Eleven Labs’ Scribe v2 (https://elevenlabs.io/speech-to-text) is a voice-to-text transcription model with two versions: v2 real time for on-the-fly transcriptions and v2 for transcriptions from an audio or video file.

Narrator recognition and 90 languages supported.

Available for testing on the page above, in Eleven Labs tools and via API.

Free and seems unlimited.