The result to aim for
After experiments in 2024, Radio France transcribes its programmes using Whisper, to make them accessible to deaf and hard of hearing people, with a human supervision and a correction form open to listeners.
What was done
On the "Sound Transcription" page of its site, Radio France states that, after successful experiments conducted in 2024, it is developing a technology derived from AI to accelerate the digital accessibility of its audio content (subtitled podcasts and broadcasts). The chosen tool is Whisper, "combining artificial intelligence and human expertise to produce quality texts". The displayed approach is an use "always subject to human supervision and validation". "Today, the transcription of the programmes from France Inter, franceinfo, This evolutionary approach will allow the entire offer to be extended by the end of the 2025 season.» For all antennas except franceinfo, all programs are transcribed "with the exception of certain formats in order to avoid transcription errors related to AI that could generate misinformation": newspapers, news flashes, morning interviews and news magazines are excluded. On franceinfo, the transcription expanded in August 2025 to include a list of additional chronicles. Transcripts are accessible on radiofrance.fr, via a quote icon in the player. Auditors can propose corrections by form, used to improve the service. The page reminds us that automatic voice recognition can be imperfect and that errors can escape readings.
Before you start
Tool cited by the source: Whisper (Radio France); other tools cited: Whisper. Fill in the bracketed items: [coller l’audio transcrit brut], [coller], [inaudible]. Strip confidential data that is not needed and check your tool’s access terms. Features and plans can change.
The method, in three steps.
Gather the context
Spot the bracketed items in the prompt (
[coller l’audio transcrit brut],[coller],[inaudible]) and prepare what they ask for.Adapt the prompt
Copy the prompt below, replace the bracketed items, then paste it into the AI assistant you use. The cited tool (Whisper (Radio France)) is not necessarily available to the public..
Reread before using
Reread the result, correct it if needed and keep the final decision. Caveats are detailed below.
Your starting point.
Replace the bracketed items, then adapt the proposal to your situation. This prompt is written by the iadujour desk It is not a quotation from the source.
Extract from issue to be transcribed or reread: [coller l’audio transcrit brut]. Antenna and type of emission: [coller]. 1. Proposes a rereading transcription, keeping useful hesitations and correcting the spelling of names if the context allows, otherwise indicates doubt. 2. Do not invent any word covered by an inaudible passage: write [inaudible]. 3. Lists the names, numbers and titles to be checked by a documentist. If the extract looks like a newspaper or flash, it reminds us that it requires a reinforced review.
What deserves your attention
Proof level: Institutional source. Source checked on 10 November 2026. Use described by the source, not tested by the desk. Original summary written by iadujour.fr: figures and claims are the source’s; the prompt is a desk proposal, not a quotation.
Comments.
A field note, a question, a precision on this use case? Comments are reviewed before publication.
Published comments
Loading comments…
Leave a comment
Moderation charter. Every comment is reviewed before publication. We refuse hate speech, insulting or discriminatory language, advertising, spam and promotional links, off-topic posts and other people’s personal data.