Mureka V9

Mureka V9 is a professional-grade music generation model developed by Mureka (Skywork AI) that synthesizes complete songs from text prompts and lyrics. The model produces end-to-end compositions including vocals, melodies, and instrumentation in a high-fidelity stereo format. It is designed to handle a wide range of musical genres and structures, emphasizing production-quality audio suitable for content creation and professional demos.

The model utilizes a specialized architecture known as MusiCoT (Music Chain-of-Thought), which allows the system to plan a song's structural layout before generating audio frames. This approach improves the coherence of transitions between song sections, such as verses and choruses, and ensures that the melodic themes remain consistent throughout the track. V9 supports a maximum duration of 330 seconds per generation and outputs audio at a 44.1kHz sampling rate.

Capabilities and Control

Mureka V9 provides granular steering through a dual-input system consisting of lyrics and style descriptions. It features multilingual support for ten languages, including English, Chinese, Japanese, Korean, Portuguese, Spanish, German, Italian, French, and Russian. Users can specify technical musical parameters such as tempo (BPM), mood, and specific instrumental layers to refine the output. The model also supports advanced features like voice cloning and stem separation for post-production workflows.

Prompting Guide

To achieve optimal structural results, users are encouraged to use functional tags within the lyric input to define song segments, such as [Verse], [Chorus], [Bridge], and [Outro]. For the style prompt, combining specific genres with descriptive textures (e.g., "Analog synth," "Acoustic folk," "Intimate jazz vocals") yields more predictable results than abstract descriptions. The model interprets punctuation and line breaks to determine the rhythmic phrasing and emotional delivery of the vocals.