AI music generation now spans large, high-fidelity foundation models and fast, user-facing tools that produce complete songs, stems, and sound effects. This topic covers the current landscape (as of 2026-06-23): multimodal models from major labs that can generate expressive audio and melody, alongside open-source and commercial systems optimized for speed, coherence, and integration. Key offerings illustrate different tradeoffs. ACE‑Step is an open-source, diffusion-based foundation model emphasizing fast, high-coherence outputs and a lightweight linear transformer conditioning strategy; its browser-first implementations focus on turning text or voice prompts into full, royalty‑free songs in seconds. Evoke Music’s relaunch as Amadeus Code centers curated AI sound packs (FUJIYAMA AI SOUND), a library of music and SFX, and MIDI topline generation for producers. Musci provides an integrated studio approach, combining multiple generation engines and audio tools to move ideas toward professional-quality tracks quickly. Trends to know: diffusion and transformer hybrids improve musical coherence; browser-first products lower the barrier to creating complete tracks; open models enable customization and local deployment; curated libraries and MIDI export facilitate DAW workflows. Practical considerations include model output quality, licensing and royalty terms, integration with existing production pipelines, and human-in-the-loop editing for arrangement and mixing. By comparing foundation models (e.g., DeepMind Lyria 3, Google Gemini) with fast consumer tools (ACE‑Step, Amadeus Code, Musci), creators can choose solutions based on fidelity, speed, openness, and workflow compatibility rather than marketing claims.