TransPal-transcriber

WhisperX Slack bot for transcribing audio files

Setup

pyannote

• Accept pyannote/segmentation-3.0 user conditions
• Accept pyannote/speaker-diarization-3.1 user conditions
• Create HUGGINGFACE_ACCESS_TOKEN at hf.co/settings/tokens.

modal

• Create an account on Modal
• Save the Hugging Face token on the secrets page, naming it my-huggingface-secret, with the key HUGGINGFACE_ACCESS_TOKEN.
• Get the token id and secret

Slack API

• Create a new Slack app
• Go to "Event Subscriptions" and subscribe to message.channels
• Go to "OAuth & Permissions" and add the following scopes:
  • chat:write
  • files:read
  • files:write
  • channels:read
  • channels:history
  • groups:history
  • im:history
• Get SLACK_BOT_TOKEN from "Install App" page (starts with xoxb-)
• Get SLACK_APP_TOKEN from "Basic Information" page > "App-Level Tokens" (starts with xapp-)

Slack

• Get SLACK_BOT_CHANNEL from the channel you want the bot to watch
• Add the bot to the channel

Deployment

docker run \
 -e SLACK_BOT_TOKEN='<YOUR_SLACK_BOT_TOKEN>' \
 -e SLACK_APP_TOKEN='<SLACK_APP_TOKEN>' \
 -e SLACK_BOT_CHANNEL='<SLACK_BOT_CHANNEL_ID>'  \
 -e MODAL_TOKEN_ID='<MODAL_TOKEN_ID>' \
 -e MODAL_TOKEN='<MODAL_TOKEN_SECRET>' \
 ghcr.io/drjuchunkoo/transpal-transcriber:latest

Environment Variables

• SLACK_BOT_TOKEN - Slack bot token
• SLACK_APP_TOKEN - Slack app token
• SLACK_BOT_CHANNEL - Watch channel for the bot
• MODAL_TOKEN_ID - Modal token id
• MODAL_TOKEN_SECRET - Modal token secret

Tech Stack

• m-bain/whisperX
• pyannote/segmentation-3.0
• pyannote/speaker-diarization-3.1
• slackapi/bolt-python
• Modal