Speech to Text

Transcribe voice notes, meetings, and audio files to clean text in your browser. Private and multilingual.

Quick answer and specifications

The core workflow runs in this browser. Check the supported formats, published limits, and any on-demand model download before you start.

Inputs
microphone, WAV, MP3, & M4A
Outputs
text
Processing
On this device, in the browser
Model languages
100 languages — see the language selector
On-demand model download
99 MB
Last verified

The interface language does not determine the model language. Recognition and pronunciation quality vary; check the result.

Ready to transcribe. Tap the mic or upload an audio file.

sherpa-onnx WASM

The first transcription downloads about 127 MB of speech engine and model files only after you start a recording or upload audio.

Tap to record

or upload an audio file

Transcription

Your transcribed text will appear here. Speak clearly into the microphone or upload an audio file…

Continue with a focused workflow

MP3 to Text →Audio to Text →

How to transcribe speech to text in your browser

Aiviko's Speech to Text tool transcribes live microphone recordings or uploaded WAV, MP3, and M4A audio files into editable text in your browser without uploading your audio. It runs a multilingual Whisper tiny (int8) model across 99 languages via sherpa-onnx WebAssembly (downloading about 104 MB of model weights, ~127 MB total with runtime on first use, then cached).

  1. 1. Choose the spoken language or Auto detect

    Pick the language spoken in your audio (such as English, Korean, Japanese, Chinese, or Spanish) or leave Auto detect enabled.

  2. 2. Record with your mic or upload WAV/MP3/M4A

    Speak into your microphone or select an audio file; the first run downloads and caches the local Whisper tiny engine and model (~127 MB total).

  3. 3. Review, copy, or download TXT

    Check names, numbers, and terminology against your recording, then copy the transcript or download it as a plain-text file.

Supported audio inputs and transcription features

  • Local MP3, WAV, and M4A audio file transcription plus live browser microphone dictation
  • Multilingual Whisper tiny (int8) speech recognition supporting 99 languages with automatic language detection
  • Cross-origin-isolated WebAssembly worker execution that keeps audio PCM samples on your device
  • Instant plain-text copy and .txt export for meeting notes, interviews, and voice memos

Whisper tiny model and browser decoding limits

  • The compact Whisper tiny model prioritizes fast in-browser execution and a manageable download (~104 MB model files) over studio-grade transcription accuracy.
  • Background noise, overlapping speakers, heavy accents, and domain-specific jargon or proper nouns can cause word substitutions—always proofread the draft.
  • Audio files must be decodable by your browser's Web Audio implementation and resampled to 16 kHz mono in memory; if you need timed SRT/VTT captions or video support, use the Subtitle Generator.

Related speech and audio tools

Subtitle Generator →

Transcribe audio or video up to 200 MB / 60 min into timed SRT, VTT, TXT, or JSON subtitles.

Text to Speech →

Synthesize natural speech from text in 31 supported languages and export WAV audio locally.

Frequently asked questions about Speech to Text

Are my voice recordings or MP3 files uploaded to a server?

No. Microphone audio and uploaded files are decoded to 16 kHz PCM and transcribed inside a local WebAssembly worker. The browser only downloads the sherpa-onnx runtime and Whisper tiny model files (~127 MB total on first use).

How can I generate timed SRT or VTT subtitles from audio or video?

Speech to Text outputs a continuous plain-text transcript. For time-coded cues, video input (MP4, MOV, WebM), and SRT/VTT exports, open Aiviko's Subtitle Generator.

Why does the first transcription take longer to start?

The tool waits until you start recording or select a file before downloading about 127 MB of WASM engine and Whisper model assets. Subsequent runs reuse your browser cache.

Before you use this tool

Product information reviewed by Aiviko Team

Where processing happens

The selected input is processed in your browser and is not uploaded to an Aiviko processing server.

What may use the network

The page may download code, fonts, runtime files or machine-learning models. Analytics and advertising load only under the choices and conditions described in the privacy and cookie policies.

Results and limits

Speed, accuracy and maximum practical file size depend on the input, browser and device. Review the guide before using the output for an important decision.

Read the full guide and troubleshooting notes →