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

The crates are not published on crates.io — depend on them via git. The crispasr-sys build.rs needs the CrispASR C/C++ sources (to build libcrispasr with cmake), which only a repo checkout provides:

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

By default crispasr-sys's build.rs builds libcrispasr from source with cmake (needs cmake + a C++ toolchain; the git dependency supplies the sources). To skip that and link a pre-built library instead, build/install it once and point CRISPASR_LIB_DIR at it:

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.