AssemblyAI

AssemblyAI is a developer speech-to-text platform with strong tooling around the transcript: speaker diarization for ten or more speakers, timestamps, summaries and a streaming mode. Arabic is supported as a single language code on its main models and in the streaming tier; the marketing page names Modern Standard, Egyptian, Levantine, Gulf and Maghrebi, but there are no locale codes to choose. What we appreciate is the candour. AssemblyAI publishes a tiering of its own languages and places Arabic in the "good" band, defined as more than 10% and up to 25% word error rate, rather than quoting a single-digit figure measured on clean read speech. That is closer to what you will see on real Gulf audio than most competitors admit. Pricing as of October 2026 runs from 15 to 21 cents per audio hour for pre-recorded files; data is processed in the US or EU.