I'll look that up. **The latest ACE Music model is ACE Music v1, plus a newer ACE Music v1.1 Preview.** Your note is right that ACE-Step v1.5 is the older release. **ACE Music v1** (https://acemusic.ai/acemusic-v1/) - It generates songs from text descriptions and lyrics, and the lyrics support performance cues in parentheses, like "(hum softly)" or "(sing with a raspy voice)". - It ranks #1 on Artificial Analysis's AA-Music-Vocal v1.1 blind-listening leaderboard with 1160 Elo. Suno v6 is second at 1144, and Suno v6-mini, Mureka V9.5, Mureka V9 and Lyria 3.5 follow. That's ACE's own page citing the leaderboard, and I haven't checked it against Artificial Analysis directly. - You can export MIDI and split stems in the ACE Music app. **ACE Music v1.1 Preview** - It generates songs from melodies and scores, and it plans the music in ABC notation before generating audio. - You can edit the score and generate a new version from it. - It's a preview, so it's not the main release yet. **ACE-Step v1.5 (the older open-source line)** - It launched on Feb 3, 2026 under an MIT license, and it runs locally in about 4GB of VRAM. - A larger **ACE-Step 1.5 XL (4B)** version followed on Apr 2, 2026. - On acemusic.ai, ACE-Step v1.5 generations are free and unlimited, with commercial use included. The newer v1 models use credits. **What's coming:** ACE says it plans to open-source **ACE Music v1 Flash** in late 2026. It's described as a different balance of quality, speed and hardware needs, aimed at local use. I haven't seen a release yet. That also fits your "no lyrics" observation if you were using the free ACE-Step v1.5 route. Try ACE Music v1 if you want lyrics with performance directions.