Audio Transcription & Meeting Notes
AI-Public can turn audio recordings into text and create meeting notes from the transcript. Transcription uses the provider from the central model catalog, such as OpenAI or European AI. When you start, choose whether the recording is for personal use, a meeting, or a lesson/presentation.
Start screen
On the transcription screen you can start a new recording or upload an existing audio file.
Providing audio
There are two ways to provide audio for transcription.
Record directly in AI-Public
Pendant un enregistrement actif, choisissez Pause pour interrompre temporairement la capture audio. Le temps et l’audio en pause ne sont pas inclus. Choisissez Reprendre pour continuer dans le même enregistrement et la même transcription ; utilisez Arrêter uniquement lorsque la réunion est terminée.
Click Start recording to begin. Before recording starts, a dialog opens with the recording settings.
Recording settings
When starting a recording you can set:
- Recording type:
- Private recording: one person close to the microphone.
- Meeting: multiple speakers in one room.
- Lesson or presentation: one main speaker with possible interaction.
- Specialist vocabulary and keywords: add names, abbreviations, product names, or terms that are often recognized incorrectly.
- Language: AI-Public uses your account language to guide transcription.
Behavior per recording type
The selected type determines how the recording is processed:
- Private recording with OpenAI uses realtime transcription. Text appears while you speak. Because this is meant for one person, speaker diarization is not applied and no interim audio files are stored.
- Meeting and Lesson or presentation use file-based processing. AI-Public processes audio parts during the recording and also processes the complete final recording when you stop. This path is suitable for longer recordings, multiple speakers, and recovery after interruptions.
- European AI/Mistral uses file-based processing. For private recordings, diarization is disabled so the transcript is not unnecessarily split into speakers.
Use an existing audio file
You can also upload an existing recording. Supported formats include MP3, WAV, M4A, and WebM. After upload, the file is processed with the same transcription approach as a recording of the same type.
Transcription and speakers
The transcript can contain time blocks and speaker labels. For conversations and meetings, the model tries to distinguish speakers. For private recordings this is disabled because the transcript is intended for one speaker. Sometimes labels such as Speaker A and Speaker B are used instead of real names. AI-Public post-processes explicit introductions in the text when possible.
Speaker recognition still depends on audio quality, overlapping speech, and the selected model. If names or terms are not recognized correctly, you can improve the transcript with AI.
Improve with AI
After processing, use Improve with AI for targeted corrections, such as renaming speakers, fixing a technical term, or applying a spelling correction consistently. Always check the result when the transcript is used for reporting or decisions.
Meeting notes
After the recording and transcription, open Meeting notes and choose Create meeting notes. The notes are based on the transcript and the active prompt.
Advanced settings
Manage prompts
AI-Public provides default prompts for general meeting notes and notes with speaker recognition. You can also add your own prompts, for example for a fixed meeting structure, action list, or public report. Custom prompts are stored in your account for future transcriptions.
Manage history
Via History you can search, load, rename, edit, or delete earlier transcriptions and play linked audio.
Use the transcription
You can copy the transcript, export it as PDF, use it in chat, or export meeting notes to PDF or Word.
Audio parts and storage
For private realtime transcriptions with OpenAI, AI-Public does not store interim audio files. For meetings, lessons/presentations, and European AI, AI-Public automatically processes audio parts for progress, reliability, and recovery. When you stop, the complete final recording is processed and has priority as the definitive basis. If one interim part fails, the recording can continue; check afterwards whether the final recording was processed correctly.
Lire et récupérer les enregistrements
Lorsque l’enregistrement complet est disponible, un seul lecteur audio apparaît. Sinon, choisissez une partie dans la liste. L’application mobile permet de renvoyer les enregistrements locaux ou de les supprimer définitivement.
En attente d’envoi. La copie locale est supprimée après un envoi réussi. Désinstaller l’application efface les enregistrements locaux.
Traitement standard : jusqu’à 1 Gio par fichier audio. Prévoyez assez d’espace libre pour l’enregistrement et le traitement.
Sur Android, rouvrir l’écran n’arrête pas l’enregistrement. Après un arrêt forcé, vous pouvez récupérer l’audio et reprendre le même enregistrement. Si un appel bloque le microphone, l’enregistrement se met en pause. Reprenez lorsque le microphone est disponible. Si le casque est déconnecté, l’application essaie le microphone intégré. Les interruptions restent visibles ; les parties internes forment un seul enregistrement.
Dans l’application Android, Enregistrer ouvre directement l’écran d’enregistrement natif. Après l’arrêt, choisissez avec ou sans reconnaissance des locuteurs et ajoutez éventuellement des termes spécialisés dans la même fenêtre. Annuler conserve l’enregistrement localement. La notification affiche la durée et permet de mettre en pause, reprendre ou arrêter, y compris sur l’écran verrouillé si vos paramètres le permettent. Après un arrêt depuis la notification, revenez dans l’application pour lancer le traitement. L’enregistrement reste disponible dans le navigateur mobile.
