article2audio VS Superstring

article2audio's answer

People who publish writing on the web - newsletter writers, blog owners and small publications - who want an audio version of each post without recording it themselves every week.

The second audience is readers who would rather listen than read. That is who the app was originally built for, and it is still the reason it exists.

article2audio's answer

I built it for myself. I prefer listening to long articles while doing other things, and the text-to-speech apps I tried read the page rather than the piece: tables recited cell by cell, code read out character by character, images skipped in silence.

article2audio started as a way to make my own reading list listenable. It turned into a narration service for the people who write the posts, because they have the same problem once a week rather than once in a while.

article2audio's answer

A Python backend fetches the post and parses the HTML structure. A vision model describes the images. A language model turns the parsed post into a script a person could plausibly read aloud - deciding what to summarise, what to explain and where to pause. A text-to-speech engine renders that script. The web front end is Next.js.

Text to speech is the last step of the pipeline rather than the product.

article2audio's answer

It narrates the post rather than reading the page.

Tables are summarised instead of read cell by cell. Code blocks and pre-formatted text are explained in spoken language rather than spelled out. Images are described, so they stop being silent gaps. Pauses land where the writing pauses - between paragraphs, after headings.

The output is an MP3 and a private podcast feed, so a post can be listened to in a normal podcast app.

article2audio's answer

Most alternatives are voice vendors: you hand them text, they hand back a voice. article2audio takes the URL of an HTML post and does the work that comes before any voice is involved - working out what the tables say, what the code is doing, what the images show, and where the writing wants a pause.

If your posts are plain prose, a text-to-speech tool will do the job. If they contain tables, code, footnotes or figures, that is the part that usually sounds wrong, and that is the part this handles.