WeDispatch 1.5.0: European preparation now covers the whole chain

The previous version brought audio quality onto a single axis: your plan determines your tier, and your tier governs the whole chain. Version 1.5.0 follows that through where it shows least and matters most — where your text is processed.

Three more operations prepared in Europe

Producing audio is not simply a matter of reading a text out loud. Upstream, several operations handle your content, and they all now follow your account's tier.

Article extraction first: you paste an address, we fetch the page and pull out the text that matters, stripped of the menu, the banner and the sidebars. The "the essentials in 30 seconds" summary next, which reads the whole article to draw out a few sentences that hold up when spoken. Image description last, the one that feeds your alt text: a newsroom infographic is editorial content like any other, and it is treated as such.

The principle is easy to state, and that is what makes it worth something: one rule decides where processing happens, across the entire preparation chain, with no exceptions and no special cases. What holds for reading an article holds for all three.

Image description, whatever your tier

Describing an image used to require one specific engine. It is now available on both, with exactly the same writing instructions. This is a point we hold to: your tier changes where the processing happens, never the result you get back.

A provenance log that stays close to the facts

Every audio keeps its log: the tier used, the processing location, the transformations applied to the text. It is what you can put in front of anyone asking where your content went — a legal team, a tender committee, a curious reader.

That log now states the location actually used for the generation. The standard and premium tiers do accept a fallback: if the usual preparation engine is briefly unavailable, your generation continues on the backup engine rather than failing. When that fallback comes into play, the log says so in plain terms.

The sovereign tier falls back on nothing. Its generation fails rather than leaving Europe, because that is exactly what it guarantees.

Legal pages at the granularity of the chain

Our privacy policy and our security page have always named our subprocessors rather than writing "partners bound by confidentiality commitments". They now describe generation as it actually runs, in two stages — text preparation, then reading — with the subprocessor involved at each stage, its precise role and its country of establishment. The technical fallback is covered too, along with what it means for each tier.

All four pages are affected, French and English, privacy and security, and they now say exactly the same thing in both languages.

What comes next

A fully European voice layer is our work in progress. Until it ships, the sovereign tier reads with the premium engine, and we keep writing that everywhere the tier is mentioned rather than letting anyone assume the chain is already complete end to end.

This is the kind of release you cannot see on screen. It simply makes what we claim verifiable — which is precisely what a sovereignty commitment ought to deliver.

Other release notes

• WeDispatch 1.5.3: every blog article now has its audio version
• WeDispatch 1.5.2: the status page now watches text preparation too
• WeDispatch 1.4.0: three tiers, one axis