crispasr 0.8.22

Lightweight on-device speech recognition via ggml. Supports Whisper, Qwen3-ASR, FastConformer, and more.

crispasr

Dart / Flutter FFI bindings for CrispASR — lightweight on-device speech recognition and speech tooling via ggml.

The native project currently ships 43 ASR backends, 48 TTS engines, language detection, punctuation, diarization, translation and chat/text helpers. This package exposes the Dart FFI API and expects the native libcrispasr library to be installed or bundled by the application.

Install

dependencies:
  crispasr: ^0.8.10

This package is pure Dart FFI and does not bundle the native library. Install libcrispasr separately or ship it with your app:

git clone https://github.com/CrispStrobe/CrispASR
cd CrispASR
cmake -B build && cmake --build build -j
sudo cmake --install build   # installs libcrispasr.{so,dylib,dll}

If libcrispasr is in a non-standard location, pass libPath: to the constructors.

Quick start

import 'package:crispasr/crispasr.dart';

final model = CrispASR.open('ggml-base.en.bin');
final segments = model.transcribe('audio.wav');
for (final seg in segments) {
  print('[${seg.start}s - ${seg.end}s] ${seg.text}');
}
model.close();

See the main repo for full API docs, the model registry, and the CLI.

License

MIT — see LICENSE.