Skip to content

No. 1In ASR Technologyaccording to independent research by the University of Łódź

Speech Recognition and Audio Transcription with Moonlight

Moonlight is our proprietary automatic speech recognition (ASR) engine that converts speech to text. It supports both 8 kHz phone calls and any audio recording above 8 kHz.

Sign in and get started

1hFree transcription, no credit card required.

Live Transcription
  1. 0:00AgentHello, how can I help you?
  2. 0:03CustomerI'm calling about last week's delivery.
  3. 0:08AgentLet me check. May I have your order number?
  4. 0:13CustomerIt's order 48213.
  5. 0:16AgentThank you. Your package is on its way and will arrive tomorrow by noon.
  6. 0:22CustomerGreat, thanks for your help.

VoiceLab in Numbers

  • 6LanguagesPolish, Danish, English, German, Russian, Ukrainian
  • 3ProtocolsWebSocket and gRPC for live audio, HTTP for files
  • 8Speakers in One RecordingWe recognize up to 8 speakers, even from a single audio track.

Get Started in 3 Steps

No need to book a demo. Sign up and start right away.

  1. Sign Up

    Your account is created automatically, with 1 hour of free audio.

  2. Create an API Key

    Keys are shown only once and can be revoked at any time.

  3. Send Your Audio

    Send your recording with a POST request and get the transcript in seconds.

Everything You Need for Speech Recognition

Models tuned to real phone calls and wideband recordings, available through a simple API.

  • Processing in Poland

    Speech recognition runs on VoiceLab servers in the EU. We work on a zero-retention model: recordings and transcripts are never stored or used to train models.

  • Recordings and Live Audio

    Recordings up to 60 minutes and 256 MB (WAV, MP3, FLAC, OGG/Opus, M4A, WebM, AMR) or a live stream over WebSocket in the OpenAI Realtime format, with text in real time.

  • Telephony and Wideband Models

    8 kHz for phone calls and contact centers, 16 kHz for microphones, meetings, and media. Text with punctuation, word timestamps, and confidence scores.

  • Speaker Recognition

    Speaker diarization (Speaker A, B…) at no extra cost, with an AI model that corrects recognition errors. Stereo and multichannel recordings are transcribed separately for each channel, so agent and customer never get mixed up.

Audio Transcription Pricing

Choose the plan that fits your needs. Transcription is billed per second of audio. Pay by BLIK, bank transfer, or card.

Transcription hour packages
PackageHoursPricePer HourChoose a package
25 hrsPLN 15PLN 0.60/hrPLN 0.60/hrChoose
100 hrsPLN 59PLN 0.59/hrPLN 0.59/hrChoose
500 hrsPLN 249PLN 0.50/hrPLN 0.50/hrChoose
1,500 hrsPLN 649PLN 0.43/hrPLN 0.43/hrChoose
5,000 hrsPLN 1,990PLN 0.40/hrPLN 0.40/hrChoose

Predictable Costs

You're billed for every second of audio transcribed. Track your balance, usage, and every request in your dashboard.