So, now I have a little more clarity on what is actually happening. As a music lover who wants to create his own music, it is essential to provide Tune with the ability to process the audio file that I attach as a reference, because if I can’t make a musical notation of my melody - and even if I can, there is no possibility of attaching a musical notation to the prompt or I don’t know about it? How else can I create my own music other than singing the basic melody and attaching it as a reference file? Tune needs to develop an algorithm that will clearly distinguish such a situation from singing an already known composition by another author, and therein seems to be the biggest problem, how can Tune distinguish an original composition in a sea of already existing music? According to the Berne Convention on Copyright (which came into force in 1971), copyright arises automatically by the act of creating a work, without the need for formal registration or certification. So, my reference audio file, recorded as my original melody, is already protected as my copyrighted work, by the act of recording it, and I am its author. If someone claims that their rights have been violated, the burden of proof falls on them, not me. This is an excellent point for my development proposal: Tunee could accept a user statement of authorship (e.g., ‘This is my original melody’) as sufficient evidence for accurate tracking, without additional certificates, because the Berne Convention protects a work from the moment of creation. That would, in my opinion, solve the problem of reinterpretation and allow for greater creative freedom.
I appreciate the response, but I have to be honest—this feels like you’re missing the point of the community’s feedback.
The Issue Isn’t How We’re Prompting
You mentioned preparing a user guide to help with “stability and music quality under the new models.” With respect, the problem isn’t that users don’t know how to prompt properly. The problem is that the models themselves produce inferior output compared to what we had before.
I’ve been using Tunee successfully for months. My prompting hasn’t changed. My creative approach hasn’t changed. What changed was your backend—and the results went from studio-quality to generic and amateur-sounding overnight.
A user guide won’t fix a fundamental model quality issue. If your new models require completely different prompting strategies just to achieve what the old system did naturally, that’s not user error—that’s a downgrade.
Understanding Your Constraints vs. Accepting Lower Quality
I genuinely understand you’re facing external constraints around copyright and compliance. That’s a real challenge, and I respect that you’re trying to keep the platform sustainable.
But here’s the thing: sustainability means keeping paying users, and we’re paying for quality. If the new models can’t deliver the same level of output as before, then they’re not a viable solution regardless of the legal/compliance benefits.
What Would Actually Help
Instead of teaching us how to work around inferior models:
-
Be transparent about whether these new models can achieve the same quality as before, or if this is the new ceiling
-
If they can’t match the old quality, tell us what the roadmap is for getting back there
-
Poll the community on potential solutions (different models, API integrations, etc.) instead of assuming you know what we need
I want Tunee to succeed. But a user guide isn’t going to change the fundamental audio quality issues we’re all experiencing. Please don’t treat this as a user education problem when it’s clearly a product quality issue.