Audio articles for French-speaking Africa: the real question is file size
Mobile-first readers, expensive data, patchy networks: for a francophone newsroom in West or Central Africa, the first test of article audio is not the voice, it is the weight of the file and what it costs to listen. The exact maths for a 1000-word article, no rounding.
For a French audience, article audio is a conversation about voice quality, pronunciation and listening comfort. For a reader in Dakar, Abidjan, Kinshasa or Lomé, the first question is different: how heavy is the file, and what does it cost to play. A French-language newsroom across West or Central Africa serves an audience that is overwhelmingly mobile, on data bundles counted by the megabyte, over a network that is not always there. Audio done well is a genuine service in that setting. Audio done carelessly becomes a cost the reader declines to pay. Here is the maths, without a flattering round-up.
The only figure that matters: the actual file weight
Take a 1000-word article, a common length for a feature. Read aloud by a French synthetic voice at a natural pace of about 145 words per minute, it runs for roughly 7 minutes. Everything else follows from that duration, because an audio file's weight does not depend on the number of words, it depends on the number of seconds multiplied by the encoding bitrate.
A mono-voice MP3 is usually encoded between 64 and 128 kilobits per second. At 128 kb/s, each second of audio weighs 16 kilobytes. Our 7 minutes (420 seconds) therefore come to about 6.7 megabytes. Drop to 96 kb/s, which a spoken voice barely loses anything from, and you land at roughly 5 megabytes. At 64 kb/s, still acceptable for speech, the same article weighs about 3.3 megabytes. This is not a marketing estimate: it is a multiplication you can redo yourself, bitrate times duration, and it will give you the same answer every time for your own articles.
Hold on to the order of magnitude: a feature in audio is 3 to 7 megabytes. A short 400-word piece lands around 1.5 to 2.5 megabytes. That is little next to a video, a lot next to text alone, which weighs a few kilobytes. Audio sits in between: it will not wreck a bundle, but it will not go unnoticed either.
What that weight costs in data, and why we will not quote you a price
The tempting move would be to announce a cost in CFA francs or in cents. We will not, because it would be wrong for half of your readers. The price of a megabyte varies enormously from one country to the next, one operator to the next, one bundle to the next, between a daily pass, a monthly plan, and out-of-bundle data billed at full rate. A single figure would be a lie by simplification.
What we can give you is the method. Take the price of a megabyte for your target audience (their everyday bundle rate, not the walk-up price), multiply it by the weight of your audio article, and you have the real cost of one download listen. Do it once for your market and you will know at once whether audio is a gift or a burden for your readers. That honesty about numbers is the same one we keep about what "free" actually covers: a hidden cost is still a cost.
The three settings that change everything
Weight is not fate, it is a choice, and three levers control it.
The first is the encoding bitrate. For a spoken voice, going from 128 to 96 kb/s cuts the weight by a third with no audible loss. It is the highest-return move, and the quietest.
The second is playback behaviour. A file downloaded once replays at no further cost; a stream reloaded on every open charges the reader as many times as they come back. On an expensive network, allowing caching or download rather than forcing streaming is a service to the audience.
The third is editorial: offer audio, do not impose it. A reader in a weak-coverage area should be able to read the light text and trigger the audio only when they want it. Audio is a door, not a toll at the entrance.
Our position, and its limit
WeDispatch produces a standard audio file, playable anywhere, whose weight follows exactly the maths above: no hidden overhead, no heavy proprietary format. That is what a constrained mobile use demands. Our tool is built to read French well, and a clean, light file is part of that standard, as much as the voice itself.
Now the limit, plainly stated: we do not do automatic network adaptation that would lower the bitrate based on each listener's detected connection. The bitrate is chosen at production, not on the fly per listener. For a newsroom targeting a high-data-cost audience, that is a setting to fix upstream, knowingly, not a magic the tool performs for you. As with where your text actually goes, we would rather show you the mechanism than sell you a black box.
Give your articles a voice with WeDispatch
This blog is itself voiced by WeDispatch. Curious how it sounds on your content?
Book a demo