Reviewing audio before it publishes: keeping the last word
Text-to-speech is excellent, but a newsroom does not always want to publish blind. How to queue each audio, listen, correct it, then approve it before it airs.
The question comes up at every demo in a serious newsroom: "What if the voice mispronounces a name?" It is a fair one. Today's text-to-speech is good, often better than people expect, but "often" is not "always", and a news title does not want to find out at the same time as its readers that a proper noun was mangled or that a sensitive story was read in a tone that jars. Publishing automatically sometimes means taking an editorial risk a newsroom will not accept. The review queue exists for that refusal.
The real risk is not average quality, it is the edge case
The problem has to be named precisely, or it gets solved badly. Average reading quality is not the issue: on the vast majority of articles the audio is good and calls for no intervention. The risk is concentrated in a minority of cases: a rare proper noun, an initialism the context makes ambiguous, a foreign-language quotation, a story where tone matters as much as words. We explain elsewhere why proper nouns are a hard test for the press: that is exactly the kind of detail a review catches and automatic publishing lets through.
For those cases, the right answer is not to give up on automation, nor to re-read everything on principle. It is to keep a human checkpoint where it matters, without slowing down the rest.
How the review queue works
The principle rests on one toggle, set per site. When review is on, generated audio no longer goes live directly: it enters a "pending" state. Nothing is served to the reader until a member of the newsroom gives the green light.
The member listens to the pending audio. Three outcomes, one rule: nothing airs without explicit approval. If all is well, they approve, and the audio publishes. If a detail is off, they apply a targeted correction to the passage in question, listen again, then approve. And if they judge that the audio does not belong, they simply do not approve it: the article stays as text, with no audio version, which is a perfectly valid editorial choice.
This control builds on team management: it is your members, each with their own account, who listen and decide, not a shared login. So you know who approved what, and access is removed cleanly when someone leaves the newsroom.
Finding the right dosage
Review is not meant to apply everywhere in the same way, and that is what makes it sustainable over time. A newsroom can turn it on for its main site and leave it off on a lower-exposure secondary feed. It can also keep it while building confidence in the reading, then ease it once several weeks have shown that problem cases are rare and well identified.
The point is that reading quality and editorial control do not pull against each other: the first cuts the number of cases to review, the second guarantees that none airs without a human eye when you require it. That balance, rather than blind trust in the machine or a distrust that would cancel the whole point of automation, is what suits a title that cares about its byline.
In practice, in a newsroom this often looks like a short pass at the start of the day. The duty editor opens the queue, listens to the handful of pending audios while proofing the morning's articles, fixes a mistaken town name or local official's name along the way, and approves. It takes a few minutes because the reading is already good in most cases: the queue is not there to redo everything, it is there to let nothing slip. It is the same gesture as checking a headline or a caption, applied to the voice.
Review and scheduled publishing
Review combines with scheduled publishing without contradiction. An audio can be prepared and approved ahead of time, then served at the planned moment: the human green light is given once, delivery follows the calendar. You keep the last word on substance, the machine handles the timing. It is the healthiest division of roles you can aim for: judgement to the newsroom, execution to the tool.
Keep editorial control over your audio
The free plan lets you test the reading on your own articles before deciding on your review setup: try it now.
Give your articles a voice with WeDispatch
This blog is itself voiced by WeDispatch. Curious how it sounds on your content?
Book a demo