YuE2 tips: KSampler ~16x faster, CFG sweet spots, LoRAs and 900s song fix
Diligent-Rub-2113 · reddit · 2026-09-20
Tested on an RTX 5060 Ti 16GB + 64GB RAM PC, the 3B local music model YuE2 gets major free wins over ComfyUI defaults: dropping KSampler steps from 32 to 6 (8x faster) plus switching dpm2 to dpmpp2m (another 2x) yields 16x faster sampling — a 5-minute song went from 34s to 2s with no audible quality loss in blind A/B tests.
Other findings: CFG 1 is fastest and most stable (CFG >3 damages frequencies); community LoRAs improve mastering (narlorajointv9) and reduce repetitive instrumental patterns (arlorainstv3abc, strength 0.7 on CLIP slots); updated ComfyUI raises max song length from 360s to 900s and fixes covers drifting out of sync after 50s. Prompt augmentation with Gemma 4 (int8 convrot) now takes 10s for style tags and lyrics, 2x+ faster. YuE2 also accepts editable ABC notation for melody/harmony/rhythm control — the author used it for a Super Mario ska cover.
Related event: YuE2 music model gets 16x sampling speedup, long-song fix(2 posts)→
More from Multimodal
- Izwi: open-source offline voice cloning in Rust, runs on 16GB M1 Pro — zinyando · 2026-11-12
- MiniMax H3 Image generates 3D game scenes in 10s, over 3x faster than GPT Image — HeyToha · 2026-11-12
- Dev releases local world model turning any image into interactive video, runs on RTX 3060 — lucidml_lover · 2026-11-12
- Ex-Meta CV researcher builds AdaL v2, a chat-driven video agent harness for GTM — Scobleizer · 2026-11-12
- MiniMax H3 image gen runs locally on AMD RX 9070, 25-30s per 4MP image — omglionheaded · 2026-11-12
- One reference, one prompt, three image models: side-by-side Nano Banana Pro vs GPT Image 2.6 test — gizakdag · 2026-11-12