crispasr

Safe Rust wrapper for CrispASR — lightweight on-device speech recognition via ggml.

Supports 17 ASR backends including Whisper, Qwen3-ASR, FastConformer, Canary, Parakeet, Cohere, Granite-Speech, Voxtral, wav2vec2, GLM-ASR, Kyutai-STT, Moonshine, FireRed, OmniASR, and VibeVoice-ASR.

Install

crispasr and crispasr-sys are on crates.io, but the FFI crate still needs a native libcrispasr to link against. Pick one of two modes:

A — build from source (git dependency). build.rs cmakes libcrispasr for you (needs cmake + a C++ toolchain; the git checkout supplies the sources):

[dependencies]
crispasr = { git = "https://github.com/CrispStrobe/CrispASR" }

B — link a pre-built library (crates.io dependency). Build/install libcrispasr once, then point CRISPASR_LIB_DIR at it — this is required because the crates.io package does not vendor the C/C++ sources, so build.rs cannot cmake it:

[dependencies]
crispasr = "0.8"

git clone https://github.com/CrispStrobe/CrispASR
cd CrispASR && cmake -B build && cmake --build build -j && sudo cmake --install build
# then, in your project:  export CRISPASR_LIB_DIR=/usr/local/lib

Quick start

use crispasr::CrispAsr;

let model = CrispAsr::open("ggml-base.en.bin")?;
for seg in model.transcribe_file("audio.wav")? {
    println!("[{:.1}s - {:.1}s] {}", seg.start, seg.end, seg.text);
}

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

License

MIT — see LICENSE.