Seed-TTS
ByteDance's foundation TTS family — near-human naturalness, research-only
About
Seed-TTS (arXiv:2406.02430, June 2024) is ByteDance Seed team's family of large-scale autoregressive text-to-speech models, trained at what the report calls orders of magnitude larger data and model scale than prior TTS. It does zero-shot in-context speaker cloning from a short enrollment clip, fine-grained emotion control, and in-context speech editing, with team-reported speaker similarity and naturalness matching ground-truth human speech. A diffusion-based variant, Seed-TTS DiT, drops the phoneme-duration dependency and performs end-to-end generation. Like VALL-E 2, it has no public release — the technical report is the product. Note the family resemblance to ByteDance's shipping Doubao voice products, but Seed-TTS itself is not available as an API or weights.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Team-reported: speaker similarity and naturalness matching ground-truth human speech
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: ByteDance technical report arXiv:2406.02430, June 2024; Human-matching naturalness claims are team-reported; no public release to verify independently; Research-only: no API or weights.