Demo

Speech to Text: API Speech Recognition

Integrated into your conversational tools: best-in-class speech recognition, built for developers.

Word Error Rate< 5%Most accurate engine in France, post-LMF
Latency< 600 msOn live conversational streams, in real conditions
PricingLowestOn the market, per-minute commitment

Trusted by leading enterprises, integrators and researchers

Language Model Factory

Express language model adaptation

With our Language Model Factory (LMF), no domain vocabulary goes unrecognised. Train custom models in as little as 15 minutes and select the one that best fits your use case.

  • A production-ready model in 15 minutes
  • Jargon, proper nouns, industry acronyms
  • Pre-trained vertical models available off the shelf
01

Vertical models

Pre-trained on industry-specific verticals.

02

Jargon & proper nouns

Vocabulary tied to your business.

03

Acronyms

Automatic recognition and expansion.

04

Business expressions

Specific to your processes and use cases.

Anonymisation

Securing and redacting sensitive data

Automatic real-time redaction of sensitive information based on your use cases: personal data, banking and health records.

Agent Acme Corp: Hello, you have reached Acme Corp[company]. What can I do for you today?

Customer: Hello, I am calling to update my personal details.

Agent Acme Corp: No problem, I will help you with that. Could you give me your full name, please?

Customer: Yes, it is John Persona[first name] [last name].

Agent Acme Corp: Thank you, John[first name]. And what is your new address?

Customer: It is 123 Rue de l’Élysée, 75000 Paris[address].

Agent Acme Corp: Noted. Do you also have a new phone number?

Customer: Yes, the new one is 06 12 34 56 78[phone].

Agent Acme Corp: Perfect. One last question: have your payment details changed?

Customer: Yes, I have a new bank card: 1234 5678 9101 1121, expiry 12/26, CVV 321[bank card].

Agent Acme Corp: Thank you, John[first name]. Your details are up to date.

Batch

Speech-to-Text API for audio recordings

Simply drop your phone conversations onto a secure FTP for transcription within minutes, or use our connectors:

NICEGenesysAxialysOdigo
Available languages

French, English, Spanish, German, Italian, +5 others (Europe)

PYTHON SDK
# Secure FTP upload
import uhlive
client = uhlive.connect("api.uh.live")
# Batch transcription
job = client.transcribe_file(
    file="call_2026-04-28.wav",
    model="en-telephony-v3",
    redaction=True
)
# Result
transcript = job.result()
print(transcript.text)
Illustration: multilingual transcription for humans
Streaming · Live

Streaming API for humans

Connect your audio streams directly via WebSocket to receive real-time multi-speaker transcription, or via Trunk SIP / SIP REC.

Available languages

French, English, Spanish and German

Streaming · Bot

Streaming API for bots

Streaming API for IVRs and voicebots. Transcribe your live interactions with our advanced solutions.

Illustration: multilingual voicebot
Our protocols
  • › MRCP v2
  • › WebSocket
Built-in for every interaction

Speech activity detection, language model selection, grammars, address recognition, dates, numbers and boolean responses.

Available languages

French, English, Spanish and German

Solutions

Relational analysis: emotions, intents, compliance, summaries.

Traceable, compliant voice consent, with no friction for the caller.

WER < 5%Most accurate engine in France
100MCalls analysed per year
40%Of analyses in real time

Ready to transcribe your first calls?

Access the uh!ive Speech-to-Text API. Set up in minutes.

Try it for free