st

st is a Go CLI for speech-to-text (stt) and text-to-speech (tts) with pluggable provider integrations.

This repository currently includes:

• OpenAI provider (official Go SDK)
• Batch and streaming transcription
• ffmpeg fallback conversion for unsupported input file extensions
• Disk-backed TOML config at ~/.st/config.toml

Install

go build -o st ./cmd/st

Initialize config

./st config init

This creates ~/.st/config.toml.

Set your API key either:

• In config: openai.api_key = "..."
• Via env var (default): OPENAI_API_KEY

Usage

Transcribe audio (batch)

./st stt ./audio.mp3
./st stt ./audio.wav -o transcript.txt

Transcribe audio (streaming)

./st stt ./audio.mp3 --stream

Synthesize speech from text file

./st tts ./script.txt -o speech.mp3

Synthesize speech from raw text

./st tts -t "hello? can you hear me?" > speech.mp3
./st tts -t "hello? can you hear me?" -o speech.mp3

Provider architecture

Providers implement internal/providers.Provider and register themselves via providers.Register(name, factory).

Add another provider by:

1. Implementing the interface in a new package under internal/providers/<name>
2. Registering it in init()
3. Adding provider-specific config section support

Notes

• OpenAI upload limit is 25 MB per transcription request.
• If an input extension is unsupported, st attempts conversion via ffmpeg to wav automatically.
• Without -o, commands write to stdout.