The French startup has developed AI speech models to detect and identify speakers for clearer audio transcriptions.