OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Audio & Voice AI

Parakeet Rs

Parakeet Rs is a Rust library that runs speech-recognition and speaker-diarization models, supporting batch transcription, streaming audio-to-text and timestamped or speaker-attributed output.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Parakeet Rs — OpenViST project profile in Audio & Voice AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when a Rust application needs to process audio with supported local model files and needs transcription, streaming output, timestamps or speaker attribution.

Who it is for

  • Rust developers adding speech recognition to an application.
  • Teams building streaming transcription or diarized transcript workflows.
  • Developers evaluating ONNX Runtime or the experimental pure-Rust Burn inference backend.

Documented uses

  • Convert audio samples into text.
  • Process audio in chunks for streaming transcription.
  • Return word or sentence timestamps where supported by the selected model.
  • Distinguish speakers in supported diarization and multitalker workflows.
Before you choose

Requirements & limits

  • ONNX Runtime is the default backend and requires one shared environment per process.
  • Burn is presented as an experimental pure-Rust backend.
  • Model capabilities differ: CTC is English-only, while TDT documents 25-language support and automatic detection.
  • Streaming examples specify model-dependent chunk sizes; the EOU example uses normalized 16 kHz mono audio.
  • The library repository is MIT-licensed, but selected model weights may have separate licenses.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Transcribe sample audio with the CTC model

Load the documented CTC model and retrieve a transcript with word-level timestamps.

Before you start

  • A Rust project using parakeet-rs and the ONNX Runtime backend.
  • The CTC model files: model.onnx, model.onnx_data and tokenizer.json.
  • Audio samples supplied to the transcription call.

Steps

  1. Initialize the ONNX Runtime environment and create an ExecutionConfig that uses it.

  2. Load the CTC model with Parakeet::from_pretrained, pointing it to the local model directory.

  3. Call transcribe_samples with the audio and TimestampMode::Words, then read the result text and token timestamps.

Expected outcome

A transcript and token-level start/end timestamps from the CTC model.

Guide limitations

  • The documented CTC model is English-only.
  • The README’s CTC example does not specify a complete audio-preparation recipe.
Documentation 1 ↗
Source-based practical guide

Process audio in chunks with EOU streaming

Feed 16 kHz mono audio to the documented EOU streaming recognizer in 160 ms chunks.

Before you start

  • A Rust project using parakeet-rs with the ONNX Runtime backend.
  • EOU model files: encoder.onnx, decoder_joint.onnx and tokenizer.json.
  • Normalized 16 kHz mono audio represented as Vec.

Steps

  1. Initialize the shared ONNX Runtime environment and create an ExecutionConfig.

  2. Load ParakeetEOU from the local EOU model directory.

  3. Split the audio into 2,560-sample chunks and pass each chunk to transcribe, printing returned text.

Expected outcome

Text is produced as successive audio chunks are processed.

Guide limitations

  • The documented chunk size assumes 16 kHz audio.
  • The example does not describe audio capture or end-to-end application deployment.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
13 October 2026
Original website title
altunenes/parakeet-rs · Official homepage title · retained in its source language
How we review sourcesSuggest a correction