Something terrible has happened with prompt > output results in the last month or so. My go-to prompts that would reliably provide amazing sounding results are now no longer doing so.
I have confirmed this by testing several different tracks generated before this regression, using the exact same prompt, Model, and seed#, and they generate results that are close but sound off, or completely not the same as what they used to be. Both models are affected.
You can view this playlist to see the comparison:
This leads me to believe that something has changed with how the prompts are being interpreted to the actual results. It takes AGES to land on prompts we like, and now it seems I have to start all over again, or (if the models are permanently nerfed in some way) abandon Udio altogether.
Something has definitely changed and I am not happy. I have already canceled my yearly subscription renewal and will not renew it unless I can find some prompts that sound good or the old prompt structure works once again.