Written by Ravindrakumar M. Purohit · September 2026
We are releasing PRISM-V: A Multidimensional Evaluation of Pretrained Neural Vocoders for Speech Synthesis.
Why Another Vocoder Benchmark?
Traditional neural vocoder evaluations predominantly report PESQ and MCD exclusively on clean, single-speaker studio recordings (e.g., LJSpeech). In production speech synthesis and voice conversion, however, practitioners deploy models on diverse voices, accented speech, and real-world noisy audio — often across hardware constrained to sub-realtime latencies.
PRISM-V fills this gap with a comprehensive, reproducible evaluation framework covering:
| Dimension |
|---|
| P — Perceptual |
| R — Reconstruction |
| I — Intelligibility |
| S — Speaker |
| M — Model Efficiency |
Live Leaderboard
Explore the full interactive leaderboard with Plotly visualizations, per-corpus breakdowns, radar charts, and audio sample comparisons:
🔗 PRISM-V: A Multidimensional Evaluation of Pretrained Neural Vocoders for Speech Synthesis
The leaderboard will be updated continuously as new pretrained models are submitted. If you have a vocoder model and want it evaluated, open a pull request or raise an issue in the GitHub repository.
Citation
If PRISM-V is useful for your research, please cite:
@misc{purohit2026prismv,
author = {Ravindrakumar M. Purohit and Hemant A. Patil},
title = {{PRISM-V}: A Multidimensional Evaluation of Pretrained Neural Vocoders for Speech Synthesis},
year = {2026},
howpublished = {\url{https://iamshreeji-copy2.github.io/open_vocoder_leaderboard/}},
note = {Open neural vocoder evaluation leaderboard}
}