This is a system I use to take notes on a voice recorder and have them transcribed for me into my Notion planner automatically.
This works by archiving mp3s on S3, transcribing them with AWS Transcribe, structuring the transcripts and attaching useful metadata including timestamps into the original media, and synchronizing the finished product into my agenda and notes page.
If you are interested, the code should be straightforward and self explanatory. You can look at the contents of scripts/
which serves as a collection of entrypoints into the project.
You need an IAM account with S3 and AWS Transcribe privileges. You also need an Notion integration and associated keys.
You should place these in ~/.voice-notes/secrets/ in environment files according to the expectations of the configuration
class in voice_notes/config.py.
Briefly these environment variables required are:
NOTION_TOKEN=: The Notion integration API key.
AWS_SLUG=: A URL prefix which is required to generate media links in Notion. The Notion API unfortunately does not support file uploads yet.
AWS credentials are managed through ~/.aws conventions, as is typical. It would be simple to refactor, if needed, to provide these credentials via the secrets .env file, so long as you modify voice_notes/config.py:Config.__post_init__ to
load the profile from environment variables rather than through the profile name.
A few other pieces of configuration are hardcoded. Some, like the name of the bucket on S3 used for mp3 storage, can be refactored. Others, like my conventions on the structure of my Notion agenda page require just a little bit more work.
$> poetry run python scripts/import_all.py