Diff-HierVC
Diffusion-based hierarchical voice conversion with robust pitch generation and zero-shot speaker adaptation.
About
Diff-HierVC tackles voice conversion as a two-stage diffusion problem. DiffPitch first generates the target speaker's F0 contour — the first use of a diffusion process for pitch generation, per the authors — and DiffVoice then renders the converted mel-spectrogram conditioned on that pitch plus a disentangled source-filter representation, with a masked prior improving speaker adaptation. The paper was presented as an oral at INTERSPEECH 2023 and the official PyTorch implementation is public.
In zero-shot conversion scenarios the authors report a character error rate of 0.83% and an equal error rate of 3.29%. It is a compact research artifact (a few hundred GitHub stars) rather than a production voice product — fine for experiments in pitch-aware conversion, not a hosted service.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
- Provider-reported (INTERSPEECH 2023 paper, arXiv 2311.04693): CER 0.83%, EER 3.29% in zero-shot VC scenarios
- No independent third-party benchmarks published
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 (TIER metrics, metrics-v1). Evidence: INTERSPEECH 2023 oral paper (arXiv 2311.04693); official PyTorch implementation public (238 GitHub stars). Strengths: peer-reviewed oral, novel diffusion pitch generation. Docked for: (1) small community footprint and unclear license; (2) eval numbers author-reported only, no reproduction studies found. License: not stated.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 3.0/10
- Maintenance · recency, releases, issue load
- 3.0/10
- Openness · license + docs
- 4.0/10
- Manual review · sanity read + red-flag check
- 5.0/10
- · INTERSPEECH 2023 oral (arXiv 2311.04693); official PyTorch implementation public
- · 238 GitHub stars (verified 2026-10-16); no commits visible since 2023
- · Author-reported CER 0.83%, EER 3.29% zero-shot; no reproduction studies found