IndexTTS
Industrial-grade controllable zero-shot TTS from Bilibili — voice cloning with emotion, speed, and pinyin control.
About
IndexTTS is Bilibili IndexTeam's GPT-style zero-shot text-to-speech system (1.0 released March 2025): clone a voice from a short reference clip, control emotion, speaking speed, and pauses via punctuation, and correct Chinese character pronunciation with pinyin. The team claims state-of-the-art results versus XTTS, CosyVoice2, Fish-Speech, and F5-TTS on their evaluations.
Licensing needs care. IndexTTS 1.x shipped under permissive terms, but IndexTTS 2.5 uses a custom Bilibili Model Use License — not OSI open source, with a separate license required above 100M monthly active users or RMB 1B annual revenue, plus downstream and prohibited-use obligations. Check the exact checkpoint's LICENSE file before commercial use; terms have tightened across versions.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-claimed (IndexTeam): SOTA vs XTTS/CosyVoice2/Fish-Speech/F5-TTS on the team's own test sets — treat as provider-claimed. Independent standardized benchmarks are limited.
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 6/10
Scored 6/10 (metrics tier). Strong feature set (emotion/speed/pinyin control) and credible team (Bilibili), with real adoption in the Chinese TTS community. Docked four points: (1) license terms tightened across versions — the 2.5 custom Bilibili license has MAU/revenue thresholds and is not open source; (2) quality claims are team-reported on their own test sets; (3) documentation is primarily Chinese-first; (4) version-to-version license drift means you must re-verify per checkpoint.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 6.3/10
- Maintenance · recency, releases, issue load
- 5.0/10
- Openness · license + docs
- 4.6/10
- Manual review · sanity read + red-flag check
- 6.0/10
- · HF downloads 12,404 (peer-normalized)
- · released 567 days ago
- · manual review: strong feature set (emotion/speed/pinyin); license tightened across versions; docs Chinese-first; license: no license tag on HF checkpoint; 2.5+ uses custom Bilibili Model Use License — not OSI; MAU/revenue thresholds; verify per checkpoint