Speaker diarization is a crucial technology in speech recognition that identifies and separates individual speakers in multi-speaker audio recordings, enhancing the readability and analysis of transcripts. Advances in Automatic Speech Recognition (ASR) have transformed diarization from basic acoustic recognition to sophisticated dual-model approaches that use segmentation and speaker embeddings, which help in accurately identifying speakers even in challenging scenarios like overlapping speech. Gladia's speech-to-text API, incorporating diarization as a core feature, is particularly adept at handling various audio file types, including mono, stereo, and multi-channel, and offers multilingual support. The API employs both mechanical and AI-based approaches to ensure high-quality speaker-based transcripts, making it suitable for a range of applications, from transcription in call centers to speaker identification in security contexts. The latest updates have significantly improved the API's speed and accuracy, even in complex situations, reinforcing its utility in streamlining transcription processes across different industries.