parakeet-rs and parakeet.cpp
These are **competitors** offering mutually exclusive implementation choices: Rust for broader cross-platform performance and streaming capabilities versus C++ for Apple Silicon optimization, requiring developers to select one based on their target deployment environment.
About parakeet-rs
altunenes/parakeet-rs
very fast speech-to-text, diarization, streaming (even in CPU) with NVIDIA Parakeet in Rust
Provides multiple specialized ASR architectures via ONNX Runtime: CTC for English with punctuation, multilingual TDT with auto-detection, streaming variants (EOU, Nemotron, Multitalker) with configurable chunk sizes, and Sortformer v2/v2.1 for speaker diarization up to 4 speakers. Supports pluggable execution providers (CUDA, TensorRT, WebGPU, DirectML, CPU) with automatic fallback, token-level timestamps, and quantized model variants (int8/int4) for reduced memory footprint.
About parakeet.cpp
Frikallo/parakeet.cpp
Ultra fast and portable Parakeet implementation for on-device inference in C++ using Axiom with MPS+Unified Memory
Supports multiple decoder types (CTC, TDT, RNNT) with optional beam search and ARPA language model fusion, plus phrase boosting for domain-specific vocabulary. Built entirely in C++ on axiom's Metal tensor compiler without external runtimes—encoder inference fuses into optimized MPSGraph operations. Offers streaming models with configurable latency, speaker diarization, VAD preprocessing, and multi-format audio (WAV/FLAC/MP3/OGG) with a flat C API for FFI bindings to Python, Swift, Go, and Rust.
Scores updated daily from GitHub, PyPI, and npm data. How scores work