About Parlessa

Why the language lab

Every language app teaches. Almost none of them make you speak under a clock. The people we built this for have done the lessons: they read French, they follow a film, they understand the boulanger, and then the sentence they know perfectly well arrives four seconds late. The missing piece is not knowledge. It is the practice that turns knowledge into speech, and that practice existed: the four-phase drill of the 1960s language laboratory, which universities ran for twenty-five years and then quietly dropped when the rooms became too expensive. The drill was never the problem. The room was.

So Parlessa is the drill without the room. Two native voices, a worksheet that shows only the model, a countdown, and nobody listening. It is not a course, not a game, not an AI tutor, and it has no speech recognition on purpose. Fifteen minutes a day, out loud.

Who is building it

Parlessa is built by Tim Fogo, a software engineer and a B1 French learner who got tired of freezing. The syllabus and drills are drafted with language models and then checked by people: an automatic validator, a second-pass teacher review, and a native French reviewer who reads every line blind before a single clip is recorded. The founder decides what ships; the models never publish anything.

A part-time native reviewer joins at the end of the validation phase. Their name and credit will be on this page, with the voice credits, once the first reviewed labs are published.

How the French is checked

1. Validation. A script checks every lab for structure, digits written as words, spoken-form overrides, duplicate items and timing budgets.
2. Teacher pass. A second model, prompted as an FLE teacher, reads each line for register, agreement, liaison traps and naturalness, and flags anything doubtful.
3. Native review. A person reads every line with the answers hidden, writes their own, and compares. Any disagreement stops the lab.

For the first five labs every audio clip is listened to before publication. After that we sample, and every lab has a flag button on every item so learners can report a line that sounds wrong. Flags are read within a day.

The voices

Two native French voices, one narrator and one responder, generated with ElevenLabs under a commercial licence and frozen once chosen, so the whole catalogue sounds like one recording session. Voice credits will be listed here when the voices are final.

The heritage

The archive photograph we want on the home page shows a French university language lab in 1964: booths, headsets, reel-to-reel decks under smoked glass. The rights are being cleared and it will appear with a full credit. Until then, the illustration stands in. The history of the format is on the blog.

Contact

Email hello@parlessa.com or use the contact form. We reply within one working day. Press and partnership questions go to the same address.