Drama3 TTS lands in audio.cpp: up to 5x real-time on RTX 5090, <8GB VRAM mode too
Acceptable-Cycle4645 · reddit · 2026-10-16
audio.cpp now supports Fish Audio's Drama3 text-to-speech model, contributed by JarodMica. On an RTX 5090 (Q4k), full CUDA mode hits a warm RTF of 0.195 (5x real-time) at 26,990 MiB peak VRAM; a CPU weight-offload mode with staged release runs 2x real-time in under 8 GB of VRAM (7,782 MiB at RTF 0.521 with 256-token chunks). Vulkan and Metal backends also work but need polish — token-limit hits on Vulkan and unexpected silence in multi-speaker generation have been observed. More iterations are planned before a minor release; contributors are welcome, though shared-framework changes must be additive and opt-in to avoid regressing 100+ other models.
Related event: Drama3 TTS Model Lands in audio.cpp with Up to 5x Real-Time Speed(2 posts)→
More from Infra
- New hardware day: local LLM deployment the moment it arrives — lxfater · 2026-10-16
- Together AI's score centering fixes RL training collapse under quantized inference — nrehiew_ · 2026-10-16
- Inference is splitting prefill and decode across specialized chips, not just H100s — AccBalanced · 2026-10-16
- Rebasely launches in beta: Git hosting built for humans and coding agents — handotdev · 2026-10-16
- Drama3 TTS lands in audio.cpp: up to 5x real-time on RTX 5090, <8GB VRAM mode too — Acceptable-Cycle4645 · 2026-10-16
- Three CoreWeave co-founders have sold $3.3B in shares since IPO amid credit worries — SumitGup · 2026-10-16