prism-v/open_vocoder_leaderboard
0
🎙️ PRISM-V: Open Evaluation of Pretrained Neural Vocoders
Quality · Robustness · Generalization · Efficiency A multidimensional benchmark for understanding how modern pretrained neural vocoders perform across datasets, architectures, acoustic conditions, and deployment constraints.
Overview
PRISM-V evaluates 15 publicly-released, pretrained neural vocoder checkpoints in a fully training-free, zero-shot setting across 4 diverse English corpora (LJSpeech, LibriTTS, VCTK, Free_ST), with GPU hardware profiling under Edge Profiling.
Key Features
Benchmark Navigation
Citation
@misc{purohit2026prismv,
author = {Ravindrakumar M. Purohit and Hemant A. Patil},
title = {{PRISM-V}: Multidimensional Evaluation of Pretrained
Neural Vocoders for Speech Synthesis},
year = {2026},
howpublished = {Hugging Face Space},
note = {Open neural vocoder evaluation leaderboard}
}