Audar ASR V1

AudarAI is an Arabic first, multilingual voice AI platform offering speech recognition, text to speech, voice translation and real time voice agents. Its speech recognition line, released as version one, is built for strong coverage of Arabic dialects as well as other languages, and the company positions the platform as an Arabic centred alternative to general voice AI providers. The platform serves enterprises, developers and governments. The speech recognition family comes in several sizes: a Flash model of 0.8 billion parameters with open weights for ultra low latency use, a Turbo model of 2.35 billion parameters with open weights for high accuracy self hosted deployments, and a commercial diarization model that separates multiple speakers in live conversations. Matching text to speech models range from a small open weight Flash model for edge and mobile, through a CPU optimised Turbo model, to a 4 billion parameter commercial Pro model for expressive long form speech. Voice agents combine streaming recognition, low latency synthesis and turn taking, and full on premises and sovereign cloud deployment is available. Usage based API prices are published in US dollars, with a note that the scheme rolls out in September 2026 and may change. Speech recognition costs $0.008 per audio minute for Turbo, $0.014 for Pro and $0.018 for diarization. Text to speech costs $0.012, $0.018 and $0.025 per 1,000 characters for Flash, Turbo and Pro. Voice translation is $0.028 per audio minute and the voice agent runtime is $0.055 per conversation minute, with telephony fees passed through. Volume and on premises pricing is by contact with sales. Billing is calculated per second and invoiced monthly per organisation.