FlashSpeech
Efficient zero-shot TTS via latent consistency models
About
FlashSpeech is a zero-shot speech synthesis system from Zhen Ye and colleagues, published in April 2024 and presented at ACM MM 2024. It uses a latent consistency model with adversarial training to generate speech in one or few steps, which the authors say makes it about 20 times faster than comparable zero-shot TTS systems without losing quality. It also supports voice conversion. The main open implementation is a community reimplementation rather than the authors' full training stack, and the license was not confirmed in the sources I read.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Paper-reported: ~20x faster inference than prior zero-shot TTS systems with comparable quality, similarity and robustness.
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, metrics-v1). Evidence: April 2024 paper, presented at ACM MM 2024; Main open implementation is a community reimplementation, not the authors' full stack; License not confirmed in sources read — verify before commercial use. License: License not confirmed in sources read this session — check the repo before commercial use..
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 5.0/10
- Maintenance · recency, releases, issue load
- 6.0/10
- Openness · license + docs
- 6.0/10
- Manual review · sanity read + red-flag check
- 6.0/10
- · April 2024 paper, presented at ACM MM 2024
- · Main open implementation is a community reimplementation, not the authors' full stack
- · License not confirmed in sources read — verify before commercial use