StyleStream 2.0: Fast and Controllable Streaming Voice Style Conversion
StyleStream 2.0 cuts streaming voice conversion latency to 520 milliseconds on a consumer GPU while adding text-based voice design and editing.
The authors report roughly twice the speed of StyleStream 1.0 while maintaining conversion quality. Natural-language prompts can define a target voice, and editing instructions can change a specific voice attribute while preserving the rest. In their experiments, the system achieved the strongest target-style fidelity across conversion, voice design and editing, while remaining competitive on intelligibility and preservation of unedited attributes.
HF Daily Papers' note
The authors report roughly twice the speed of StyleStream 1.0 while maintaining conversion quality. Natural-language prompts can define a target voice, and editing instructions can change a specific voice attribute while preserving the rest. In their experiments, the system achieved the strongest target-style fidelity across conversion, voice design and editing, while remaining competitive on intelligibility and preservation of unedited attributes.
HF Daily Papers' note
score 5