AI narration

Our AI narration system is made to serve the longtermism, effective altruism and progress studies communities.

There are no other AI narration services that edit complex writing for a great narration—they are all focussed on articles with a simpler structure, namely: continuous prose.

How we edit your writing

Click "play" to hear some examples.

Sample narration #1

The post "Advice on communicating in and around the biosecurity policy community" contains headings and footnotes that are important to understand the piece, and one reference footnote which should be mentioned but not narrated.

Comparison to Nonlinear Library

Nonlinear reads the footnotes directly after the end of the article (see 10:50). It does not indicate that it is reading footnotes, so the listener experience is confusing.

The author makes frequent use of the "/" character, which is usually best replaced with "or" in the narration. Nonlinear just ignores the "/" character, so it says (e.g.) "scientists who do DURC ePPP research" instead of "scientists who do DURC or ePPP research".

Nonlinear pronounces the "ePPP" acronym as ("ep") rather than "e P P P". This may or may not be desirable, depending on the norms within the field. TYPE III AUDIO spells each letter of an acronym by default, and has a growing database of case-specific pronunciations.

Sample narration #2

The post "Shallow Problem Review of Landmines and Unexploded Ordnance" features lists, nested lists, statistics and specialist terminology.

Comparison to Nonlinear Library

Nonlinear makes several errors of pronunciation (e.g. "IED" is pronounced "E D" instead of "I E D"; "2/3rds" is spoken as "two three R D S" instead of "two-thirds").

There are various other issues. For more details, see our thoughts on how the Nonlinear Library narration could be improved.

Sample narration #3

Here's how Nonlinear reads a table that compares the cost-effectiveness and tractability of different interventions:

The TYPE III AUDIO system detects special content, such as tables, and replaces them with an audio note such as:

There's a table here, see the original post.

In the future, we may take screenshots of content such as tables and graphs and include them in the episode description.