3. A speech-to-text workflow for Cornish

This sounds like one of your most valuable posts. You already have a Cornish speech-to-text process that is roughly “95% accurate”, involving recording audio, Whisper, and NotebookLM with Cornish grammar sources.

Why people would read it:

Automatic transcription for Cornish is a huge practical problem. Even a partly manual workflow would be very interesting to teachers, researchers, learners, and language activists.

You could cover:

• recording Cornish audio
• running Whisper
• using a Python script such as your transcribe1.py
• pasting the output into NotebookLM
• using 19 Cornish grammar sources to help regularise/correct the transcript
• what works well
• what remains slow or fiddly
• why this is not yet a one-click system