles dozens of languages; the picker offers 18 common ones plus auto-detect. Choosing the spoken language explicitly gives noticeably better accuracy than auto-detect, which can mis-default to English.

Why is this tool desktop-only?

The Whisper speech model and the local text model are too heavy for phone browsers — they run out of memory and crash the tab. Open the tool in a desktop browser; a WebGPU-capable one (Chrome or Edge) is fastest.

How do I use the generated YouTube chapters?

Copy the chapters block and paste it into your YouTube video description. YouTube requires the first chapter to start at 0:00 (the tool enforces this) and at least three chapters, each 10 seconds or longer, for the chapter markers to appear on the player.