Neural Text to Speech

This text to speech tool reads your text aloud with neural voices that run entirely in your browser. Download a voice model once, then paste text or drop a folder of text files and get a WAV file for each, ready for videos, e-learning or accessibility. Choose SpeechT5 for natural English or MMS voices for seven other languages, or use your device's built-in voices with no download. Nothing you type leaves your device.

Voices

ModelLanguagesDownload
SpeechT5 + HiFi-GANEnglish≈ 210 MB
MMS (VITS)One per language≈ 40 MB each
Browser voicesDepends on deviceNone

When to use it

  1. Narrate videos and slides.
  2. Make audio versions of articles.
  3. Proofread by listening.
  4. Create e-learning voice-overs.
  5. Batch audio prompts for apps.

How to Use

  1. Choose a voice and download it once.
  2. Type text or drop text files.
  3. Make audio.
  4. Play and download WAVs.

Features

  1. Neural voices on-device
  2. 8 languages
  3. Batch text files
  4. WAV and ZIP download
  5. Browser voices alternative
  6. Clear downloaded models

FAQs

Is this text to speech free and private?

Yes. The voice model runs on your device, so your text is never sent to a server.

Which languages are supported?

Natural English with SpeechT5, plus MMS voices for English, French, German, Spanish, Russian, Arabic, Vietnamese and Romanian.

Can I download the audio?

Yes, as WAV files — one per text or file, or all together in a ZIP.

Is anything uploaded?

No. Files and text are processed in your browser. Only the model itself is downloaded, once, when you ask for it.

Can I delete the downloaded model?

Yes. The Downloaded models panel lists what this tool stored and deletes it after a warning; you'd need to download it again to use it.

Technical Details

Text is split into sentences of up to 250–400 characters. Each is synthesised by a Transformers.js text-to-speech pipeline running in a Web Worker: Microsoft SpeechT5 with a HiFi-GAN vocoder and a fixed speaker embedding for English, or Meta's MMS VITS models for French, German, Spanish, Russian, Arabic, Vietnamese, Romanian and English. Clips are joined with a chosen pause, peak-normalised to −1 dBFS and written as 16-bit mono WAV. The model downloads from Hugging Face into Cache Storage only on request and can be deleted. The browser's own speech voices are offered as a no-download alternative for listening.

Audio Tools

Audio Tools

  • Audio Noise Remover
  • Audio to MIDI Converter
  • Audio Trimmer
  • Audio Merger
  • Audio to WAV Converter
  • Audio Speed & Pitch Changer
  • BPM Detector
  • Tone Generator