Open-source music gen lacks good cover mode: ACE-Step 1.5 disappoints, Minimax holds back audio input
maxiedaniels · reddit · 2026-09-11
A Reddit user asks for a decent "cover mode" music generation model, noting that nearly every new music gen release is text-to-audio only. They found ACE-Step 1.5's cover mode awful in testing, and had hoped Minimax Music 3 would ship audio input for its open weights — but it still hasn't.
More from Multimodal
- Native ComfyUI Workflow Does Instant Character/Style References With Voice, No refmod Needed — crinklypaper · 2026-09-12
- Runway launches MCP to generate images and videos inside ChatGPT, Claude and Cursor — runwayml · 2026-09-12
- Dev Builds Browser Battle Royale With 45 Smash Bros Characters in Days Using GPT-6 Astra — mattshumer_ · 2026-09-12
- Lev Manovich proposes 'Seven Worlds of Digital Art', a sociology-based new taxonomy — matdryhurst · 2026-09-12
- ElevenLabs ships Music v2.5 with richer melodies and commercial use on all plans — aziz4ai · 2026-09-12
- A Dreamina + GPT-6 Astra pipeline turns multi-angle product images into 360° interactive pages — HeyNayeem · 2026-09-12