AI Processing Disclosure
How AI is used
Audio and text you provide may be processed by Darija to English and listed service providers to generate app results.
Processing Flow
- You record or upload audio, or record an English reply.
- Our API stores the audio temporarily and queues background processing.
- OpenAI may process audio/text for transcription, translation, and speech generation.
- A Modal-hosted Darija speech recognition service may process audio for Moroccan Darija recognition.
- Convex stores app state, consent proof, temporary processing state, private artifacts, and saved outputs.
- Trigger.dev coordinates background jobs using job identifiers and processing state.
Accuracy Limits
AI-generated transcripts, translations, reply text, and speech may be wrong. Names, numbers, dates, places, official documents, and legal or administrative terms may be misrecognized or mistranslated.
Retention Limits
By default, Darija to English does not keep original recordings permanently in app storage. Provider-side processing, logs, backups, or security retention may still apply under each provider's terms.
Consent
You must consent before AI processing begins. Withdrawing consent blocks future translation and speech-generation requests until you accept the current disclosure again.