SongBlend

SongBlend — AI Music Region Blending Workstation

SongBlend is a full-stack, browser-based audio workstation for decomposing AI-generated songs into musical regions and blending them into seamless new compositions. AI song generators (Suno and friends) are great at generating but lousy at editing — you cannot tell them to drop the second chorus, swap the bridge for a different one, or remix the intro and outro. That is exactly the gap SongBlend fills.

SongBlend handles all the audio science — BPM detection, key analysis, phrase segmentation, tempo stretching, and intelligent crossfading — so you can focus on creative curation.

Features

• Smart Region Detection — automatically segments songs into musical phrases, with BPM, key, and energy analysis.
• Visual Canvas — drag-and-drop interface for arranging regions into new compositions.
• Intelligent Joins — auto-selected blend modes (hard cut, crossfade, tempo stretch, pitch shift) based on compatibility scores, with manual override.
• Live Preview — in-browser playback with caching for instant re-preview.
• Compatibility Scoring — BPM, key, and timbre analysis to score region transitions, so the best join is usually the one the tool picks.
• Project Management — save and load multi-canvas projects with all settings preserved.
• Professional Export — high-quality WAV rendering with beat-matched joins.

How it works

1. Upload WAV files — AI-generated songs from Suno work great.
2. The app automatically analyzes and segments regions (verse, chorus, bridge), each with its detected BPM, key, and energy.
3. Drag regions onto the canvas to build your new song.
4. Preview joins and adjust blend settings.
5. Export your final composition.

Blend modes

SongBlend automatically selects the best blend mode based on region compatibility:

• Hard Cut — clean cut at zero-crossing (for similar BPM/key).
• Crossfade — equal-power crossfade (for smooth transitions).
• Beat Match — tempo-stretched join (for BPM differences).
• Pitch Shift — key-adjusted crossfade (for key mismatches).
• Silence Gap — brief silence (for dramatic transitions).

All modes can be manually overridden in the join settings panel.

Tech stack

• Frontend: React 18 + Vite + TypeScript, shadcn/ui, Zustand, WaveSurfer.js, Tone.js, dnd-kit.
• Backend: FastAPI (Python 3.11+), librosa for audio analysis (BPM, key, spectral features), pyrubberband for time-stretching and pitch-shifting, pydub for mixing and export.

All audio processing happens server-side for accuracy, and rendered previews are cached in-memory so replaying a join is instant.

Status

Deployment is pending. The app is dockerized and ready to deploy (Nginx serving the pre-built frontend and proxying /api/* to the FastAPI backend with a named volume for data persistence); a deployment target is being finalized.

Code: github.com/abulka/songblend (private for now)