Inconsistent transcriptions
When using whisper (currently only the tiny version), I obtain different transcriptions every time I run them. I use the transcription using the code outlined below. Can anyone explain this or help me obtain consistent transcriptions?
import whispermodel = whisper.load_model("tiny")result = model.transcribe(audio)text = result["text"]
Hey @RikRaes - I would advise that you ask this question directly on the OpenAI Whisper repo, since you're using their package! https://github.com/openai/whisper
If you decide to use 🤗 Transformers, we can give you a hand here! See a code example here: https://huggingface.co/openai/whisper-tiny#long-form-transcription
@sanchit-gandhi Thanks, I do not have this issue when using Transformers so I might stick to that although I would like to know where the issue arises from, when using the original Whisper repo but I will ask that on their repo!