Transcription Pipeline

Plugin transcription-pipeline from marketplace danielrosehill in danielrosehill/Claude-Code-Plugins. Commands read from the repo's manifest — today.

Skills (11)

• batch-pipeline

  Run the transcription pipeline over several recordings at once - the same process for each file, run concurrently, one worker per file (or per small batch). Use on "transcribe these three files", "do the same for all the voice notes in this folder", "batch these calls into minutes", or any request t

• clean-transcript

  Clean a raw ASR transcript into a readable one without losing anything - light (remove disfluencies, resolve self-corrections, punctuation, paragraphs) or heavy (also headings, de-duplication, reordering of passages the speaker flagged as out of order, and marked fixes of likely mistranscriptions).

• format-document

  Typeset a transcript or any of its reshaped versions as a PDF with Typst - a clean A4 document from Markdown, or a conversation layout with speech bubbles per speaker, title, summary and decisions (calls, meetings, interviews). Use on "make it a PDF", "nice document", "with speech bubbles", "somethi

• label-speakers

  Replace generic speaker labels (Speaker A/B, Speaker 1/2) in a diarized transcript with real names, from a mapping the user gives, from voice descriptions ("I'm the male voice, Hannah is the female voice"), or from short clips the user listens to and identifies. Use after a multi-speaker transcripti

• preprocess-audio

  Prepare a recording before transcription - optional denoise (FFT or DeepFilterNet for very noisy audio), optional speech EQ (shipped presets or the user's own), conservative silence trimming, and downsampling to mono 16 kHz in a small codec. Smaller uploads, lower cost, cleaner input. Use before any

• reshape-transcript

  Rewrite a cleaned transcript into another document type - development spec or brief, notes, summary, meeting minutes, to-do list, email, blog outline, blog post, or any format the user describes or has saved as a template. Keeps the transcript files and adds the new one. Use on "turn this into a spe

• retrieve-audio

  Collect the recording to transcribe - the newest file in a Google Drive folder the user named, a local folder or path, an S3/MinIO/Wasabi/R2 bucket, or a URL (including YouTube). Also gets a local file to where the transcription will run when the agent's tools live on another machine (staging). Firs

• run-pipeline

  Transcribe a recording end to end - fetch it, prepare it, transcribe it, clean it up, reshape it (spec, minutes, to-do, blog...), format it (PDF) and save every artefact where it belongs. The entry point for "transcribe this", "I just recorded a voice note about X", "turn this call into minutes and

• save-outputs

  Put the finished transcription artefacts where they belong - the current repo, a local folder, a Google Drive folder (with per-type subfolders), or a bucket - with consistent names, the recording itself only when the user wants it, and, in a repo, an optional "Original idea recording" link in the RE

• setup

  First-time setup and later changes to how transcription works for this user. Writes a plain preferences file (sources such as Google Drive folders, keys, default backend, OpenRouter model slots, local ASR servers, EQ preset, save targets, named output sets) and checks the API keys work. Use on "set

• transcribe-audio

  Turn audio into a raw transcript with one ASR backend - OpenRouter audio models (default, latest Gemini Flash, or a model slot from preferences), AssemblyAI (long-form, timestamps, speaker labels), Gemini direct (long files via Files API), or a local/self-hosted OpenAI-compatible server (Whisper, Pa

Similar plugins