AutoNota

autonota is a Node CLI for downloading YouTube audio, transcribing it with OpenAI Whisper into timestamped JSON, and summarizing transcript JSON into Markdown.

Prerequisites

• Node.js 22+
• yt-dlp
• ffmpeg
• ffprobe
• OPENAI_API_KEY

Install

npm install -g autonota

Or run without installing:

npx autonota transcribe <youtube-url> --output <base-path>
npx autonota summarize <transcript-json> --output <summary-md>

Usage

autonota transcribe <youtube-url> --output <base-path>
autonota summarize <transcript-json> --output <summary-md>

Examples:

autonota transcribe "https://www.youtube.com/watch?v=..." --output ./out/demo
autonota summarize ./out/demo.transcript.json --output ./out/demo.summary.md

Commands

Transcribe

autonota transcribe --help

Flags:

• --output <basePath>: base path used to derive <base>.transcript.json
• --model <name>: Whisper model override, defaults to whisper-1
• --lang <code>: transcription language override, defaults to auto detection
• --browser <name>: optional yt-dlp --cookies-from-browser source
• --force: overwrite an existing transcript output
• --base-url <url>: optional OpenAI-compatible base URL override

Notes:

• Downloaded MP3s are stored beside the --output base path as <base>-<youtube-title-slug>.mp3 and reused if that file already exists.
• If --output ends with / or \, it is treated as a directory, producing:
  • transcript: <output-dir>/<slug>.transcript.json
  • audio: <output-dir>/<slug>.mp3
• Omit --lang to let the API auto-detect the source language.

Summarize

autonota summarize --help

Flags:

• --output <summaryPath>: explicit summary file path
• --model <name>: summary model override; defaults to gpt-5-mini (OpenAI), claude-sonnet-4-6 (--claude), or gpt-5.4-mini (--codex)
• --summary-lang <code>: summary language override, defaults to en
• --force: overwrite an existing summary output
• --base-url <url>: optional OpenAI-compatible base URL override (OpenAI path only)
• --claude: use Claude Code CLI for summarization
• --codex: use Codex CLI for summarization

Notes:

• --claude and --codex are mutually exclusive.

Progress Output

autonota transcribe streams progress for the two slow steps:

Download — yt-dlp progress lines appear under the spinner as they arrive:

↻ downloading audio
    [download]  45.2% of 32.4MiB at 1.2MiB/s ETA 00:14

Transcription — shows which chunk is being uploaded, and counts down rate-limit waits:

↻ transcribing audio
    chunk 2/5...

↻ transcribing audio
    rate limited, waiting 4m32s (attempt 2/3)...
    retrying in 3m40s...

In non-TTY contexts (pipes, CI) all progress writes to stderr; stdout remains machine-readable artifact paths only.

Artifacts

• autonota transcribe --output ./out/demo writes ./out/demo.transcript.json
• autonota summarize ./out/demo.transcript.json --output ./out/demo.summary.md writes ./out/demo.summary.md

Environment

Copy .env.example to .env if you want local environment loading through your shell setup.

Required:

• OPENAI_API_KEY

Optional:

• OPENAI_BASE_URL