Skip to content

[Suggestion] GitHub Actions workflow for automated audio and video file transcription (TypeScript) #307

Description

@deepgram-robot

What to build

A reusable GitHub Actions workflow that automatically transcribes audio and video files when they are added to a repository — useful for transcribing meeting recordings, podcast episodes, lecture videos, or any media files committed to a repo. Outputs transcripts as Markdown files alongside the source media, with optional Audio Intelligence features (summary, topics, chapters).

Why this matters

Developer teams increasingly store meeting recordings, design review videos, and podcast episodes in Git repositories or as release assets. Manually transcribing these is tedious and often skipped. A GitHub Actions workflow that runs automatically on push (when media files are added) brings transcription into the CI/CD workflow developers already use daily. This integration reaches developers who want transcription as infrastructure rather than a manual step, and demonstrates Deepgram working in a headless, server-side automation context.

Suggested scope

  • Language: TypeScript (GitHub Actions)
  • Deepgram APIs: Pre-recorded STT, Audio Intelligence (summarize, topics, detect_language)
  • What to include:
    • Composite GitHub Action with action.yml
    • Trigger: on push when .mp3, .wav, .mp4, .webm, .m4a files are added or modified
    • Transcription output as .transcript.md files alongside source media
    • Optional Audio Intelligence: summary block, topic tags, detected language
    • Configurable via action inputs: model, language, Intelligence features
    • PR comment with transcription summary when triggered by pull request
    • Example workflow file for easy adoption

Acceptance criteria

  • Runnable with minimal setup (add Deepgram API key as GitHub secret, copy workflow file)
  • README explains setup, configuration options, and output format
  • Uses current SDK version
  • Handles common audio/video formats (mp3, wav, mp4, webm, m4a)
  • Generates clean Markdown transcripts with timestamps and speaker labels
  • Works as both a push trigger and a reusable workflow

Raised by the DX intelligence system.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions