AI text to speech in Spanish and English, in your browser

How to turn text into speech, step by step

You do not need a file open: by default the speech becomes a new file.

1. Click Open Auditius to generate speech.
2. Open the Generate menu and choose Speech…
3. Type or paste your text in Text. Add [pause] wherever you want half a second of silence.
4. Set Language to Spanish or English and choose a Voice. The list groups the voices by engine and says which ones are already downloaded.
5. The first time you use a voice, click Download. It stays in this browser for next time.
6. Adjust Speed if you need to, from 0.5× to 2×, and leave Output on New file.
7. Click Generate, listen with the Space bar and save it with File ▸ Save As… as WAV, FLAC, MP3 or Ogg Vorbis.

Which voice to choose

There are three speech engines, and each has voices for Spanish and English.

• Kokoro: natural and fast. One 96 MB download covers its seven voices: Dora, Alex and Santa in Spanish, and Heart, Bella, Michael and Fenrir in US English.
• Piper: very fast and light, one download of 28 to 64 MB per voice. davefx and carlfm speak Spanish from Spain, ald Mexican Spanish, and LJSpeech and Norman US English.
• Chatterbox: the most expressive, with an Exaggeration control instead of Speed and the option to clone a voice. It is a 963 MB download and needs WebGPU; without it, it runs on the processor at around 15 to 25 seconds per second of speech.
• The first voice you download also brings a 27 MB runtime that every model shares.

Pauses, paragraphs and where the speech goes

Pause tags work the same with every engine, in English or Spanish.

• [pause] or [pausa] adds 0.5 seconds of silence; [pause 2s], [pausa 1.5] and [pausa 300ms] set the length.
• A blank line between paragraphs adds a 0.8-second pause.
• Other words in square brackets are not read aloud: the window lists them and skips them.
• Insert at the cursor puts the speech into the active file at the cursor and pushes the rest of the audio to the right, in one step you can undo.

Cloning a voice, only with consent

With Chatterbox you can save a voice from a short recording. Open Clone a voice, use a selection of the active file or A file as Reference (5 to 10 seconds of one person speaking clearly, without music), type a Name and tick This is my voice or I have its owner’s permission. Save voice stays disabled until you do.

Cloning needs a separate 592 MB voice encoder. Saved voices stay in this browser only, and you choose them later under Speaker.

Generated speech is labeled

Every result is marked as synthetic speech in its metadata, with the engine and the voice that produced it. The note is written into WAV files, and when you insert speech into a recording, that file gets the note too.