Benchmarking CPU-Based TTS Models via UTMOS
gvij · reddit · 2026-07-06
The author benchmarked four small TTS models on an Intel Xeon (4-core) CPU, scoring them with the objective UTMOS MOS metric: Kokoro 82M, Supertonic 3, Inflect-Nano (4.6M parameters), and Kyutai's newly released Pocket TTS (about 100M parameters). The results showed Kokoro 82M had the best audio quality (UTMOS approx. 4.44); thanks to its streaming LM architecture, Pocket TTS maintained a nearly flat RTF (0.69-0.76) regardless of text length variations, meaning inference cost scales linearly with output length and exhibits a uniquely flat scaling profile.
Related event: Kyutai Open-Sources Pocket-TTS for CPU-Based Voice Cloning(4 posts)→
More from Multimodal
- MOSS Image Edit vs Fibo-2.0-edit: 5-grid comparison of instruction-based image editing — Illustrious_Row_9971 · 2026-10-27
- Cinematic title sequence built with ComfyUI: LTX-2.5 + MiniMax H3 image-to-video — magratheya_64 · 2026-10-27
- Fish Audio ships open-weight Drama 3 TTS; creator shows turning old posts into narrated videos — alex_verem · 2026-10-27
- Redditor crafts photorealistic 'morning bus ride' short film entirely in ComfyUI — no3us · 2026-10-27
- New blur/anti-blur slider LoRA for MOSS lets you dial background sharpness with weights — NoMouse9610 · 2026-10-27
- PixGR: Pixel-space generative decoding for fast feed-forward 3D reconstruction — kwangmoo_yi · 2026-10-27