can anyone help me build this function I want to evaluate a class of student audio. so I can give marks based on the gpt evaluation.
1 Like
Not yet possible, likely before the end of the year.
3 Likes
This is not currently true. OpenAI has not released the audio input capabilities of gpt-4o.
2 Likes
You do realize what you are describing is not what we are discussing, yes?
Whisper is a speech-to-text model, so the output of that is going to be… <drumroll> text.
The model isn’t evaluating the audio.
Evaluating the audio would include things like,
- Understanding tone of voice
- Pronunciation
- Tempo
- Other sounds
- Etc
All whisper does is make a best guess at the words which might be present in an audio file.
Perhaps this class is, for instance, an ESL class. Being able to tell if the speaker is correctly pronouncing things correctly would be important.
Or maybe it’s a drama class and we need to evaluate if the student’s vocal performance is compelling?
So, again, there is no currently released model from OpenAI which can evaluate audio.
Only Gemini Pro 1.5 currently can evaluate audio
3 Likes