TTS audio is speech generated from written text by a computer. The acronym stands for text-to-speech. You give the system text, it gives you back an audio file of a voice reading that text aloud.

Until a few years ago, TTS audio sounded robotic. The kind of voice you heard on phone menus and accessibility tools. That changed when neural network models replaced the old concatenation and parametric systems. Modern TTS voices read with realistic intonation, pacing, and emotion. In a 2025 blind test reported by Alphasignal, Microsoft MAI Voice 2 fooled 45% of listeners into thinking they were hearing a human, the closest of any TTS tested.

How TTS audio is generated

The pipeline is roughly:

  1. Text in. You paste a script, an article, or a full document. Plain text, Markdown, or extracted from a PDF.
  2. Normalize. The model expands abbreviations, disambiguates homographs (read vs. read), parses numbers, dates, and currency into spoken form.
  3. Spectrogram prediction. The neural model produces a time-frequency representation of the voice saying the text.
  4. Vocoder. A second model converts the spectrogram into a waveform you can play as an MP3 or WAV.

On tts.audio the whole loop runs in one to three seconds for short text. Long documents up to 100,000 characters are split at sentence boundaries, synthesized in chunks, and stitched back together losslessly.

What TTS audio is used for

  • Voiceovers. YouTube, TikTok, explainer videos, ads. Recording a human voiceover costs time and money. TTS gives you a take in seconds.
  • Accessibility. Visually impaired readers, dyslexic readers, and anyone who prefers listening. TTS turns any written page into audio on demand.
  • Language learning. Hearing native pronunciation for arbitrary text is one of the fastest ways to internalize a new language.
  • Proofreading. Errors that hide on the page jump out when heard. Many writers run their drafts through TTS as a final pass.
  • Podcasts and audiobooks. Intro and outro voiceovers, narration for indie audiobooks, draft passes before hiring a human reader.

How much TTS audio costs

Pricing in 2026 is almost universally per character. tts.audio charges around $0.13 per day for 100,000 characters on the Pro plan, or $0 for 10,000 characters a month on the free plan. Other major providers cluster between $5 and $40 per month for similar volume. You can compare them on the homepage pricing table.

The voices

tts.audio routes through three neural TTS engines: Microsoft MAI Voice 2 (the default, with expressive styles like cheerful, sad, and whispering), Google Gemini 3.1 Flash (multilingual, with inline mood tags), and Alibaba Qwen Audio (multilingual with low-latency PCM output). 76 voices across 17+ languages are listed on the voices page, each with a playable sample.

Try TTS audio yourself

The fastest way to understand TTS audio is to generate some. The free TTS audio page lets you type up to 300 characters, pick any voice, and hear the result instantly, no sign-up required. If you want longer audio, PDF upload, or MP3 downloads, the free tier includes 10,000 characters a month with no credit card.

Further reading