Kunlun Wanzhi Unveils AI Music Model Mureka V7.5 and MoE-TTS
Kunlun Wanzhi Unveils Cutting-Edge AI Music and Speech Models
On August 15, 2025, Kunlun Wanzhi made waves in the AI industry with the launch of two groundbreaking technologies: the Mureka V7.5 AI music model and the MoE-TTS text-to-speech framework. These releases capped off the company's successful SkyWork AI Technology Launch Week, which showcased innovations across multiple AI domains.
Mureka V7.5: Revolutionizing AI Music Creation
The Mureka V7.5 model represents a significant leap forward in AI-generated music, particularly for Chinese compositions. The model demonstrates remarkable improvements in:
- Tone accuracy and vocal techniques
- Lyric articulation and emotional expression
- Understanding of Chinese musical styles and cultural elements
The system covers diverse genres including traditional folk songs, operas, classic pop hits, and contemporary folk music. Its enhanced Automatic Speech Recognition (ASR) technology analyzes minute vocal details like breath control and emotional fluctuations, resulting in more natural-sounding performances that significantly reduce the "mechanical" quality often associated with AI-generated vocals.
MoE-TTS: A Breakthrough in Speech Synthesis
The company simultaneously introduced MoE-TTS (Mixture of Experts Text-to-Speech), a novel framework that allows precise voice control through natural language descriptions. This innovation addresses long-standing challenges in descriptive TTS by:
- Enabling accurate generation of complex speech patterns like metaphors
- Combining large language model text capabilities with specialized voice modules
- Maintaining "zero knowledge loss" between modalities during transfer
Initial tests show MoE-TTS matching or surpassing commercial products in character alignment, even when using only open-source data.
Industry Impact and Future Applications
These technologies promise to transform multiple sectors:
- Music Production: Mureka V7.5 enables faster, more culturally authentic song creation
- Digital Assistants: MoE-TTS allows for more natural voice interactions
- Content Creation: Both models enhance immersive media experiences
The company plans to integrate MoE-TTS into its Mureka-Speech platform as a base model for character dubbing, making these advanced capabilities available to global developers.
Key Points:
- Mureka V7.5 delivers unprecedented quality in AI-generated Chinese music
- MoE-TTS introduces natural language control for voice characteristics
- Both models demonstrate Kunlun Wanzhi's leadership in multimodal AI
- Technologies available for public testing at www.mureka.ai