YuE2 hands-on: local music model rivals Suno in some cases
noxsanguinis · reddit · 2026-09-16
A Reddit user shared hands-on results with the open music model YuE2: quality beats all previous local models (Ace, MiniMax Music 3) and in some cases rivals Suno — with no subscription or credits, making it the best local option today.
Known flaw: distortions and skipped/mangled lyrics tend to appear toward the end of longer songs.
Workflow:
- YuE2 text-to-music template in ComfyUI (update to latest)
- BF16 model instead of the int8 quant
- Only changed setting: maxduration=600
- Hardware: RTX 4090 + 32GB RAM
The author also demoed a full song generated from a very detailed heavy-metal style prompt (Drop C tuning, 180 BPM, A minor, horror atmosphere instrumentation).
Related event: Open-Source Music Model YuE2 Rivals Suno in User Tests(3 posts)→
More from Multimodal
- Midjourney wallpaper recipe: sref + profile mix for eerie film-blur portraits — michaelrabone · 2026-11-02
- Cavallaero Teaser: A Painterly Rider Races Through a Golden Forest — CommunicationTop4244 · 2026-11-02
- Anima's snapref Offers a New Reference-Based Fix for Character Consistency — Intrepid_Swim_8669 · 2026-11-02
- Pencil-textured folding fan: surreal product concept render goes viral — Delahuntagram · 2026-11-02
- Free Midjourney SREF gallery adds 1,200+ handpicked style codes overnight — tisch_eins · 2026-11-02
- Adobe's EditAgent beats prior image-editing agents but needs ~10x the compute — emmanuelvivier · 2026-11-02