Omi transcription - Better accuracy / reduce hallucinations

Hi team,

On all my transcripts, there’s plenty of errors, invented words, mistaken words, or sometimes, just full hallucinated convos. My main language being french, it seems like :

• the STT is not as accurate as in English

• the STT have trouble switching between languages when there a 2 separate conversations happening in parallel

I can provide additional transcripts samples for you to look at, but below is a first example :

https://h.omi.me/memories/93874361-2e12-4a2e-b0ee-48fffb4e3f92. Attached are the 3 segments that are recorded on the approximate date/time.