Real-time Valorant coaching overlay powered by Gemini. Watches your screen during a match, gives live voice and text advice, records the session, extracts key moment clips, and syncs everything to a web dashboard.
spectAI/
├── spectai-overlay-v2/ # PyQt6 overlay — runs during your game
├── Vision-Model/ # Gemini vision + OCR pipeline
├── Dashboard/spectAI/ # React/Vite web dashboard
├── sessions/ # Auto-generated session recordings & clips
└── requirements.txt # All Python dependencies
- Python 3.11+
- Node.js 18+
- Valorant running on Windows
- Supabase project (for dashboard sync)
Create a .env file in the project root:
GEMINI_API_KEY=
ELEVENLABS_API_KEY=
HENRIK_API_KEY=
SUPABASE_URL=
SUPABASE_SERVICE_KEY=
SUPABASE_ANON_KEY=
SUPABASE_PROFILE_ID=
RIOT_API_KEY=pip install -r requirements.txt
playwright installcd Dashboard/spectAI
npm installCreate a public bucket named match-clips in your Supabase project:
Supabase Dashboard → Storage → New bucket → name: match-clips → Public: on
Launch Valorant first, then:
cd spectai-overlay-v2
python main.pyThe overlay starts in standby — no recording or AI until you press F8.
| Key | Action |
|---|---|
F8 |
Start / stop session (recording + AI coaching) |
F9 |
Push-to-talk — hold to ask the AI a question |
F10 |
Toggle AI voice mute |
F12 |
Quit SpectAI |
ALT + M |
Cycle through demo plays |
ALT + H |
Hide / show minimap overlay |
ALT + X |
Toggle speech bubble overlay |
cd Dashboard/spectAI
npm run devOpen http://localhost:5173 in your browser.
- Press F8 to start — begins screen recording and AI coaching
- AI gives real-time coaching via overlay text and voice
- Press F8 again to stop — session is saved to
sessions/<match_id>/ - Analysis pipeline runs automatically:
- Finds key moments →
key_moments.json - Extracts video clips →
clip_1/throughclip_5/ - Uploads clips to Supabase Storage
- Inserts a new row into
match_datawith clip URLs and coaching notes
- Finds key moments →
- Open the dashboard to review clips and AI feedback
If you have a session with clips already extracted but not yet uploaded:
cd spectai-overlay-v2
python upload_clips.py- Match history with expandable clip review
- Per-clip coaching feedback panel
- AI-generated player summary (strengths / weaknesses)
- Tracker.gg stat sync
- Training plan generation