anyscribecli 0.8.2

CLI tool to download, transcribe, and convert video/audio to structured markdown

anyscribecli

Download. Transcribe. Markdown. Turns YouTube videos, Instagram reels, and local audio/video files into structured, searchable markdown — browsable in Obsidian.

Three equivalent surfaces, pick whichever fits:

• Web UI (scribe ui) — click-through, for humans who prefer browsers.
• Terminal wizard (scribe onboard, scribe "<url>") — arrow-key prompts, for humans who live in terminals.
• Headless flags (scribe onboard --provider X --yes --json, scribe "<url>" --json) — flag-driven, structured output, for AI agents / CI / scripts.

None of the three is "primary"; all three cover the full product. Human users reach the full product through Web UI or terminal without ever needing the other. Agents reach it through flags alone. Shared backend, shared state — a transcription started from any surface is visible to all of them.

Private by default, local-first

• The Web UI is a local server, not a cloud service. It runs at 127.0.0.1:8457 on your own machine. No account, no sign-up, no telemetry. There is no "anyscribe.com" backend.
• Internet is only involved when you ask for it. Three cases, all transparent:
  • Downloading a YouTube or Instagram source — obvious; you gave it the URL.
  • Calling an API provider (OpenAI, Deepgram, ElevenLabs, Sarvam, OpenRouter) — your audio goes to the provider you picked, and nothing else. Your data stays between you and them.
  • Pulling a Whisper model (one-time, only if you enable local transcription) — weights download from Hugging Face.
• Fully offline is available. Local files + the local provider (scribe local setup --model base) = zero network traffic. Your audio never leaves your machine. Same pipeline, same output format as the cloud providers.
• No analytics, no phone-home. scribe update --check reaches PyPI to compare versions, but only when you run it.

---

What it does

URL or local file → Download/convert audio → Transcribe → Formatted Markdown → Obsidian Vault

• 6 transcription providers — OpenAI Whisper, Deepgram Nova, ElevenLabs, OpenRouter, Sarvam AI, Local (offline)
• Speaker diarization — --diarize flag for multi-speaker transcripts (meetings, interviews, podcasts)
• 3 input sources — YouTube, Instagram (reels + posts), local files (mp3, mp4, m4a, wav, opus, ogg, flac, webm)
• Obsidian-native output — YAML frontmatter, word count, reading time, tags
• Master index + daily logs — browse everything in Obsidian
• Download-only mode — grab video or audio without transcribing
• Batch processing — transcribe a list of URLs from a file
• Web UI — scribe ui launches a local dashboard (transcribe, browse history, manage settings, first-run onboarding wizard) at 127.0.0.1:8457 — served from your own machine, no cloud backend
• Local-first, no account — no sign-up, no telemetry, no SaaS layer; fully offline with the local provider + local files
• Agent-friendly CLI — --json output, structured exit codes, --yes for non-interactive runs on every consequential command; no silent defaults for choices an agent might make on the user's behalf
• Three-surface onboarding parity — wizard modal in the Web UI, interactive prompts in scribe onboard, flag-driven in scribe onboard --yes ...; all three write the same config

Quick Start

Install

macOS / Linux (one command — installs Python, ffmpeg, and scribe):

curl -fsSL https://raw.githubusercontent.com/rishmadaan/anyscribecli/main/install.sh | bash

Windows (PowerShell — installs Python, ffmpeg, and scribe):

irm https://raw.githubusercontent.com/rishmadaan/anyscribecli/main/install.ps1 | iex

Or install manually:

pip install anyscribecli

Get started

scribe ui    # opens web dashboard — guides you through setup

On first launch the web UI opens a full-screen onboarding wizard — pick a provider, paste the API key (with a live Test button), optionally enable offline transcription, confirm your workspace, done.

Windows: If scribe isn't recognized, use python -m anyscribecli ui

Alternative paths (same end state):

• scribe onboard — interactive terminal wizard.
• scribe onboard --provider openai --api-key "$OPENAI_API_KEY" --yes --json — headless mode for agents, CI, and scripts.

Transcribe

# From a URL
scribe "https://www.youtube.com/watch?v=VIDEO_ID"

# From a local file
scribe /path/to/podcast.mp3

Always wrap URLs in quotes — shells like zsh break URLs with ? and &.

Download (no transcription)

scribe download "https://www.youtube.com/watch?v=VIDEO_ID"            # video
scribe download "https://www.youtube.com/watch?v=VIDEO_ID" --audio-only  # audio

Commands

Command	Description
scribe onboard	Interactive setup wizard (TUI)
scribe onboard --provider X --api-key $KEY --yes	Headless setup for agents / CI
scribe transcribe "<url or file>"	Transcribe a video or local file to markdown
scribe download "<url>"	Download video or audio only
scribe batch <file>	Batch transcribe URLs or file paths from a file
scribe config show/set/path	View and change settings
scribe providers list/test	Manage transcription providers
scribe local setup --model <size>	Install faster-whisper + download a Whisper model
scribe local status / scribe local teardown	Report / remove offline transcription
scribe model list / pull / rm / reinstall / info	Manage cached Whisper model weights
scribe ui	Launch the web UI in your browser
scribe install-skill	Install Claude Code skill
scribe update	Update to the latest version
scribe doctor	Check system health

Transcribe options

scribe transcribe "<url>"
  --provider, -p <name>    # Override provider (openai, deepgram, elevenlabs, local, etc.)
  --language, -l <code>    # Language code (default: auto-detect)
  --diarize, -d            # Enable speaker diarization (multi-speaker transcripts)
  --json, -j               # JSON output for scripting/AI agents
  --keep-media             # Keep the downloaded audio file
  --clipboard, -c          # Read URL from clipboard
  --quiet, -q              # Suppress progress output

Provide a URL, file path, or use interactive mode:

scribe transcribe "https://..."     # quoted URL (primary)
scribe transcribe /path/to/file.mp3 # local audio/video file
scribe transcribe                    # interactive prompt (no quoting needed)
scribe transcribe --clipboard        # read URL from system clipboard

Download options

scribe download "<url>"
  --video / --audio-only     # Video (default) or audio only
  --json, -j                 # JSON output
  --quiet, -q                # Suppress progress
  --clipboard, -c            # Read URL from clipboard

Batch options

scribe batch <file>
  --provider, -p <name>      # Override provider
  --language, -l <code>      # Override language
  --json, -j                 # JSON output
  --keep-media               # Keep audio files
  --quiet, -q                # Suppress progress
  --stop-on-error            # Stop at first failure

JSON output

scribe transcribe "https://youtube.com/watch?v=abc123" --json

{
  "success": true,
  "file": "~/anyscribe/sources/youtube/video-title.md",
  "title": "Video Title",
  "platform": "youtube",
  "duration": "12:34",
  "language": "en",
  "word_count": 1500,
  "provider": "openai"
}

Prerequisites

The onboarding wizard checks for these and offers to install them:

Dependency	Required	Install
Python 3.10+	Yes	python.org
yt-dlp	Yes	brew install yt-dlp or pip install yt-dlp
ffmpeg	Yes	brew install ffmpeg or ffmpeg.org
API key	Yes (for cloud providers)	See Provider Guide

Directory structure

~/anyscribe/                              # Obsidian vault (configurable)
├── _index.md                             # Master index (newest first)
├── sources/
│   ├── youtube/<slug>.md
│   ├── instagram/<slug>.md
│   └── local/<slug>.md
└── daily/YYYY-MM-DD.md

~/.anyscribecli/                          # App internals (hidden)
├── config.yaml                           # Settings (no secrets)
├── .env                                  # API keys + passwords
├── downloads/                            # Downloads (separate from vault)
│   ├── audio/<platform>/                 # Kept audio (if keep_media=true)
│   └── video/<platform>/                 # Downloaded videos
├── sessions/                             # Login sessions
└── logs/                                 # Processing logs

Workspace is visible and configurable — transcripts default to ~/anyscribe/ (no hidden dot-dir). Change it with scribe config set workspace_path /your/path. Downloads stay separate to keep the vault lightweight.

Providers

Provider	Best for	API key
OpenAI Whisper (default)	General purpose, multilingual	OPENAI_API_KEY
Deepgram Nova	Diarization (auto-selected with --diarize), Hinglish	DEEPGRAM_API_KEY
ElevenLabs Scribe	High accuracy, 99 languages, word timestamps	ELEVENLABS_API_KEY
Sarvam AI	Indic languages (Hindi, Tamil, Telugu, etc.)	SARGAM_API_KEY
OpenRouter	Access to various AI models	OPENROUTER_API_KEY
Local (faster-whisper)	Offline, free, no API key needed	None

See Provider Guide for detailed comparison, pricing, and setup.

Configuration

# ~/.anyscribecli/config.yaml
provider: openai          # Transcription provider
language: auto             # Language (auto-detect or ISO code)
keep_media: false          # Keep audio files after transcription
output_format: clean       # clean | timestamped | diarized
diarize: false             # enable speaker diarization
prompt_download: never     # never | ask | always — download video after transcription
local_file_media: skip     # skip | copy | move | ask — what to do with local files
workspace_path: ""         # empty = ~/anyscribe (default), or set a custom path

API keys and passwords live in ~/.anyscribecli/.env (separate from config, never committed). You can set API keys directly:

scribe config set deepgram_api_key YOUR_KEY
scribe config set openai_api_key YOUR_KEY

Diarization auto-routing: When you use --diarize without specifying a provider, scribe automatically switches to Deepgram (if configured) for best speaker detection. Override with -p openai if needed.

Web UI labels: The CLI's --diarize flag is shown as Multi-speaker in the web UI, and the diarized output format is labelled with-speaker-labels. Wire values are unchanged — the rename is display-only so the UI reads in plain English.

See Configuration Guide for all options.

Claude Code Integration

scribe ships with a Claude Code skill that teaches Claude how to transcribe, configure providers, and troubleshoot on your behalf. After installing scribe:

scribe install-skill

Or run scribe onboard — it auto-detects Claude Code and offers to install the skill. Once installed, Claude can use /scribe or auto-activate when you ask it to transcribe something.

Documentation

For	Where
First-time users	Getting Started
Command reference	Commands
All config options	Configuration
Provider comparison	Providers
AI developers	CLAUDE.md
Agent directives	AGENTS.md
Developer memory	Building Docs

Development

git clone https://github.com/rishmadaan/anyscribecli.git
cd anyscribecli
pip install -e ".[dev]"

ruff check src/          # lint
ruff format src/         # format
pytest                   # test

License

MIT