6 podcast hours transcribed: MAI Streaming $3.24 vs Sume STT $3.60

Six hours of audio costs $3.24 on MAI-Transcribe-2-Streaming at its introductory $0.54 per hour, which Microsoft lists through the end of the year. On Sume STT at $0.01 per audio minute it costs $3.60: 360 minutes, run as 36 jobs of 10 minutes because one job takes up to 600 seconds.

Microsoft's rate is from its October 1 announcement, and Sume's limits from the Sume API reference (read 2026-10-06).

What is the comparison?

These are different kinds of product. MAI-Transcribe-2-Streaming returns partial text while someone is speaking, which Microsoft says begins just over 100 ms after audio arrives. Sume STT is a batch job on a recorded file with no streaming, so the price compares cost, not what you get.

What changes after December?

Microsoft states the $0.54 rate is for the rest of the year, so the comparison for 2027 needs a fresh read of its page. Sume's per-minute STT price is the catalog rate, and GET /v1/catalog returns the current one.

Which one do you need?

If you need live captions while someone talks, a recorded-file job does not fit. If the audio is already a file, such as a podcast episode, a batch job is the simpler fit, and Sume returns words[] with start and end times for every word.