Markdown to MP3 Converter — AI Voice Narration

Your writing, read aloud

A lot of the text we produce is written to be read silently: meeting notes, draft blog posts, newsletters, study summaries, scripts for videos. Written text lives on a screen, which makes it easy to skim and hard to consume anywhere else. Turn the same text into audio and it starts commuting with you, surviving a workout, or playing while your eyes are busy with something else. For anyone reviewing their own writing, hearing it read back is also a surprisingly effective edit pass — clunky sentences announce themselves the moment they are spoken.

This page does that one job for Markdown files. Upload a single Markdown document and you get back an MP3 of its content spoken aloud: a narration of the prose itself, not a robotic spelling of punctuation and markup. The rest of this page explains what the voice sounds like, which parts of your document it will actually read, and what lands in your download folder.

An AI voice with a broadcast cadence

The narration is synthesized, not recorded: the speech comes from the Qwen3-TTS engine, and we say so plainly because you deserve to know that no human read your document. The default voice is Vivian, and it holds a fixed news-anchor tone — the measured, even delivery of a broadcaster reading headlines. That register suits informational text of exactly the kind people keep in Markdown: notes, summaries, articles, briefings.

There are deliberately no knobs today. You cannot pick a different voice, and you cannot adjust the speaking rate — one voice at one cadence is what the engine offers, and the page does not pretend otherwise. If your text needs a dramatic read or a slowed-down walkthrough, this tool is not the right fit yet.

Input is capped at 2000 characters per conversion, which yields roughly 9 minutes of audio. Longer documents are rejected rather than silently truncated, so split a long piece at its natural section boundaries and convert the parts one after another.

What gets read, what stays silent

Before synthesis, the Markdown goes through a make-it-speakable pass. Code blocks are not read aloud: a fenced block of configuration or source code narrated line by line would be unlistenable and useless. Tables are skipped for the same reason — a grid of cells read out in row order is noise, not information. Anything the engine cannot speak cleanly, it leaves out.

Links contribute their anchor text only. A link labelled "the release notes" is spoken as the release notes, never as a URL spelled out character by character. Image syntax is ignored outright: alt text is not read, and images are not described.

What remains is your prose, and its structure survives in the pacing. Heading levels, emphasis and the order of lists shape how the narration flows, so a well-structured document sounds organized rather than flat. Write for the ear and the audio rewards you.

Two files, one honest timeline

The engine first produces a WAV of the narration, then transcodes it into a 192 kbps MP3. The two sound identical; the MP3 is markedly smaller and is the right copy to share, send to a colleague or load onto a phone. You can download both — keep the WAV only if some other tool in your workflow wants raw audio.

A conversion usually finishes within a few minutes; the task ceiling is about 25 minutes, so a busy queue can add waiting. The service is free, no account required, and both the uploaded Markdown and the generated audio are deleted automatically within 24 hours. No email address is requested at any point.