- Who it helps
- Speech researchers comparing controllable multilingual synthesis
- Availability
- Public paper first submitted April 20 and revised November 9, with lab demo and toolkit links. Revision contents were not diffed.
Inspect language-level results
The authors report that rankings vary by language and that combining controls remains difficult. An overall score can hide weaknesses in a particular language or speaking instruction.
Reproduction still needs checking
The lab publishes comparison and implementation links. Audio examples were not listened to and the evaluation toolkit was not run, so results remain author evidence.
MINT-Bench demo and leaderboard ↗MINT-Bench implementation ↗
What you can do
- Read the evaluation setup and its limitations before applying the result. Open reference ↗
Where the evidence stops
Paper and public project pages checked; no speech generated or evaluation reproduced.
Sources read. Public text checked November 15, 2026 UTC. No installation, model inference or full workflow executed.
Sources & further reading
Paper · paper
MINT-Bench paper ↗
April 20 paper revised November 9; ten-language evaluation and toolkit described. No reproduction.
Huakang Chen and coauthors · page checked 2026-11-15
GitHub · implementation
MINT-Bench implementation ↗
Repository reached from lab demo; setup not executed.
ASLP lab · page checked 2026-11-15
Web · demonstration
MINT-Bench demo and leaderboard ↗
Public comparison and code links inspected; audio not listened to.
ASLP lab · page checked 2026-11-15
Coverage note: Paper and public project pages checked; no speech generated or evaluation reproduced.
The record over time (1)
Dated source checks and published explanations. An observation date can differ from the original release date.
- · Added to Current
MINT-Bench evaluates generated speech across content, instruction following and perceived quality in ten languages.