How does it work?
An AI speech model listens to audio and turns it into text. Quality depends on the model, the language, the accent and the jargon. For Dutch, and certainly for Flemish with dialect and technical jargon, standard models often perform worse.
What is it for?
- Turning phone calls and conversations into text and tasks automatically
- Meeting notes that write themselves
- Subtitling and live translation of trainings or events
- Dictation and hands-free input
The challenge: dialect and jargon
Standard models miss dialect, proper names and company-specific terms. A custom dictionary (client names, systems, technical terms) significantly improves transcription quality. After that, you can split the text per topic into separate action items.
Related terms
- EU hosting & zero retention: where and how the audio and text are processed securely.
- RAG: lets an assistant answer based on those transcripts.
- Vector store: makes the transcripts searchable by meaning.
