Legora utilizes Google's Gemini models for audio transcription and speaker attribution tasks