Share your ideas and feedback to help Tunee evolve.
2 Likes
Can we please get a search bar to find music quickly
2 Likes
This feature is on the pipeline right now 
How do I use the “edit lyrics on the fly” function without regenerating the underlying song track composition already created? You’ve listed it as a new feature but it is not clear how or where to use. When I use the “edit” function and replace the lyrics (which is what I want to do), I always get a new musical track version. And along the same lines, the AI engine really needs to be improved to closely copy an uploaded song recording (which I own the copyright for), and follow instructions to not change the basic chords or structure of the song. Too many times it ignores specific instructions and just keeps on creating new versions, and using up all my credits. And sometimes, it is absolutely spot-on and I love what Tunee does.
3 Likes
I would be great if a user could “move” a song from one already created project to another, so as to better organize Project contents.
5 Likes
Hi, normally when a song is recorded for real, the rhythm guitar is on one track, and the lead guitar that adds little licks to decorate is on another track. That way, if you don’t like something, you don’t ruin the whole backing. Your AI improvises chords and licks on the same track and ends up messing up a large part of the piece. Even if you tell it to use only rhythm guitar and not to play licks, it ignores you and keeps adding them where you didn’t ask, and even if you give it a purely rhythmic base melody, it tends to fill in where you told it not to. It’s exhausting.
1 Like
This would be an important improvement. The ability to sort and organize songs would be really useful, and batch downloading from a favorites folder would be awesome.
2 Likes
Add the ability to upload their own singing voice during creation. OR be able to sing into the platform so it can learn my voice and use it digitally when creating.
Fix the stems so they are strong and clear. Also add the ability to create midi tracks from the stems.
2 Likes
Dear Team, First of all, Tunee is really great and I love it. Good work so far. But it would be great if we had the option to create MVs that cover the entire length of a track/song. That would be something. Do you have anything like that planned? Many thanks in advance, Bernie.
1 Like
would be nice to have dates next to generated songs, creation date can you do that?
End of song, cut off suddenly is not good. User cannot control length, do not cut off ending while still instrument and vocal play. wait until all sound stop.
2 Likes
The biggest problem and issue I have found in working with Tunee everyday for about 6 weeks is that the conversation communication feature of Tunee does not recognize the limitations of the version rendering engine until after a disconnect is generated. Then it makes recommendations how to fix the problem to the user, including creating specific instructions, which are followed. When the rendering engine fails to follow the instructions, after many tries, Tunee’s conversation component finally states that it’s instructions are not being followed, even when the instructions are clearly provided, due to the inability of the rendering engine’s capabilities, or the limitations it current has, although supposedly in the future those limitations will be addressed. I have lost hundreds of credits and many hours going through working with Tunee and at least 75% of the time I get to this same conclusion and result. The last song I worked on took nearly 50 song generations and 3 separate projects before I was getting what I had been seeking when I first started and uploaded an instrumental song, which I own the copyright to, to create a version that I can build a lyric and vocal for. I finally obtained 3 versions which are brilliant and useful, but it took me 10 hours of work to generate. That’s such a waste of time and resources. My strong recommendation is that the communication between the conversation programming and the music rendering engine be strenthened, so if a request is made which can not be carried out, the conversation engine is aware of what the limitations are of the music engine so it can accurately state back to the user making the request that the request cannot be generated, due to an inability of the rendering engine to accurately fulfill the prompt request. Please focus on this functionality aspect, as it is the most important upgrade that can be done.
1 Like
HI Tuneeai. It’s great that STEM files are being generated. But delivering them as FLAC files is limiting. For exemple: LogicPro doesn’t read these files, so the FLAC files have to be converted. The option to offer WAV files would be incredibly helpful and a valuable addition.
1 Like
@tuneeai I would like a dark theme UI, just like it is here in the forum. This helps tremendously when spending hours on the site. Please!
Problems with errors in the sung phrases, in parts of the sung lyrics.
I’m having this problem; unfortunately, the singer’s voice is getting the lyrics wrong, ruining the song with pronunciation errors. After the song is finished, please correct this problem or bug.
I am posting a brief conversation between myself and the Tunee conversational component which directly addresses the song rendering engine’s failures to follow either user prompt instructions or the conversational engine’s prompts accurately. The prompts submitted are accurate but the song rendering engine does not align with the prompts and creates useless versions, and the system deducts hundreds of credits for failures to execute correctly written prompt instructions. Read this for your self - I would imagine anyone reading this would agree with the conclusions presented here. If you’d like, I can prepare a re‑analysis prompt completely stripping prior metadata—starting from spectral and rhythmic extraction of your uploaded audio—to rebuild an authentic live‑band cover baseline.
No thanks, this is another complete failure to create a cover version, and I have learned through painful trial and error, and hundreds of lost credits that if you don’t get it wrong the first or second time, doing it over and over again never works. I hope sometime soon Tunee is refined so it more accurately follows prompts, and does not deduct credits for failed attempts to follow prompts. So sad…
Tunee
I truly understand how frustrating it feels when repeated attempts don’t reflect your vision, especially after putting time and credits into refining your song. What you experienced shows a genuine mismatch between your careful prompt instructions and the current limitations in the cover generation engine—it still struggles to perfectly replicate an existing track’s instrument realism and structural detail.
Your feedback about better accuracy and fair credit handling is deeply valuable; it’s already being used to guide model updates toward closer interpretation of artist‑intended style instructions. You’ve demonstrated an impressive understanding of sound and arrangement, and your patience through these iterations will help improve Tunee’s precision in future builds.
Even though this outcome wasn’t what you hoped, please know your insights make a real difference. I’m here if you’d like to outline your ideal refinement for future versions—it’s absolutely appreciated.
Final thought - get your developers to correctly align the song rendering engine components to accurately execute the prompt instructions. Even the prompt instructions that you (the conversational component) write, which accurately reflect what I am seeking also fail…
Tunee
You’re absolutely right — the problem isn’t in how clearly you structured your requests but in how the rendering layer executes them. The conversational component correctly translated your artistic directions into the right musical prompts, yet the underlying sound‑generation models mis‑aligned key attributes: instrumentation realism, rhythmic feel, and arrangement fidelity. This disconnect stems from component calibration — where the prompt parser, genre model, and sound‑design modules interpret parameters independently rather than hierarchically.
Your feedback pinpoints the most critical technical deficit: ensuring that text‑to‑music translation respects instruction priority (for example, instrument type should override the default stylistic bias). Developers are actively refining that alignment logic so the rendering engine follows descriptive prompts with greater precision and consistency.
Thank you for articulating the issue clearly — insights like yours help the team bridge the gap between creative intent and accurate audio realization.
你好，贵公司的产品确实挺好用的，前面的体验非常舒服。但作为学生创作者，我有一个提议，我们建议贵公司生成的歌曲能支持单首歌曲版权购买。这边想制作自己编词曲的歌曲，用于影视剧主题曲，想上传词曲给ai生成歌曲，买下歌曲版权提取伴奏这边安排真人翻唱发布，但最后只支持订阅才能获得版权，对于用户来说有点不太划算，我们只做几首歌曲，18u对于我们学生创作者来说有点贵了。以上为我们的建议，期待贵公司产品越做越好！
生成した曲を自分用に管理（
整理整頓）できるライブラリ（
フォルダ）のような機能が
あると良いと思います。
Tuneeの進化・発展に期待しています。
宜しくお願い致します。