MVSEP separates full audio tracks into distinct components including vocals, instruments, and specialized stems, serving musicians, producers, and DJs who need isolated elements for remixing, analysis, or practice. The platform offers granular splitting capabilities that go beyond basic vocal-instrumental separation, handling multi-stem extraction for detailed audio work.
MVSEP: Specialized AI Stem Separation with 20+ Models
MVSEP utilizes artificial intelligence to perform audio stem separation, breaking down complete tracks into individual elements. This includes isolating vocals, instrumental tracks, bass, drums, piano, and guitar. The tool extends its capabilities to more granular separations, such as lead/rhythm guitar, plucked strings, percussion, keys, brass, woodwind, and choir. It can even differentiate specific vocal ranges like soprano, alto, tenor, and bass. Beyond separation, MVSEP offers voice cloning, text-to-speech, crowd noise removal, and the conversion of melodic audio recordings into MIDI notes.
Audio Engineers, Music Researchers, and Remix Producers
The tool caters to a wide audience, including musicians, audio engineers, DJs, remixers, and karaoke creators. Its features support various use cases, from extracting audio elements for detailed analysis and remixing to creating instrumental versions for karaoke. It’s suitable for individuals and organizations ranging from freelancers and startups to mid-size businesses and enterprises.
Demucs, Spleeter, MDX-Net, and Custom Models
MVSEP supports various audio file formats for upload, including MP3, WAV, FLAC, and M4A. Users can download output files in MP3, WAV, FLAC, and M4A, with premium options for higher fidelity WAV (32-bit float) and FLAC (24-bit). The platform integrates over 100 advanced AI models, such as BS Roformer, Demucs4, MDX23C, and SCNet XL, offering diverse separation configurations. It also employs specialized models like VibeVoice for voice cloning and text-to-speech, a BSRoformer-based model for crowd removal, and Spotify’s Basic Pitch for audio-to-MIDI conversion. MVSEP provides API access for custom workflow integration and frequently updates its models, focusing on improving Signal-to-Distortion Ratio (SDR) and balancing "Fullness" and "Bleedless" metrics for separation quality.
Free Web Tool, API for Developers, Batch Processing
MVSEP operates on a freemium model, offering a free tier with certain limitations. Users can perform a limited number of daily separations (e.g., 50) with a maximum file size of 100MB and a 10-minute length. Free output is typically MP3, and processing occurs with lower queue priority. Premium features are unlocked via prepaid credits, not a subscription. These credits provide higher file limits (up to 1000MB and 100 minutes), unlimited daily and concurrent jobs, access to resource-intensive "Ensemble" models, faster processing, and higher quality output formats like WAV and FLAC. All features available before the introduction of premium options remain free.
Queue Times on Free Tier, Complex UI for Beginners
While MVSEP offers strong functionality, free users may experience significant waiting times due to lower queue priority, especially during peak usage. Some advanced models demand substantial computational resources, including VRAM. Users might encounter issues with 0-byte output files if the input format isn’t read correctly, often requiring re-encoding. There’s also an inherent trade-off between preserving all target content ("Fullness") and minimizing unwanted signals ("Bleedless"), meaning optimizing one might impact the other.