transcribe

name	transcribe
description	Transcribe audio files with speaker diarization via AssemblyAI. Supports local files and Plaud AI cloud recordings.
triggers	transcribe, transkribiere, transcription, transkription, plaud, diarization, speaker recognition
requires_standards	["english-only"]

/transcribe

Transcribe audio with speaker diarization via AssemblyAI API. Supports local files and Plaud AI cloud recordings.

Workflow

Step 1: Run transcription

# Local file
transcribe ~/recording.mp3

# With language hint
transcribe ~/meeting.mp3 --language de

# Plaud AI: list recordings
transcribe --plaud

# Plaud AI: transcribe specific recording
transcribe --plaud --id a8747e81

Step 2: Ask user about speakers

After the transcription completes, the CLI outputs the transcript with generic labels (Speaker A, Speaker B, ...) and saves it to ~/.local/share/transcribe/.

Always ask the user:

X Speaker erkannt. Wer ist wer?

• Speaker A: (erster Satz zitieren)
• Speaker B: (erster Satz zitieren)

Step 3: Replace speaker labels in saved file

Once the user provides names, use Edit to replace labels in the saved transcript file:

Speaker A → Malte
Speaker B → Stefan

Replace all occurrences of Speaker A and Speaker B in the saved markdown file.

Alternative: If the user already knows the speakers upfront, pass them directly:

transcribe ~/meeting.mp3 --speakers "Malte,Stefan"

This maps Speaker A → first name, Speaker B → second name (in order of first appearance).

Storage

Transcripts are auto-saved to ~/.local/share/transcribe/:

~/.local/share/transcribe/
├── 2026-03-27_combined-recording.md
├── 2026-03-28_besprechung-praxissoftware.md
└── ...

Credentials

Stored in ~/.config/transcribe/.env (primary) with 1Password fallback:

Secret	Env Var	1Password Path
AssemblyAI API Key	ASSEMBLYAI_API_KEY	op://API Keys/AssemblyAI/API Key
Plaud AI Token	PLAUD_TOKEN	op://API Keys/Plaud AI/API Key

CLI Location

~/.local/bin/transcribe

Supported Formats

MP3, WAV, OGG/Opus, M4A, FLAC, WebM, MP4 (audio track)

Pricing

• AssemblyAI Universal-3 Pro + Diarization = $0.23/h
• With Speaker ID add-on = $0.25/h
• Free tier: 185h pre-recorded
• 80-min meeting costs ~$0.33
• Cost is shown in the transcript header

Plaud AI Notes

• EU region (api-euc1.plaud.ai) — auto-detected from JWT
• Plaud token is a long-lived JWT (~10 months), extracted from web.plaud.ai localStorage
• Short IDs supported (e.g. a8747e81 matches full UUID)