An AI Foley artist and soundscape blueprint generator. Upload an image or a sound, and get a cinematic sound design blueprint — layered Foley textures, ambient layers, and Suno-ready music prompts — powered by whichever LLM provider you choose.
![UI theme: dark obsidian with HarpStar electric blue / hot pink / coral / gold accents]
Creating sound effects and music for visual media takes time, skill, and expensive specialized tools. Most creators just need a starting point: what does this scene sound like, and what layers should I build?
Sonic Blueprint Engine analyzes your media asset (an image, a recording, even an MP4 video) and returns a structured soundscape blueprint:
- Suno Prompt Generator — audio in, three highly detailed Suno AI music prompts out
- Phoenix Oracle Foley Engine — image or audio in, structured JSON sound design out: mood, environment, audio categories, background sound suggestions, and engine logic
- Real-Time Sound Generator — describe a sound, preview a sample track immediately
# 1. Install dependencies
npm install
# 2. Configure your LLM provider
cp .env.example .env
# -> set at least LLM_API_KEY (or GEMINI_API_KEY) in .env
# 3. Run (dev mode — starts the API server + Vite dev server)
npm run dev| Env Var | Description | Default |
|---|---|---|
LLM_PROVIDER |
Provider: gemini, openai, anthropic, openai-compatible |
gemini |
LLM_MODEL |
Model name for the chosen provider | provider default (e.g. gemini-2.5-flash) |
LLM_API_KEY |
API key for your provider | — |
GEMINI_API_KEY |
Gemini-only fallback for LLM_API_KEY |
— |
LLM_BASE_URL |
Base URL for openai-compatible endpoints (Ollama, vLLM, LM Studio, Groq...) |
http://localhost:11434/v1 |
PORT |
HTTP port | 3000 |
The app is provider-agnostic. No code changes needed to switch models — just edit .env:
# OpenAI
LLM_PROVIDER=openai
LLM_MODEL=gpt-4o
LLM_API_KEY=YOUR_PROVIDER_API_KEY
# Anthropic Claude
LLM_PROVIDER=anthropic
LLM_MODEL=claude-sonnet-4-5
LLM_API_KEY=YOUR_ANTHROPIC_API_KEY
# Local models via Ollama (OpenAI-compatible endpoint)
LLM_PROVIDER=openai-compatible
LLM_MODEL=llama3.1
LLM_BASE_URL=http://localhost:11434/v1Note: inline audio input is currently only supported by the Gemini provider. Image analysis and text prompts work with every provider; audio analysis degrades with a clear message when the provider can't accept audio.
npm run build
NODE_ENV=production npm startSonic Blueprint Engine exists to democratize Foley and sound design — making cinematic audio creation accessible to every creator, not just studios. It started as an experiment in Google AI Studio and grew into a model-agnostic sound-design workbench.
Built by Harp★Star. The app carries the HarpStar brand watermark, which it verifies at startup and at render time — so the branding stays part of the product.
- Generates sound design blueprints and music prompts, not final mixed-ready audio (the Real-Time Sound Generator plays sample tracks as a preview).
- Blueprint quality depends on the underlying LLM.
- Audio analysis requires the Gemini provider (see above).
MIT