whis-core

whis-core

Core library for whis voice-to-text functionality.
Website · CLI · Desktop · Mobile

Features

• Audio recording — capture microphone input
• Transcription — cloud providers or local models
• Progressive chunking — transcribe long recordings in real-time
• LLM post-processing — transform transcripts with custom prompts
• Clipboard — copy results to system clipboard
• Config management — persistent settings (platform-aware paths)

Usage

use whis_core::{
    AudioRecorder, TranscriptionProvider, Settings,
    ProgressiveChunker, ChunkerConfig, ProgressiveChunk,
    progressive_transcribe_cloud, copy_to_clipboard, ClipboardMethod,
};
use tokio::sync::mpsc;

// Load settings and get provider config
let settings = Settings::load();
let provider = settings.transcription.provider.clone();
let api_key = settings.transcription.api_key_for(&provider).unwrap();

// Set up progressive chunking channel
let (chunk_tx, chunk_rx) = mpsc::unbounded_channel::<ProgressiveChunk>();
let chunker = ProgressiveChunker::new(ChunkerConfig::default(), chunk_tx);

// Record audio
let mut recorder = AudioRecorder::new()?;
recorder.start_recording()?;
// ... feed audio samples to chunker during recording ...
let recording = recorder.stop_recording()?;

// Transcribe progressively
let text = progressive_transcribe_cloud(
    &provider,
    &api_key,
    None, // language hint
    chunk_rx,
    None, // progress callback
).await?;

// Copy to clipboard
copy_to_clipboard(&text, ClipboardMethod::Auto)?;

For simpler use cases, see the CLI implementation in whis-cli which handles the recording → chunking → transcription → clipboard pipeline.

Feature Flags

Modules

License

MIT