공시 • Jul 23
Skywork Officially Launches Mureka V7
On July 23, Skywork unveiled its latest music model, Mureka V7, alongside an all-new audio model, Mureka TTS V1. The latter introduces groundbreaking Voice Design capabilities for timbre customization, delivering an exceptional experience to users worldwide. Previous iterations, including the Mureka O1 and V6 models, have garnered widespread global recognition, fueling nearly 3 million new user registrations since their official launch in late March. The latest upgrade Mureka V7 transforms users' musical inspiration into a single click. It marks not just a quantity leap in efficiency, but a fundamental evolution in sound quality. Mureka V7 delivers superior music quality. With enhanced melodic development, better arrangements, and more realistic vocal /instrument rendering, it ensures radio-ready excellence in every creation. Mureka V7 redefines musical innovation. Compositions generated by Mureka V7 can actively inspire creator workflows. Of course, these auditory enhancements are rooted in measurable technical advancements: An increase in average performance rating of V6's 43.4% to V7's 57.7% represents a significantly higherability of generating desirable songs from lyric inputs. With 44% enhanced vocal realism and expression, the output consistently sparks reactions like "This can't be AI-generated!" With nearly 100% improvement in overall audio quality, Mureka V7 overcomes homogenized AI sound profiles. Mureka V7 introduces major enhancements to MusiCoT (Analyzable Chain-of-Musical-Thought Prompting, a specialized Chain-of-Thought (CoT) prompting framework designed for music generation. This upgrade delivers significant improvements in the coherence and sonic performance of AI-generated music mainly through three key innovations: Structure-First, Generation-Second: Cognitive Modeling of Human Compositional Processes: MusiCoT first guides the model to generate a comprehensive musical structure - including sections, emotionalarcs, and instrumentation layouts - prior to audio token generation. This approach effectively resolves the "local optimization trap" inherent in traditional AR models, enabling compositions with more natural progression and dynamic development. Interpretable musical Chain-of-Thought with Style-Adaptive Capabilities: By integrating CLAP (Contrastive Language-Audio Pretraining), MusiCoT constructs a semantically grounded musical chain-of- thought. The improvement not only enhances structural analyzability and controllability but also accepts reference audio inputs of variable lengths as style prompts, significantly improving model flexibility in stylistic replication and variation while mitigating risks of direct copying. Extensive experimental results demonstrate MusiCoT's superior performance across both subjective and objective metrics. The technology significantly outperforms conventional solutions in structural integrity, melodic cognition, and overall musicality, achieving state-of-the-art benchmarks in multiple evaluations. The latest iteration of MusiCoT achieves unprecedented alignment with human compositional cognition at the structural level. Through scale-up of the proprietary datasets of Mureka and embedding refined information granularity, the latest solution features enhanced controllability and scalability, largely bridging the text-audio modality gap. Future iterations will further advance the system's creative potential and industrial-grade capabilities. Mureka TTS V 1: Unlimited Timbre Choices, One-Click Customization: Moreover, the newly launched Mureka T TS V1 supports the Voice Design feature, which means users can generate custom vocal timbres simply by describing desired voice characteristics in text. Mureka TTSV1 also demonstrates significant improvements across key technical metrics in voice design function: Scoring 4.6 for speech sound quality, surpassing ElevenLabs (4.36), with significantly superior auditory performance across all competitors; Scoring 4.64 for word segmentation and phrasing accuracy, indicating superior comprehension of input intent and linguistic expression; Significantly improved overall listening experience, delivering markedly smoother and more coherent audio output. Mureka V7 represents a paradigm shift in AI music generation - from mere technical utility to genuine artisticempathy. This isn't just an upgrade; it'spathy. This isn't Just an upgrade; it's not just an upgrade; it's a paradigm shift in an upgrade; it's paradigm shift in AI music generation. This isn't just an update; it's not just an increase in an upgrade; it's a fundamental evolution in an upgrade; it's an upgrade; it's's's not just an upgrade; It's an upgrade; it's an increase in an upgrade; It's's an upgrade; it's new experience, it's an upgrade. This isn't just an increase in AI music generation - from a paradigm shift in an upgrade.