Audio newsletters: a synthetic voice, or your own?

An audio version of an issue you have already written, delivered in the same email or on the same web page, usually as a player at the top of the post. Subscribers listen instead of reading, generally while doing something else.

It is not a podcast, and the difference matters. A podcast is a separate show with its own feed and its own discovery problem. An audio newsletter is the same issue in another format, for people who already subscribe. Nobody finds you through it, and that is fine, because retention rather than reach is the job.

Most newsletter platforms now offer some version of this. beehiiv generates one with text to speech in a couple of clicks. Substack expects you to supply a file. Some writers run theirs as a private podcast feed so subscribers get it in a normal podcast app.

There is only one real question and it is not technical. Is the voice part of what people subscribed to?

If a reader could not pick your writing out of a line-up, and subscribes because the information is useful, a synthetic voice is genuinely the right answer. It costs nothing, takes seconds, and no one is disappointed.

If people subscribe partly because of how you think and phrase things, a synthetic reading strips out exactly the thing they are paying for. Not because it sounds robotic, since good text to speech no longer does, but because it does not know which clause you meant to lean on.

Daily or near-daily publication, where twenty minutes per issue is twenty minutes you do not have. News summaries, link roundups and digests, where the value is the information and nobody is listening for a performance. Long archives you want to make accessible in bulk. And accessibility as a baseline, where the point is that a reader who cannot read comfortably has an option at all.

beehiiv has this built in. ElevenLabs is what most people reach for when their platform does not. Neither needs a guide, and if you are in this category the tooling is not your problem.

Essays and anything with an argument, where emphasis carries meaning. Personal writing, where the whole proposition is that this is a person talking to you. Anything behind a paywall, because a paying subscriber notices the difference between something made for them and something generated for them.

And anything you plan to push to a podcast app, where you are suddenly next to shows with producers and a synthetic read is immediately identifiable.

There is a middle path worth knowing about: record your own voice for the flagship piece each month and generate the rest. Most readers never notice the pattern and you spend the effort where it is visible.

Read it once, out loud, all the way through, before you record. You will find three sentences that work on the page and collapse in the mouth. Change them in the written version too, because they were probably not working there either.

Do not re-record when you fluff a line. Pause, say the sentence again from the start, and cut the first attempt afterwards. Writers who restart from the top on every mistake stop making audio newsletters within about a month.

Skip anything that does not survive the transfer. Pull quotes, footnotes, image captions and the subscribe button are all reading furniture. Say so once at the top if a piece leans on a chart, rather than describing it.

Record in the softest room you have. Carpet, curtains, a wardrobe of clothes. A cheap microphone in a soft room beats an expensive one in an empty one, every time, and it is the single largest quality difference available to you.