speak2text

speak2text CLI tool. Drop audio or video files into input/, run s2t transcribe, get transcripts in output/. Powered by OpenAI Whisper.

Installation

pnpm add -g speak2text

Requirements

• Node.js 22+
• OpenAI API key (or Grok / Gemini)
• ffmpeg - only needed for unsupported formats (mkv, avi, flac, ogg...)

Install ffmpeg (if needed)

macOS

brew install ffmpeg

Linux (Debian/Ubuntu)

sudo apt install ffmpeg

Windows

winget install ffmpeg

Configuration

Create ~/.speak2text/.env and add your API key:

OPENAI_API_KEY=your-key-here

Other providers:

GROK_API_KEY=your-key-here
GEMINI_API_KEY=your-key-here

Set default provider (OpenAI is default):

s2t config set provider openai

How It Works

1. Put any audio or video file in input/
2. Run s2t transcribe
3. Transcripts appear in output/

Files sent directly to OpenAI: mp3, mp4, m4a, wav, webm

Other formats (mkv, avi, flac, ogg...) are converted to mp3 via ffmpeg first. If ffmpeg cannot process the file, an error is shown.

Usage

Transcribe

s2t transcribe

Processes all files in input/, writes transcripts to output/.

$ s2t transcribe
Transcribing 1 file(s)...
  chile_promo.mp3 - transcribing...
  ✓ output/chile_promo.txt [19dee66d]

Done.

Output:

September 11, 1973, a military coup overthrows the government in Chile, ending the longest
democratic tradition in Latin America. It was a bloody, bloody coup.

Transcribe a single file:

s2t transcribe path/to/audio.mp3

Options

s2t transcribe --format srt
s2t transcribe --format json
s2t transcribe --language fi
s2t transcribe --translate fi
s2t transcribe --translate en
s2t transcribe --provider gemini
s2t transcribe --out my-transcript.txt
s2t transcribe --out transcripts/

Translation

s2t transcribe --translate fi
s2t transcribe --translate en
s2t transcribe --translate sv

--translate en uses Whisper's native translation endpoint - one call, fast and cheap.

--translate <other> transcribes first, then translates via GPT-4o-mini.

Translation requires OpenAI provider.

Example - English audio translated to Finnish:

$ s2t transcribe --translate fi
Transcribing 1 file(s)...
  chile_promo.mp3 - transcribing...
  ✓ output/chile_promo.txt [acc2fb12]

Done.

Output:

11. syyskuuta 1973 sotilasvallankaappaus kaataa hallituksen Chilessä, mikä päättää
Latinalaisen Amerikan pisimmän demokraattisen perinteen. Se oli verinen, verinen vallankaappaus.

Manage transcripts

s2t list
s2t show <id>
s2t export <id> --format srt
s2t delete <id>
s2t delete --all

Database

s2t db stats
s2t db list
s2t delete --all

s2t db stats shows transcript count, database file size, provider breakdown, and oldest/newest dates.

s2t db list shows each source file that has been transcribed, with transcript count and last transcribed date.

s2t delete --all removes all stored transcripts.

List supported languages

s2t languages

Output Formats

| Format | Description | |--------|-------------| | txt | Plain text (default) | | srt | SRT subtitles with timestamps | | json | Full JSON with timestamps, confidence, and metadata |

Providers

| Provider | Flag | Notes | |----------|------|-------| | OpenAI Whisper | --provider openai | Default. $0.006/min | | Grok | --provider grok | OpenAI-compatible API | | Gemini | --provider gemini | OpenAI-compatible API |

Storage

Transcripts are stored locally in SQLite:

• macOS/Linux: ~/.speak2text/transcripts.db
• macOS/Linux config: ~/.speak2text/.env

Supported Languages

99 languages supported. Use the code with --language or --translate.

| Code | Language | Code | Language | Code | Language | |------|----------|------|----------|------|----------| | af | Afrikaans | hi | Hindi | pt | Portuguese | | am | Amharic | hr | Croatian | ro | Romanian | | ar | Arabic | ht | Haitian Creole | ru | Russian | | as | Assamese | hu | Hungarian | sa | Sanskrit | | az | Azerbaijani | haw | Hawaiian | sd | Sindhi | | ba | Bashkir | hy | Armenian | si | Sinhala | | be | Belarusian | id | Indonesian | sk | Slovak | | bg | Bulgarian | is | Icelandic | sl | Slovenian | | bn | Bengali | it | Italian | sn | Shona | | bo | Tibetan | ja | Japanese | so | Somali | | br | Breton | jw | Javanese | sq | Albanian | | bs | Bosnian | ka | Georgian | sr | Serbian | | ca | Catalan | kk | Kazakh | su | Sundanese | | cs | Czech | km | Khmer | sv | Swedish | | cy | Welsh | kn | Kannada | sw | Swahili | | da | Danish | ko | Korean | ta | Tamil | | de | German | la | Latin | te | Telugu | | el | Greek | lb | Luxembourgish | tg | Tajik | | en | English | ln | Lingala | th | Thai | | es | Spanish | lo | Lao | tk | Turkmen | | et | Estonian | lt | Lithuanian | tl | Tagalog | | eu | Basque | lv | Latvian | tr | Turkish | | fa | Persian | mg | Malagasy | tt | Tatar | | fi | Finnish | mi | Maori | uk | Ukrainian | | fo | Faroese | mk | Macedonian | ur | Urdu | | fr | French | ml | Malayalam | uz | Uzbek | | gl | Galician | mn | Mongolian | vi | Vietnamese | | gu | Gujarati | mr | Marathi | yi | Yiddish | | ha | Hausa | ms | Malay | yo | Yoruba | | he | Hebrew | mt | Maltese | yue | Cantonese | | nn | Nynorsk | my | Myanmar | zh | Chinese | | no | Norwegian | ne | Nepali | | | | oc | Occitan | nl | Dutch | | | | pa | Punjabi | ps | Pashto | | |

Tech Stack

| Tool | Purpose | |------|---------| | TypeScript | Language | | Commander | CLI framework | | better-sqlite3 | Local transcript history | | OpenAI Whisper API | Transcription | | ffmpeg | Converts unsupported formats before transcription (optional) | | Vitest | Testing | | Biome | Lint & format | | pnpm | Package manager |

License

See MIT