AI brief
A question on AI Stack Exchange asks whether RTFx, a metric calculated as audio length divided by transcription time, is a good measure for ASR models, noting that some models such as Whisper generate output tokens autoregressively.
Written by AI from AI Stack Exchange's published text.
Source
“The performance of the transcription task for ASR models is evaluated using RTFx, which is the audio length divided by the time the model takes to transcribe the audio.”
Interpretation or community commentary rather than straight reporting.
Read original story ↗