Skip to content

Resonant 0.4.0

Choose a tag to compare

@calesthio calesthio released this 02 Aug 11:41
· 8 commits to main since this release

Resonant 0.4.0

Resonant 0.4 adds an optional local generative-music engine and a substantial visual refresh while keeping ordinary projects and startup lightweight.

Highlights

  • Optional ACE-Step 1.5 post-install module with isolated dependencies and shared model storage outside project folders.
  • Hardware-aware RTX 4060 Laptop profile: ACE-Step 2B Turbo, 0.6B planner, PyTorch backend, batch size 1, and CPU offload.
  • New ACE-Step Studio for prompting instrumental or vocal music, choosing duration and tempo, and adding generated WAVs directly to the Audio track.
  • Engine start, stop, storage reporting, generated-output count, and complete independent removal controls.
  • MCP get_ai_generator_status and generate_ai_music tools for agent-driven local generation followed by Resonant import, arrangement, mixing, analysis, and rendering.
  • Copy-ready MCP onboarding for Codex, Claude Code, generic STDIO clients, and PowerShell, with workspace confinement and sanitized examples.
  • Songwriter Studio language packs for English, Standard Mandarin, Hindi, Korean/K-pop, Latin American Spanish, and Japanese, exposed through both the desktop app and MCP discovery.
  • Three creative starters: Neon Raga, Midnight Drive, and Velvet Orbit.
  • Premium dark workstation refresh with clearer panel depth, teal/violet hierarchy, stronger contrast, and offline-safe Space Grotesk/Inter/JetBrains Mono font stacks.
  • New Reso mascot and theme-aware Resonant wordmark for the public project page.
  • Improved hover, focus, disabled, scrollbar, modal, transport, clip, mixer, and keyboard-accessibility states.
  • Model caches remain centralized so multiple projects never duplicate the multi-gigabyte runtime.

Validation

  • Unit/domain suite, TypeScript checks, delivery-graph validation, and production build pass.
  • MCP smoke covers all 29 registered tools, language-pack discovery, optional AI-generator status, path confinement, revision conflicts, deterministic render, analysis, and undo.
  • Electron smoke covers the optional ACE-Step post-install screen plus instruments, Flow, audio import/record, Arrange, Mix, save/open/recovery, playback, and WAV export.
  • Real ACE-Step model installation and generated-sample measurements are recorded with the release artifacts.