Seed-Music
ByteDance's unified framework for controllable vocal music generation and in-track editing.
About
Seed-Music is a suite of music generation and editing systems from ByteDance's Seed Team, described in a September 2024 technical report. Rather than betting on one modeling approach, it combines autoregressive language modeling with diffusion to cover two workflows: controlled generation and post-production editing. For generation it takes lyrics, style descriptions, audio references, musical scores and voice prompts, and produces vocal or instrumental music. For editing it lets a user change lyrics, melody or timbre directly inside an existing audio track.
The report is unusually broad for a music paper: vocal and instrumental generation, singing voice synthesis and conversion, plus interactive editing tools, all aimed at both novices and professional producers. Everything shown so far is demo audio on the project page, though. No weights, code or API have been released, and there are no published benchmarks, so the quality claims rest entirely on ByteDance's own examples.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
No published benchmarks; evaluation is demo-based (none found).
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (verified tier, metrics-v1). Evidence: detailed technical report from ByteDance's Seed Team, broad capability suite covering generation and editing. Strengths: unified AR plus diffusion design, fine-grained control over lyrics, melody and timbre, serious treatment of pro workflows. Docked for: (1) no public weights, code or API; (2) no published benchmarks, quality claims rest on provider demos. License: proprietary research, no public release.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 6.0/10
- Maintenance · recency, releases, issue load
- 6.0/10
- Openness · license + docs
- 2.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · ByteDance Seed Team technical report, Sep 2024 (arXiv 2409.09214)
- · Unified AR language modeling plus diffusion framework
- · Generation from lyrics, style, audio refs, scores, voice prompts
- · In-track editing of lyrics, melody, timbre
- · No public weights, code or API