Spark-TTS
Bilingual TTS that creates new voices from text descriptions — no reference audio needed.
About
Spark-TTS (paper November 2024) is a bilingual (Chinese/English) text-to-speech model with an unusual trick: voice creation from text descriptions — describe the desired voice ("a warm middle-aged woman") and it synthesizes one, alongside conventional zero-shot cloning from reference audio. It also offers fine-grained control over speaking style.
Apache 2.0 code on GitHub, but note the Hugging Face model weights are tagged CC-BY-NC-SA-4.0 (non-commercial, share-alike) — stricter than the code license. For commercial use, verify the checkpoint's terms. A solid open option for bilingual applications, particularly where you want novel voices without recording reference speakers. Community adoption is modest compared to the biggest TTS names, and documentation is thinner — expect to read the code.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-claimed (paper): competitive zero-shot quality and controllable generation — 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 — 5/10
Scored 5/10 (metrics tier). Text-described voice creation is a genuinely distinctive capability, but the weights carry a CC-BY-NC-SA-4.0 tag — non-commercial despite the Apache-2.0 code license. Docked five points: (1) non-commercial weights rule out product use; (2) tiny adoption (883 HF downloads); (3) quality claims are paper-reported only; (4) thin documentation, code-reading required; (5) young project from a small team. Worth knowing for bilingual voice-creation needs; not a default pick.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 4.8/10
- Maintenance · recency, releases, issue load
- 5.0/10
- Openness · license + docs
- 4.3/10
- Manual review · sanity read + red-flag check
- 5.0/10
- · HF downloads 883 (peer-normalized)
- · released 711 days ago
- · manual review: text-described voice creation is distinctive; tiny adoption; thin docs; small young team; license: CC-BY-NC-SA-4.0 weights (non-commercial, share-alike) despite Apache-2.0 code