| Repository | Description |
|---|---|
| obsidian-neural-central | Central inference server |
| obsidian-neural-provider | Provider kit — run a GPU node on the network |
| obsidian-neural-frontend | Storefront & dashboard |
| obsidian-neural-controller | Mobile MIDI controller app |
| ai-dj ← you are here | VST3/AU plugin (client) |
Live AI music generation in your DAW
"I've cycled through almost every AI music tool on the market, but Obsidian is the first one that actually feels like a real production tool rather than a novelty. While other AI apps try to replace the songwriter, Obsidian treats AI like a powerful, playable instrument. The 8-track MIDI-triggering is a total game-changer. Because it lives directly in my DAW, there is zero latency and zero break in my workflow. It stays perfectly locked to my project's tempo and vibe, serving as the ultimate intelligent jam partner VST."
— Moteka, Electronic Music Producer SoundCloud · Instagram
OBSIDIAN Neural now features 8 specialized AI engines in a single interface. You can now assign a different "brain" to each of your 8 tracks, each optimized for specific tempos and styles:
- stabilityai/stable-audio-open-1.0 — The versatile foundation for full-mix textures and drum loops.
- Best BPM: Versatile (80–160 BPM).
- RoyalCities/Foundation-1 — Surgical tag-based control for melodic and harmonic phrasing.
- Best BPM: 100–150 BPM (Sweet spots: 120, 128).
- adlb/Audialab_EDM_Elements — High-energy EDM leads, supersaws, and plucks.
- Best BPM: 100–150 BPM (Ideal at 128).
- RoyalCities/RC_Infinite_Pianos — High-fidelity grand and electric piano performances.
- Best BPM: 100–150 BPM (Optimal at 120).
- RoyalCities/Vocal_Textures_Main — Choral, operatic, and atmospheric vocal chord progressions.
- Best BPM: 100–150 BPM (Best for atmospheric pads).
- santifiorino/SAO-Instrumental-Finetune — Melodic trap, lofi jazz rap, and modern indie stems.
- Best BPM: 75–160 BPM (Great for 90 BPM Lofi).
- gab-gdp/StableBeaT — The drum machine: specialized in trap beats and 808 grooves.
- Best BPM: 75–160 BPM (Tightest at 140).
- atoof/gluten_v1 — Specialized engine for loopable melodic trap and wavy motifs.
- Best BPM: 90–160 BPM (Optimal at 135).
🎁 New accounts now receive 20 free credits on signup — no credit card required.
⚡ Quick Start: Get your API key and start generating in minutes.
📄 Late Breaking Paper — AIMLA 2025 — Presented at the AES International Conference on AI and Machine Learning for Audio, Queen Mary University London.
🎓 Tutorial — From DAW setup to live performance (French + English subtitles)
Type words → Get musical loops in ~30s. No stopping your creative flow.
- 8-track sampler with MIDI triggering (C3-B3)
- 4 pages per track (A/B/C/D) — Switch variations instantly
- Per-page ADSR envelope — Shape the dynamics of each variation independently, editable directly on the waveform
- 8 sequences per page — 256 total patterns for complex live sets
- 16-step sequencer with multi-measure support
- Quantized page changes — Seamless transitions locked to measure boundaries
- Draw-to-sound — Sketch your ideas visually and let AI interpret them musically
- Perfect DAW sync — Auto time-stretch to project tempo
- Non-blocking generation — No pre-recorded samples, renders in background
OBSIDIAN Neural is NOT a song generator like Suno or Udio. It's a performance tool: you build your track loop by loop, you're the composer, AI is your loop generator.
OBSIDIAN Neural runs on a distributed GPU provider network. When you generate a loop, the request is routed to an available community provider. If none is available, the system falls back to a cloud inference service.
VST Plugin → Central server → Provider GPU pool → WAV returned to DAW
↓ if unavailable
Cloud inference fallback
Revenue sharing — full transparency:
Subscription revenue is redistributed strictly equally among all eligible providers each month via Stripe Connect, after a 15% platform fee covering infrastructure costs. Redistribution history is public:
obsidian-neural.com/public.html
No authentication required. No data is ever deleted.
Provider eligibility:
- Uptime score ≥ 80% (based on random unpredictable pings)
- At least 1 real job processed during the month
🚀 We're looking for our first 10 GPU providers.
If you have a GPU and want to earn a share of the monthly revenue while supporting an open-source project: → Provider kit
- Download VST3 from Releases
- Get your API key from obsidian-neural.com
- Load the VST in your DAW → Settings → Enter Server URL + API key
📘 Getting Started
🎚️ First Step
🎨 Draw-to-Audio
🎛️ Bank Management
Pricing:
| Plan | Price | Credits/month |
|---|---|---|
| Free | — | 20 samples |
| Base | €7.99/month | 150 samples |
| Starter | €11.99/month | 300 samples |
| Pro | €14.99/month | 500 samples |
Best for privacy, customization, and unlimited generations.
- Get Stability AI access
- Follow INSTALLATION.md
- Run
python server_interface.py→ point the plugin tohttp://localhost:8000
Windows: install-win.bat · macOS: ./install-mac.sh · Linux: ./install-lnx.sh
Runs completely offline. No GPU required, 16GB+ RAM.
- Get Stability AI access
- Download models from innermost47/stable-audio-open-small-tflite
- Copy to
%APPDATA%\OBSIDIAN-Neural\stable-audio\→ choose "Local Model" in plugin
Note: Fixed 10s generation, higher RAM usage, some timing limitations.
| Feature | Cloud API | Self-Hosted | Local Models |
|---|---|---|---|
| Setup | ⭐ Easy | ⭐⭐⭐ Advanced | ⭐⭐ Moderate |
| Hardware | None | GPU + CUDA/Metal | 16GB+ RAM |
| Quality | ⭐⭐⭐ Best | ⭐⭐ Good | ⭐ Basic |
| Variable duration | ✅ Up to 30s | ✅ Yes | ❌ Fixed 10s |
| Cost | Subscription | Free after setup | Free |
| Internet | Required | Not required | Not required |
Featured in 8 countries and 6 languages:
- 🇺🇸 Synthtopia — "Brings AI-Generated Samples To Your DAW"
- 🇨🇳 MIDIFAN — Leading Chinese music tech publication
- 🇳🇱 Rekkerd — "Real-time AI music generation VST3"
- 🇫🇷 Audiofanzine — Major French music tech publication
- 🇪🇸 FutureMusic — Spanish coverage
- 🇰🇷 S1 Forum — Korean music production community
- 🇯🇵 DTM Plugin Sale — Japanese music production community
- 🇺🇸 Bedroom Producers Blog — "FREE AI-powered jam partner"
"Too many AI projects focus on the things AI can save you from doing rather than how AI can help you get better at what you do."
— James Nugent, Bedroom Producers Blog
We hit 200 stars! 🌟 Huge thanks to the entire community.
The winner of the 1-year Starter Pack (€143.88 value) was randomly drawn from all stargazers: congratulations @mxmilkiib!
| Platform | Install path |
|---|---|
| Windows | C:\Program Files\Common Files\VST3\ |
| macOS VST3 | ~/Library/Audio/Plug-Ins/VST3/ |
| macOS AU | ~/Library/Audio/Plug-Ins/Components/ |
| Linux | ~/.vst3/ |
- Installation Guide
- Video Tutorial — French + English subtitles
- Online Documentation
- Press Coverage
📧 Contact · 💬 GitHub Discussions · 🌐 obsidian-neural.com
🥁 BeatCrafter — Intelligent MIDI drum pattern generator VST3
🎛️ Randomizer — Generative music studio
🎵 YouTube — Original compositions (electronic, ambient, metal, experimental)
🆓 GNU Affero General Public License v3.0 — Stability AI Community License for the AI model.
Developed by InnerMost47 (Anthony Charretier)
Special thanks to A.D.[167] for the original draw-to-audio concept, Moteka for the testimonial and early adoption, Stability AI, and all beta testers.
Made with 🎵 in France
