Dream GPT

Welcome to Dream GPT...

🌙 A 10-minute lucid dream simulation artwork

🔐 SecureAuth Integration Active
• API keys managed securely via Railway
• Realtime API (voice) + Standard API (visuals)
• No exposed credentials in code
• Invisible authentication (zero UI impact)

🎤 Speech Transcription: ENABLED
• Your voice is transcribed with Whisper-1
• Transcripts appear in chat automatically
• Saved to logs with AI responses

🧪 Test Mode Available
• Toggle in editor (⚙️ Edit Phases & Prompts)
• Disables stream and music to save resources
• Perfect for testing dialogue without visuals

⚙️ NEW: Fullscreen Editor
• Click "⚙️ Edit Phases & Prompts" button (top-left)
• "🎤 Start Now" button in editor - quick start recording!
• Edit phases.txt and visual.txt in one place
• Live visual anchor display - see current anchor & history
• Live Auto Go On indicator - shows question detection & pause time
• 🎭 Metaphorical Portrait Generator - analyzes conversation via GPT-4o
• Your edits are saved in browser (localStorage)
• Priority: Your edits > Files > Defaults
• Reset anytime to original files

⏱️ Smart Auto Go On Timer:
• If AI asks question (?) → waits 20 seconds
• If AI uses word "describe" → waits 20 seconds
• If AI makes statement → waits 4 seconds
• Automatically continues dialogue after pause
• Live indicator in editor shows detection

Quick Start:
1. Press "Connect" to begin the experience
2. Use "Speak" button for voice OR type answers in text field
3. Press "Send" to submit text answers
4. Answer questions when asked - the AI will wait for you
5. Make choices during the dream construction phase
6. Press "💾 Download Log" in editor to save conversation anytime
7. Press "Disconnect" when finished
8. Press "📺 Show Stream" or V key to view TouchDesigner visuals

🎤 Voice Transcription:
• All your spoken words are transcribed automatically
• Transcripts appear in chat messages
• Saved to conversation logs
• Powered by OpenAI Whisper-1 model

🎭 Performance Phases System:
• Uses ONLY phases.txt - no hardcoded scenarios
• Auto-loads from local phases.txt file on page load
• Phases auto-switch ON by default
• Each phase has: temperature, speed, duration, instructions
• NEW: Invisible Nudge - optional message sent to AI on phase switch
• NEW: Auto-Disconnect - if nudge contains "final message", disconnects after AI finishes
• Phases change automatically based on response count
• Current phase shown in UI with progress

📝 phases.txt Format:
===PHASE===
temperature: 0.9
speed: slow
duration: 3
nudge: ***Now begin the extraction phase***

Your phase instructions here...
Nudge (optional): Invisible message to AI on phase switch. Use *** or plain text.
Special: If nudge contains "final message" → auto-disconnect after AI finishes speaking

🎙️ Experience Structure:
• Controlled entirely by phases.txt
• Typically: Induction → Extraction → Dream → Ending
• Each phase can have different tone/temperature
• Auto-continue after 20 seconds of silence

🎨 Visual System:
• ✅ NEW FRAGMENT-BASED GENERATION
• AI text split into 8-word fragments in real-time
• UI slideshow: 3 seconds per fragment (stops at last)
• Generation trigger: Every fragment switch in UI
• Fragment 1/N → generate anchor → Fragment 2/N → generate anchor → ...
• Each fragment = one GPT-4o-mini call
• ✅ CONTEXT-AWARE: GPT-4o-mini sees last 5 generated anchors
• Prevents repetition, creates coherent visual sequence
• Anchors sent to TouchDesigner stream immediately
• 🚫 NO FACES: People only as silhouettes/figures

💡 Important:
• The AI will STOP and WAIT for your answers
• Your answers become "meaningful anchors" in the dream
• Be specific, concrete, and honest in your responses
• Use voice (Speak) or text (Type + Send) - both work!
• The Operator speaks VERY VERY SLOW
• Experience driven by phases.txt only

Note: This is an art experience, not therapy or meditation. The Oneiric Operator will guide you through an intimate, uncanny journey...