Picovoice/cheetah

Cheetah

GitHub release GitHub

Maven Central Maven Central npm npm npm npm Nuget CocoaPods Pub Version PyPI

Made in Vancouver, Canada by Picovoice

Twitter URL YouTube Channel Views

Cheetah is an on-device streaming speech-to-text engine. Cheetah is:

• Private; All voice processing runs locally.
• Accurate
• Compact and Computationally-Efficient
• Cross-Platform:
  • Linux (x86_64), macOS (x86_64, arm64), and Windows (x86_64, arm64)
  • Android and iOS
  • Chrome, Safari, Firefox, and Edge
  • Raspberry Pi (3, 4, 5)

Table of Contents

• Cheetah
  • Table of Contents
  • AccessKey
  • Language Support
  • Demos
    • Python
    • C
    • iOS
    • Android
    • Flutter
    • React Native
    • Java
    • Node.js
    • .Net
    • Web
      • Vanilla JavaScript and HTML
      • React
  • SDKs
    • Python
    • C
    • iOS
    • Android
    • Flutter
    • React Native
    • Node.js
    • Java
    • .Net
    • Web
      • Vanilla JavaScript and HTML (ES Modules)
      • React
  • Releases

AccessKey

AccessKey is your authentication and authorization token for deploying Picovoice SDKs, including Cheetah. Anyone who is using Picovoice needs to have a valid AccessKey. You must keep your AccessKey secret. You would need internet connectivity to validate your AccessKey with Picovoice license servers even though the voice recognition is running 100% offline.

AccessKey also verifies that your usage is within the limits of your account. You can see your usage limits and real-time usage on your Picovoice Console Profile. To get, renew or adjust your usage limits, please reach out to our Enterprise Sales Team or your existing Picovoice contact.

Language Support

• Cheetah Streaming Speech-to-Text currently supports English, French, German, Italian, Portuguese, and Spanish.
• Support for additional languages is available for commercial customers on a case-by-case basis.

Demos

Python Demos

Install the demo package:

pip3 install pvcheetahdemo

cheetah_demo_mic --access_key ${ACCESS_KEY}

Replace ${ACCESS_KEY} with yours obtained from Picovoice Console.

C Demos

If using SSH, clone the repository with:

git clone --recurse-submodules git@github.com:Picovoice/cheetah.git

If using HTTPS, clone the repository with:

git clone --recurse-submodules https://github.com/Picovoice/cheetah.git

Build the demo:

cmake -S demo/c/ -B demo/c/build && cmake --build demo/c/build

Run the demo:

./demo/c/build/cheetah_demo_mic -a ${ACCESS_KEY} -m ${MODEL_PATH} -l ${LIBRARY_PATH}

Replace ${ACCESS_KEY} with yours obtained from Picovoice Console, ${LIBRARY_PATH} with the path to appropriate library under lib, and ${MODEL_PATH} to path to default model file (or your custom one).

iOS Demos

To run the demo, go to demo/ios/CheetahDemo and run:

pod install

Replace let accessKey = "${YOUR_ACCESS_KEY_HERE}" in the file ViewModel.swift with your AccessKey.

Then, using Xcode, open the generated CheetahDemo.xcworkspace and run the application.

Android Demo

Using Android Studio, open demo/android/CheetahDemo as an Android project and then run the application.

Replace "${YOUR_ACCESS_KEY_HERE}" in the file MainActivity.java with your AccessKey.

Flutter Demo

To run the Cheetah demo on Android or iOS with Flutter, you must have the Flutter SDK installed on your system. Once installed, you can run flutter doctor to determine any other missing requirements for your relevant platform. Once your environment has been set up, launch a simulator or connect an Android/iOS device.

Run the prepare_demo script from demo/flutter with a language code to set up the demo in the language of your choice (e.g. de -> German, ko -> Korean). To see a list of available languages, run prepare_demo without a language code.

dart scripts/prepare_demo.dart ${LANGUAGE}

Replace "${YOUR_ACCESS_KEY_HERE}" in the file main.dart with your AccessKey.

Run the following command from demo/flutter to build and deploy the demo to your device:

flutter run

React Native Demo

To run the React Native Cheetah demo app you will first need to set up your React Native environment. For this, please refer to React Native's documentation. Once your environment has been set up, navigate to demo/react-native/CheetahDemo to run the following commands:

For Android:

yarn android-install          # sets up environment
yarn android-run ${LANGUAGE}  # builds and deploys to Android

For iOS:

yarn ios-install              # sets up environment
yarn ios-run ${LANGUAGE}      # builds and deploys to iOS

Node.js Demo

Install the demo package:

yarn global add @picovoice/cheetah-node-demo

With a working microphone connected to your device, run the following in the terminal:

cheetah-mic-demo --access_key ${ACCESS_KEY}

For more information about Node.js demos go to demo/nodejs.

Java Demos

The Cheetah Java demo is a command-line application that lets you choose between running Cheetah on an audio file or on real-time microphone input.

To try the real-time demo, make sure there is a working microphone connected to your device. Then invoke the following commands from the terminal:

cd demo/java
./gradlew build
cd build/libs
java -jar cheetah-mic-demo.jar -a ${ACCESS_KEY}

For more information about Java demos go to demo/java.

.NET Demo

Cheetah .NET demo is a command-line application that lets you choose between running Cheetah on an audio file or on real-time microphone input.

Make sure there is a working microphone connected to your device. From demo/dotnet/CheetahDemo run the following in the terminal:

dotnet run -c MicDemo.Release -- --access_key ${ACCESS_KEY}

Replace ${ACCESS_KEY} with your Picovoice AccessKey.

For more information about .NET demos, go to demo/dotnet.

Web Demos

Vanilla JavaScript and HTML

From demo/web run the following in the terminal:

yarn
yarn start

(or)

npm install
npm run start

Open http://localhost:5000 in your browser to try the demo.

React Demo

From demo/react run the following in the terminal:

yarn
yarn start

(or)

npm install
npm run start

Open http://localhost:3000 in your browser to try the demo.

SDKs

Python

Install the Python SDK:

pip3 install pvcheetah

Create an instance of the engine and transcribe audio in real-time:

import pvcheetah

handle = pvcheetah.create(access_key='${ACCESS_KEY}')

def get_next_audio_frame():
    pass

while True:
    partial_transcript, is_endpoint = handle.process(get_next_audio_frame())
    if is_endpoint:
        final_transcript = handle.flush()

Replace ${ACCESS_KEY} with yours obtained from Picovoice Console.

To get word-level metadata, use the annotated API:

import pvcheetah

handle = pvcheetah.create(access_key='${ACCESS_KEY}')

def get_next_audio_frame():
    pass

while True:
    partial_output = handle.process_annotated(get_next_audio_frame())

    partial_transcript = partial_output.transcript
    partial_words = partial_output.words
    is_endpoint = partial_output.is_endpoint

    if is_endpoint:
        final_output = handle.flush_annotated()
        final_transcript = final_output.transcript
        final_words = final_output.words

C

Create an instance of the engine and transcribe audio in real-time:

#include <stdbool.h>
#include <stdio.h>
#include <stdlib.h>

#include "pv_cheetah.h"

pv_cheetah_t *handle = NULL;
const pv_status_t status = pv_cheetah_init("${ACCESS_KEY}", "${MODEL_PATH}", "${DEVICE}", 0.f, false, false, &handle);
if (status != PV_STATUS_SUCCESS) {
    // error handling logic
}

extern const int16_t *get_next_audio_frame(void);

while (true) {
    char *partial_transcript = NULL;
    int32_t partial_num_words = 0;
    pv_word_t *partial_words = NULL;
    bool is_endpoint = false;
    const pv_status_t status = pv_cheetah_process(
            handle,
            get_next_audio_frame(),
            &partial_transcript,
            &partial_num_words,
            &partial_words,
            &is_endpoint);
    if (status != PV_STATUS_SUCCESS) {
        // error handling logic
    }
    // do something with transcript
    pv_cheetah_transcript_delete(partial_transcript);
    pv_cheetah_words_delete(partial_num_words, partial_words);
    if (is_endpoint) {
        char *final_transcript = NULL;
        int32_t final_num_words = 0;
        pv_word_t *final_words = NULL;
        const pv_status_t status = pv_cheetah_flush(handle, &final_transcript, &final_num_words, &final_words);
        if (status != PV_STATUS_SUCCESS) {
            // error handling logic
        }
        // do something with transcript
        pv_cheetah_transcript_delete(final_transcript);
        pv_cheetah_words_delete(final_num_words, final_words);
    }
}

Replace ${ACCESS_KEY} with yours obtained from Picovoice Console and ${MODEL_PATH} to path to default model file (or your custom one). Finally, when done be sure to release resources acquired using pv_cheetah_delete(handle).

iOS

The Cheetah iOS binding is available via CocoaPods. To import it into your iOS project, add the following line to your Podfile and run pod install:

pod 'Cheetah-iOS'

Create an instance of the engine and transcribe audio in real-time:

import Cheetah

let modelPath = Bundle(for: type(of: self)).path(
        forResource: "${MODEL_FILE}", // Name of the model file name for Cheetah
        ofType: "pv")!

let cheetah = Cheetah(accessKey: "${ACCESS_KEY}", modelPath: modelPath)

func getNextAudioFrame() -> [Int16] {
  // .. get audioFrame
  return audioFrame;
}

while true {
  do {
    let partialTranscript, isEndpoint = try cheetah.process(getNetAudioFrame())
    if isEndpoint {
      let finalTranscript = try cheetah.flush()
    }
  } catch let error as CheetahError {
      // handle error
  } catch { }
}

Replace ${ACCESS_KEY} with yours obtained from Picovoice Console and ${MODEL_FILE} with a custom trained model from Picovoice Console or the default model.

Android

To include the package in your Android project, ensure you have included mavenCentral() in your top-level build.gradle file and then add the following to your app's build.gradle:

dependencies {
    implementation 'ai.picovoice:cheetah-android:${LATEST_VERSION}'
}

Create an instance of the engine and transcribe audio in real-time:

import ai.picovoice.cheetah.*;

final String accessKey = "${ACCESS_KEY}"; // AccessKey obtained from Picovoice Console (https://console.picovoice.ai/)
final String modelPath = "${MODEL_FILE}";

short[] getNextAudioFrame() {
    // .. get audioFrame
    return audioFrame;
}

try {
    Cheetah cheetah = new Cheetah.Builder().setAccessKey(accessKey).setModelPath(modelPath).build(appContext);

    String transcript = "";

    while true {
        CheetahTranscript transcriptObj = cheetah.process(getNextAudioFrame());
        transcript += transcriptObj.getTranscript();

        if (transcriptObj.getIsEndpoint()) {
            CheetahTranscript finalTranscriptObj = cheetah.flush();
            transcript += finalTranscriptObj.getTranscript();
        }
    };

} catch (CheetahException ex) { }

Replace ${ACCESS_KEY} with yours obtained from Picovoice Console and ${MODEL_FILE} with the default or custom trained model from console.

Flutter

Add the Cheetah Flutter plugin to your pub.yaml.

dependencies:
  cheetah_flutter: ^<version>

Create an instance of the engine and transcribe audio in real-time:

import 'package:cheetah_flutter/cheetah.dart';

const accessKey = "{ACCESS_KEY}"  // AccessKey obtained from Picovoice Console (https://console.picovoice.ai/)

List<int> buffer = getAudioFrame();

try{
    Cheetah _cheetah = await Cheetah.create(accessKey, '{CHEETAH_MODEL_PATH}');

    String transcript = "";

    while true {
        CheetahTranscript partialResult = await _cheetah.process(getAudioFrame());
        transcript += partialResult.transcript;

        if (partialResult.isEndpoint) {
            CheetahTranscript finalResult = await _cheetah.flush();
            transcript += finalResult.transcript;
        }
    }

    _cheetah.delete()

} on CheetahException catch (err) { }

Replace ${ACCESS_KEY} with your AccessKey obtained from Picovoice Console and ${CHEETAH_MODEL_PATH} with the the path a custom trained model from Picovoice Console or the default model.

React Native

The Cheetah React Native binding is available via NPM. Add it via the following command:

yarn add @picovoice/cheetah-react-native

Create an instance of the engine and transcribe an audio file:

import {Cheetah, CheetahErrors} from '@picovoice/cheetah-react-native';

const getAudioFrame = () => {
  // get audio frames
}

try {
  while (1) {
    const cheetah = await Cheetah.create("${ACCESS_KEY}", "${MODEL_FILE}")
    const {transcript, isEndpoint} = await cheetah.process(getAudioFrame())
    if (isEndpoint) {
      const {transcript} = await cheetah.flush()
    }
  }
} catch (err: any) {
  if (err instanceof CheetahErrors) {
    // handle error
  }
}

Replace ${ACCESS_KEY} with yours obtained from Picovoice Console and ${MODEL_FILE} with the default or custom trained model from console. When done be sure to explicitly release the resources using cheetah.delete().

Node.js

Install the Node.js SDK:

yarn add @picovoice/cheetah-node

Create instances of the Cheetah class:

const Cheetah = require("@picovoice/cheetah-node");

const accessKey = "${ACCESS_KEY}"; // Obtained from the Picovoice Console (https://console.picovoice.ai/)
const endpointDurationSec = 0.2;
const handle = new Cheetah(accessKey);

function getNextAudioFrame() {
  // ...
  return audioFrame;
}

while (true) {
  const audioFrame = getNextAudioFrame();
  const [partialTranscript, isEndpoint] = handle.process(audioFrame);
  if (isEndpoint) {
    finalTranscript = handle.flush()
  }
}

Replace ${ACCESS_KEY} with yours obtained from Picovoice Console.

When done, be sure to release resources using release():

handle.release();

Java

Create an instance of the engine with the Cheetah Builder class and transcribe audio in real-time:

import ai.picovoice.cheetah.*;

final String accessKey = "..."; // AccessKey provided by Picovoice Console (https://console.picovoice.ai/)

short[] getNextAudioFrame() {
    // .. get audioFrame
    return audioFrame;
}

String transcript = "";

try {
    Cheetah cheetah = new Cheetah.Builder().setAccessKey(accessKey).build();

    while true {
        CheetahTranscript transcriptObj = cheetah.process(getNextAudioFrame());
        transcript += transcriptObj.getTranscript();

        if (transcriptObj.getIsEndpoint()) {
            CheetahTranscript finalTranscriptObj = cheetah.flush();
            transcript += finalTranscriptObj.getTranscript();
        }
    }

    cheetah.delete();

} catch (CheetahException ex) { }

.NET

Install the .NET SDK using NuGet or the dotnet CLI:

dotnet add package Cheetah

The SDK exposes a factory method to create instances of the engine as below:

using Pv;

const string accessKey = "${ACCESS_KEY}";

Cheetah handle = Cheetah.Create(accessKey);

Replace ${ACCESS_KEY} with yours obtained from Picovoice Console.

When initialized, the valid sample rate is given by handle.SampleRate. Expected frame length (number of audio samples in an input array) is handle.FrameLength. The engine accepts 16-bit linearly-encoded PCM and operates on single-channel audio.

short[] GetNextAudioFrame()
{
    // .. get audioFrame
    return audioFrame;
}

string transcript = "";

while(true)
{
    CheetahTranscript transcriptObj = handle.Process(GetNextAudioFrame());
    transcript += transcriptObj.Transcript;

        if (transcriptObj.IsEndpoint) {
        CheetahTranscript finalTranscriptObj = handle.Flush();
        transcript += finalTranscriptObj.Transcript;
    }
}

Cheetah will have its resources freed by the garbage collector, but to have resources freed immediately after use, wrap it in a using statement:

using(Cheetah handle = Cheetah.Create(accessKey))
{
    // .. Cheetah usage here
}

Web

Vanilla JavaScript and HTML (ES Modules)

Install the web SDK using yarn:

yarn add @picovoice/cheetah-web

or using npm:

npm install --save @picovoice/cheetah-web

Create an instance of the engine using CheetahWorker and transcribe an audio file:

import { CheetahWorker } from "@picovoice/cheetah-web";
import cheetahParams from "${PATH_TO_BASE64_CHEETAH_PARAMS}";

let transcript = "";

function transcriptCallback(cheetahTranscript: CheetahTranscript) {
  transcript += cheetahTranscript.transcript;
  if (cheetahTranscript.isEndpoint) {
    transcript += "\n";
  }
}

function getAudioData(): Int16Array {
  // ... function to get audio data
  return new Int16Array();
}

const cheetah = await CheetahWorker.create(
  "${ACCESS_KEY}",
  transcriptCallback,
  { base64: cheetahParams }
);

for (;;) {
  cheetah.process(getAudioData());
  // break on some condition
}
cheetah.flush(); // runs transcriptionCallback on remaining data.

Replace ${ACCESS_KEY} with yours obtained from Picovoice Console. Finally, when done release the resources using cheetah.release().

React

yarn add @picovoice/cheetah-react @picovoice/web-voice-processor

(or)

npm install @picovoice/cheetah-react @picovoice/web-voice-processor

import { useCheetah } from "@picovoice/cheetah-react";

function App(props) {
  const {
    result,
    isLoaded,
    isListening,
    error,
    init,
    start,
    stop,
    release,
  } = useCheetah();

  const initEngine = async () => {
    await init(
      "${ACCESS_KEY}",
      cheetahModel,
    );
  };

  const toggleRecord = async () => {
    if (isListening) {
      await stop();
    } else {
      await start();
    }
  };

  useEffect(() => {
    if (result !== null) {
      console.log(result.transcript);
      console.log(result.isComplete);
    }
  }, [result])
}

Releases

v4.1.0 - June 24th, 2026

• Added annotated transcriptions that include word timestamps and confidences

v4.0.0 - March 19th, 2026

• Improved engine performance and latency
• Added text normalization feature

v3.0.0 - December 17th, 2025

• Improved engine performance
• Added support for running on GPU or multiple CPU cores
• Node.js min version bumped to Node 18
• iOS min version bumped to iOS 16
• Flutter min version bumped to 3.10
• React Native min version bumped to 0.73

v2.3.0 - August 29th, 2025

• Added support for low latency Cheetah Fast models
• Improved engine accuracy

v2.2.0 - July 3rd, 2025

• Improved automatic punctuation insertion

v2.1.0 - December 10th, 2024

• Added language support for French, German, Italian, Portuguese and Spanish
• Various bug fixes and performance improvements

v2.0.0 - November 27th, 2023

• Improvements to error reporting
• Upgrades to authorization and authentication system
• Improved engine accuracy
• Various bug fixes and improvements
• Node min support bumped to Node 16
• Bumped iOS support to iOS 13+
• Patches to .NET support

v1.1.0 - August 11th, 2022

• Added true-casing by default for transcription results
• Added option to enable automatic punctuation insertion
• Cheetah Web SDK release

v1.0.0 - January 25th, 2022

• Initial release