Convert any text to speech using your browser's built-in voices. 50+ languages, runs offline once voices are installed.
⚡ On-device voice (offline) · 🌐 Network-backed voice
Recording to file is not supported in this browser. Try Chrome or Edge, or switch to Engine B (Piper) for a downloadable WAV.
Web Speech runs locally in your browser. No text is sent to a server (cloud voices stream within the browser process).
Trim, resample, mono/stereo, gain, fade — convert any audio to 16-bit WAV right in your browser.
Normalize audio loudness to Spotify/YouTube/Podcast/Broadcast targets via peak or LUFS — entirely in-browser.
Detect the tempo (BPM) of any song or beat by analyzing peak intervals — runs entirely in your browser.
The Text-to-Speech tool exposes two complementary engines, both running entirely in your browser. Engine A (Web Speech) uses the voices already installed on your operating system — instant, no download, available offline once voices are installed, but quality varies by OS and language coverage is uneven for low-resource languages. Engine B (Piper Neural via VITS) loads compact ONNX models from Hugging Face on first use and runs them through onnxruntime-web, producing voices that rival commercial cloud TTS like Google Cloud and Amazon Polly. Piper covers 35+ languages with 100+ voice options — including Vietnamese (vais1000, vivos), Japanese (built-in language coverage via standard phonemizer fallbacks), Chinese (huayan), Korean, and many low-resource European languages where Web Speech often has nothing usable. The first time you select a Piper voice the tool downloads a model (typically 10–60 MB depending on quality preset) and caches it in your browser's OPFS storage; subsequent uses are instant and fully offline. Both engines expose voice picker by language, and Piper output is a downloadable WAV file you can drop directly into a video editor or podcast workflow.
Use Web Speech when you need quick narration in a major language (English/Spanish/French/German/Japanese/Chinese) and the OS-shipped neural voice is acceptable. Use Piper when (1) the language is low-resource and Web Speech has no usable voice, (2) you need a downloadable WAV file for editing, (3) you need consistent quality across all platforms (Web Speech sounds different on Windows vs macOS vs Linux), or (4) the content is sensitive and you want zero network calls during synthesis (Piper synthesizes locally after the one-time model download).