Data security/privacy when using pyannote diarization on Huggingface #1401
-
Hi, I'm working on a basic tool for interview transcription. I'd like to use pyannote for diarization but I'd like to understand what information is sent to the Huggingface servers when pyannote calls their API. I'm aware of Huggingface's docuementation on security (https://huggingface.co/docs/hub/security) and security and compliance (https://huggingface.co/docs/inference-endpoints/security#data-securityprivacy). This question is to better understand what information is being transacted. What's in the payload and what comes back. Many thanks. |
Beta Was this translation helpful? Give feedback.
Replies: 3 comments 5 replies
-
pyannote can be used independently of Huggingface. |
Beta Was this translation helpful? Give feedback.
-
Hi @hbredin your guide was really good and I have it running. However, my pipeline seems to run smoother when working with the Huggingface connection. For reference I'm using it with WhisperX. Could you please clarify what the Huggingface connection does? Is just authenticating use and providing downloads of the model's and pipelines or is any of the text I'm processing being sent over the connection to huggingface's servers? Many thanks. |
Beta Was this translation helpful? Give feedback.
-
I have the same question. What exactly does the Huggingface connection do? Is any of my audio data sent to the servers? |
Beta Was this translation helpful? Give feedback.
pyannote can be used independently of Huggingface.
See "Can I use gated models and pipelines offline?" FAQ.