MAI-Transcribe-2-Streaming $0.54/hour vs Scribe v2 Realtime $0.39

On listed prices, MAI-Transcribe-2-Streaming ($0.54 per hour) is the most expensive of three live speech-to-text options: ElevenLabs Scribe v2 Realtime is $0.39 per hour and AssemblyAI Universal-3.6 Pro Realtime is $0.45 per hour. Microsoft calls its price introductory through the end of 2026, so it may change in January. Price is one input; Microsoft also claims first hypotheses in just over 100ms and a no. 1 Artificial Analysis rank, both vendor claims that I cannot check here.

Live transcription price per hour

Differences against MAI at $0.54: Scribe v2 Realtime is $0.15 less ($0.54 - $0.39), which is 28% lower; AssemblyAI Universal-3.6 Pro Realtime is $0.09 less ($0.54 - $0.45), 17% lower. Per minute, $0.54 / 60 = $0.009, $0.39 / 60 = $0.0065 and $0.45 / 60 = $0.0075.

Billing details that move the real cost

• Session duration vs audio duration. AssemblyAI says streaming is billed on session duration. If you hold a socket open through silence, you pay for the silence. Check how each vendor meters before you compare headline rates.
• Introductory price. Microsoft's $0.54 runs through the end of the year; build a margin for a different number after.
• Add-ons. AssemblyAI lists Medical Mode at +$0.15 and diarization at +$0.02 on its page; ask whether the live model you pick has equivalents.

Where Sume fits

Sume does not offer live streaming transcription. Its speech-to-text is a feature of POST /v1/video-inspect (transcribe: true) on a hosted clip, billed at $0.01 per audio minute, which is $0.60 an hour, on top of the job's compute. That is a file workflow, so the table above is a reference for live needs, not a Sume price list. If you transcribe recordings after the fact, see the Scribe v2 per-minute comparison.