Dev builds a vocal synth that models the vocal tract as branching acoustic tubes, not phoneme stitching
flngr · x · 2026-11-10
Developer jamesonhodge showcased a self-built vocal synth that, unlike Vocaloid's phoneme-recording stitching and pitch-shifting, models the vocal tract as a network of branching acoustic tubes. He is still tuning naturalness against recordings of real singers but says the results already sound impressively cool.
More from Multimodal
- Collector compares AI Mona Lisas: Fable's version beats Opus — morqon · 2026-11-10
- Fable 5.2 wired to Runway via MCP: video generation driven by the model — tlakomy · 2026-11-10
- Disabled artist open-sources project files of 154 artworks under CC for AI and human remixing — joonasvirtanen · 2026-11-10
- Turbo LoRAs are 'copium': oily look, no negative prompts, only for under 24GB VRAM — Rare_Subject2343 · 2026-11-10
- Ex-Apple researchers' Nuance AI debuts Emma, a full-duplex avatar that reacts while you talk — ditzikow · 2026-11-10
- ComfyUI long-video node breaks Viggle Remake's 345-frame limit with overlapping windows — Hoje-Na-IA · 2026-11-10