One product, five different reasons to use it
We sell neither five offers nor five prices. But a regional newsroom, a museum and a local authority are not buying the same thing: the first wants automated volume, the second wants several languages, the third wants a legal obligation met. Pick your case.
A hundred articles a day, and no time to spend on audio
You publish at a pace no manual process can follow. The audio has to happen on its own, the moment an article goes live, without a sub-editor having one more box to tick.
This is our original use case, and the one French reading was built for. Your articles are full of scores, town names, local acronyms and council reports: precisely where other engines go wrong.
Your journalists will never add a step to their publishing chain, and you cannot re-listen to a hundred files a day to check that a mayor's name came out right.
- French reading. Scores, dates, acronyms and legal references are read by written rules, and our public record proves it before you pay.
- Shared lexicon. A journalist fixes one town's pronunciation once, and every later article follows, across all your titles.
- Targeted correction. A typo is repaired by resynthesising only the passage concerned, with no new address and no second payment.
- Multi-site. Your local editions live under one account, each with its own domain and settings, on a shared volume.
- Voice per section. Sport, culture and obituaries are not read the same way, and the routing happens without anyone intervening.
A hundred and twenty articles a month go to audio automatically through the WordPress extension, averaging a thousand words. The newsroom changed nothing: journalists publish, and the player appears at the top of the article.
At a thousand words an article runs to roughly 6,000 characters, which is 6 credits: one credit covers 1,000 characters, and any started block is paid in full. Your average length therefore decides the bill as much as your article count, and the Pricing page calculator runs the sum on your own figures before you sign.
An AI charter to honour, and to be able to prove
You have published a charter on the use of artificial intelligence, and it binds you. The question is no longer whether audio is possible, but what you will be able to tell your readers and your journalists' society.
Our position is plain: no language model touches your text, only its voicing, and charter mode lets you demonstrate it. Every article has a provenance page documenting how the audio was produced, with which voice, down to any technical fallbacks.
One national newsroom compared seven solutions and kept none. It had to build its own normalisation dictionary, and it lives with known residual errors.
- Charter mode. The verifiable guarantee that no language model rewrote your text. No competitor on our panel offers it.
- Provenance page. One document per article, on the record, saying how the audio was made.
- Pre-publication review. On a sensitive story, the audio waits for a human to listen before it is served.
- Signature voice. A named column keeps its author's voice, with their written consent.
- Ad slots and registration wall. Audio becomes an acquisition and monetisation channel, on your own inventory.
Eighty investigations and long reads a month, averaging two thousand words, with human review before publication and a signature voice for two columnists.
At two thousand words a long read runs to roughly 12,000 characters and costs 12 credits, the block being 1,000 characters. An article read by a cloned voice then counts for 3 times that price: on a long read that is worth working out before deciding which columns get one, and the client area announces it before every generation.
A legal obligation, and a purchase to justify
Accessibility is not a selling point for you but an obligation, and your purchase has to survive an audit. Two questions will be put to you: where the data is hosted, and what proves compliance.
Hosting is European, and an end-to-end sovereign tier is being built. We do not announce it as available before it is, and the product checks that itself. On compliance, we document the scope precisely: the audio component, not your whole site, and our declaration template is not legal advice.
A municipal bulletin, a resolution or a by-law is packed with legal references and numbers that engines read digit by digit, which makes them unlistenable.
- Legal references. Code articles, by-law numbers and ordinals are read correctly, and the public record shows those very cases.
- Document voicing. Your reports, resolutions and bulletins become listenable without being retyped.
- Accessibility declaration. A prefilled, copyable template, with the scope written in plain sight.
- European hosting. Your texts serve only to produce the audio, never to train a model.
- No listener data. No IP address, no cookie, no profile, which considerably shortens your impact assessment.
Thirty municipal news items a month, around three hundred words each, plus voicing for council documents, in French and English for the tourism section. The communications team manages alone, with no dedicated IT department.
Multilingual editions are included, with no supplement and no tier, but each one is a separate audio file and spends its own credits: 60 editions a month, at 2 credits for a news item of roughly 1,800 characters. Council documents add to that count according to their length. Each language has its own feed and its own declared language for screen readers, which counts in an accessibility audit.
Your gallery texts, listenable in 23 languages
Your public arrives with a phone and leaves with a memory. In between there are texts few visitors read in full, and audiences who do not read your language or cannot see your labels.
An audio version replaces an audioguide you have to hire out, and it updates the day the exhibition changes. Image descriptions make a floor plan or a work accessible to blind visitors, which is often the hardest part of a cultural accessibility audit.
An audioguide is expensive to produce, goes out of date with every new exhibition, and almost never exists in more than two languages.
- 23 languages. French, English, Spanish, Hindi, Arabic, Japanese… with one edition and one feed per language.
- Image descriptions. Alt text and an audio description for a plan, a map or a work.
- Chapters and second-level links. A visit route can be split up, and each room gets its own direct link.
- Sound bed. Discreet music under the reading, adjustable by type of content.
- No visitor tracking. You measure listening without setting a cookie or collecting an identifier.
Forty gallery texts and two press packs per quarter, published in four languages of your choice among the 23. Each new exhibition replaces the previous texts, with no fresh audio production to commission.
Each language spends its own credits, since each edition is a separate audio file. Forty gallery texts of around two hundred words in four languages make 160 editions per quarter, at 2 credits each: 320 credits per quarter, not per month. That is the sum to run before picking a volume, since the volume itself renews every month.
One more channel, with no extra production
You already write, and you have neither a studio nor time to record. Audio extends attention on your longest pieces, and it opens up the listening apps where you do not yet exist.
For an agency the appeal is different: one account is enough to equip several client sites, each with its own domain, key and player settings, while volume and billing stay at account level. The MCP server even lets you drive production from an assistant.
A podcast needs a studio, an edit and a regularity nobody keeps up. Your articles already exist, but they do not reach the people who listen rather than read.
- Automatic podcast feed. Your articles become a show in listening apps, with no studio and no edit.
- Shareable clips. A short file taken from existing audio, ready for social, with no credit spent.
- Multi-site. For an agency: several client sites under one account, each with its own settings.
- MCP server. The same reach as the API, relayed to an assistant, to drive production from a conversation.
- Registration wall. Listening becomes a way to win subscribers, with no cookie and no profile.
Five sites, ten articles a month each, averaging eight hundred words. One account, one volume, one invoice, and a different player configuration per client.
At eight hundred words an article runs to roughly 4,800 characters: it starts 5 blocks of 1,000 characters, so 5 credits. A short piece costs less than a long one, but at that length none of them fits in a single credit: only the last started block is paid without being filled, and that is the only rounding in the model.
What everyone gets, whatever the trade
There is no press offer and no public-sector offer. One product, one volume price, and these five commitments that hold for everyone.
No sector-specific offer, no feature reserved for one trade. A paying account has every feature, whatever its volume or its sector.
The slider does not know your trade, only how many articles you publish. Nobody pays more for being a local authority or a press group.
Anonymous, aggregated counts, never an IP address, a cookie or a profile. True for a daily paper and for a museum alike, and it shortens your paperwork.
Your texts serve only to produce the audio and train no model. An end-to-end sovereign tier is being built, and will be announced on the day it is genuinely available.
Enough to hear the reading on your own texts. Nothing turns into a subscription by itself: with no card on file, there is nothing to charge.
Your case is not on the list?
Any organisation that publishes text in French has a reason to make it listenable. The free account asks for no card: the simplest route is to send us your own content and hear the result.
Going further: the features, one by one