Official documentation
whisper/transcribe.py (openai/whisper, main branch)
OpenAI, GitHub (openai/whisper), n.d.
Read the originalRead on October 17, 2026
- What it is
- The open-source Whisper code for the transcribe function, including the docstring for the no speech threshold that decides whether a 30-second segment is treated as silence.
- How much weight it carries
- The reference implementation published by the model's authors. It is the code that runs when the open-source model is used, but it is not the documentation for OpenAI's hosted transcription API.
Facts taken from this source
- The docstring of OpenAI's open-source Whisper code says a segment is treated as silent when its no-speech probability is above the threshold and its average log probability is below the log probability threshold.
“If the no_speech probability is higher than this value AND the average log probability over sampled tokens is below” (transcribe docstring)