Adaptation is a list of words or phrases that biases the recognition output.
Create adaptations offline and use them later in recognition.
Accurate recognition of new and never before seen words.
Quick and cost-free process with minimal effort.
Speech Recognition Reinvented
Facts
Why Soniox?
One speech model.
Near error-free recognition.
Works on all kinds of audio.
Near error-free recognition.
Works on all kinds of audio.
Speech adaptation.
No need for data labeling and model retraining.
No need for data labeling and model retraining.
Easy-to-use APIs.
Simple integration.
Best developer resources.
Simple integration.
Best developer resources.
Near Error-Free Speech Recognition For All Audio
24%
higher accuracy than the best leading speech recognition systems.
Competition
- ONE extremely accurate speech model for ALL kinds of audio.
- No need for customized models that require expensive data labeling, model retraining and deployment.
- Accurate recognition of conversational and spontaneous speech.
- Robust recognition in noisy audio environments.
Speech Adaptation
Adapt recognition by adding context words and make it error-free.
Development
Redefined Speech Integration
No more difficult speech integration. Focus on building your application.
Simple and generic gRPC API that can be used from virtually any programming language.
Easy-to-use Python client for batch, real-time and low-latency speech recognition.
Provided solutions for most common speech recognition use cases.
1. Install Soniox Python package
pip3 install soniox2. Create Soniox Account
Create a Soniox Account to obtain an API key. The API key gives you free access to Soniox speech recognition.
Create Account3. Try It Out
from soniox.transcribe_live import transcribe_microphone
from soniox.speech_service import Client, set_api_key
set_api_key("<YOUR-API-KEY>")
def main():
with Client() as client:
print("Transcribing from your microphone ...")
for result in transcribe_microphone(client):
print(" ".join(w.text for w in result.words))
if __name__ == "__main__":
main()
Web Development
Soniox Web Voice
Make your website or app accessible and interactive via voice.
Integrate real-time low-latency speech recognition right into your website.
Easy-to-use Javascript library that requires only a few lines of code.
Free voice recognition for up to 30 seconds per session.
1. Include Soniox Javascript package
<script src="https://js.soniox.com/soniox-web-voice.js"></script>
Check out our complete code example on how to use Soniox Web Speech Javascript package.
Get Started2. Try It Out
// Create RecordTranscribe object.
var record_transcribe = soniox.RecordTranscribe()
// Set event handler.
record_transcribe.setOnPartialResult(printWords)
function printWords(result) {
for (let i = 0; i < result.words.length; ++i) {
console.log(result.words[i]);
}
}
// Start transcribing from your microphone.
record_transcribe.start()
Support
Developer
Developers can sometimes get stuck and need help to move forward.
Resources
Check out our resources page with all the information for development in one place.
Resources
Research
Breakthrough AI Technology
We developed a novel approach to training speech recognition models on unlabeled data. It is an iterative process with each iteration yielding higher accuracy models. We built a Soniox speech model with no human labeled data.
Read More
Engineering
Distributed Stream Processing
We developed a proprietary framework for distributed real-time low-latency processing of audio streams that can run in any cloud environment. To support streaming speech recognition, we built a proprietary inference engine and decoder.
Read More