Serious degradation of music generation quality in Mureca 9 model since May 12, 2026

Usage Issues
1

Dear Support Team,

I am writing to report a recurring issue regarding the quality of music generation on your platform. Ever since the release of Happy Horse 1.0 on May 12, 2026, the music generation process using the Mureca 9 model has significantly deteriorated. As of today, May 18, 2026, these issues remain unresolved.

Specifically, I am experiencing the following critical problems:

  1. Lack of User Control: The AI agent now automatically selects generation variants for me, completely bypassing my preferences and choices.
  2. Pronunciation Issues: The AI heavily mispronounces words, ruining the lyrical delivery.
  3. Audio Artifacts: The generated tracks are filled with severe digital artifacts, background noise, and technical glitches that ruin the composition.

Please look into this and resolve these issues as soon as possible. Unfortunately, this is not the first time that the introduction of new features has directly damaged the core music generation quality. It has become a frustrating pattern that severely disrupts our workflow.

Thank you for your prompt attention to this matter. I look forward to your update and a swift fix.

Best regards,
lime_lim

2

Thanks for taking the time to write this out so clearly. We really appreciate the detailed feedback, especially with the dates and the specific issues you noticed~~~

We understand that you’re seeing a quality change in Mureka 9 after the recent update, including less control over generation variants, pronunciation problems, and more audio artifacts or noise. These are important points, and we’ll pass them to the team for review. :headphone:

Model updates can sometimes affect generation behavior in ways that are not ideal for every workflow, so examples like this are very helpful for checking what changed and where the experience needs improvement.

For now, you may also get more stable results by making the prompt more specific, such as clearly describing the genre, vocal style, mood, language, pronunciation needs, and arrangement. If certain words are often mispronounced, we also suggest avoiding very difficult or unusual words when possible, or spelling them in a simpler phonetic way. AI generation still has a small amount of unpredictability, so clearer prompts and easier pronunciation can sometimes help improve the result. :sparkles:

If you have 【specific project links or examples】 showing the pronunciation issues, artifacts, or unwanted variant selection, feel free to share them as well. That would help the team investigate more accurately.