Writing for the ear: what reads well and listens badly

When you turn an article into audio, you quickly discover something uncomfortable: some very well-written texts are hard to listen to. They are not bad texts, quite the opposite. They are often dense, precise, worked for the eye. But the eye and the ear do not read the same way. The eye goes back, skips a parenthesis, rereads a complex sentence at will. The ear moves at the pace of the voice and never goes back. What loses a listener is not the difficulty of the subject, it is the impossibility of catching their breath.

The good news is that listening quality is decided first in the writing, before any technical setting. A good French text-to-speech tool reads a well-built sentence correctly; it will not rescue a six-line sentence with no full stop. Here are the four faults that come up most often, and how to fix them without betraying the substance.

The nested sentence

The first fault is the single sentence that stacks clause upon clause, each opening the next. Reading it, you keep up. Hearing it, you are lost by the third aside, because the voice has nowhere to land. As we explain in punctuation and how TTS breathes, a full stop is not a comma: the full stop is the only real rest for the ear, the place where the listener files away what they have just heard before the next thing arrives.

Take a real paragraph, the kind you find in a company note or a board summary:

> Q3 revenue, up 12.4% YoY (3.2M vs 2.85M), was presented by the CFO at the 14/09 ExCo, where the CEO, revisiting the HR trade-offs agreed in June, reminded everyone that the H2 OKRs remained, despite EBITDA down 1.8 points, perfectly achievable provided the BUs prioritise the three workstreams flagged by the board.

Read aloud, this paragraph lasts about 21 seconds, and it is a single sentence: one full stop, so one real rest, right at the end. For 21 seconds the listener runs after the voice and never catches up. Here is the same point, rewritten for the ear:

> Third-quarter revenue reached 3.2 million. That is up 12% year on year. The chief financial officer presented it at the executive committee on 14 September. The chief executive made one point: the second-half targets still stand. Margin is down nearly two points, true. It will hold if each unit tackles the three priority workstreams first.

This version lasts about 23 seconds: barely longer. Writing for the ear does not mean cutting things short. But where the first offered a single foothold, this one offers six. Six times the voice truly stops and the listener breathes. The duration is almost identical, the comfort is nothing alike.

Acronyms never spelled out

The second fault is visible in the example above: CFO, ExCo, CEO, HR, OKRs, EBITDA, BUs. On the page, a used-to-it reader decodes them without thinking. Out loud, an acronym is either spelled letter by letter or said as a word, and either way it means nothing to someone who does not already know it. The writer's rule is simple: spell it out the first time, then use it. The rewritten version says "chief financial officer" and "executive committee" in full, and the ear follows. A good tool handles acronym pronunciation, and we devote a whole article to it, but no tool will guess that your reader does not know what an ExCo is.

Numbers in bursts

Third fault: numbers glued to one another. "12.4% YoY, 3.2M vs 2.85M" is crystal clear to the eye, which compares the columns at a glance. To the ear, three numbers in six words make a mush nobody retains. Space them out, keep one figure per idea, and round when precision does not serve the point: "12%" is often enough where "12.4%" clutters. Each kind of number also has an expected reading and a wrong one, a subject we cover in reading numbers aloud.

Piled-up asides

The last fault is the most insidious, because it hides in good writing habits: the aside between commas, which qualifies, nuances, adds a special case. One aside is fine. Three in the same sentence, and the main clause is out of sight before it finishes. Out loud, the listener who hears "the targets, revisiting the trade-offs agreed in June, despite a margin drop, still stood" has forgotten the subject before the verb. Move the qualifications into sentences of their own, or drop the ones that do not serve the main clause.

In short

Writing for the ear is not writing simpler, it is writing straighter: one idea per sentence, a full stop where the listener needs to breathe, acronyms spelled out, numbers spaced apart. These are writer's rules, not technician's rules, and they hold whether your article ends up read by a synthetic voice or by a presenter. The difference, in the listening, does not come from the tool: it comes from the number of places where the listener can breathe.

Measure it on one of your own texts

Take a dense paragraph from one of your articles, turn it into audio, and listen. Where you lose the thread is where a full stop is missing.

Read next

• An accessible audio player: the criteria the player itself has to meet
• Internal comms as audio: what changes when the text must stay inside the company
• Listening to a long text in French TTS: listening fatigue, and what actually helps