Your Scenarios
Bienvenue sur Unlimited PocketAI
Tap a card to begin
New here?
Nouveau scénario
L'IA fera référence au lore, aux personnages et à la construction du monde de cette œuvre.
Enables a "Generate Image" button in chat. Uses Miniapps AI (credits required). Requires disclaimer accepted.
Uses platform AI image models (costs credits) instead of free Pollinations.
Paramètres du scénario
Scenario Details
No greeting set.
Danger Zone
Sélection du modèle
Credit Limiter Beta
Limit Credits per Message
Reduces context size (conversation history, journal, memory) to stay within budget. Tier 1 content (system rules, characters, D&D state) is never trimmed.
When active: World state, memory, and chapter limits are reduced. Older conversation rounds are dropped first. Smart Compression synergizes when both are enabled.
Credit Usage
No AI messages yet
Costs are per-request from the platform's creditsCharged. Includes input + output tokens. Free models show no cost. Tap a row to jump to that message.
Cost Display
Show estimated cost on the Send button
Show actual cost badges on AI responses
Type de scénario
L'IA fera référence au lore, aux personnages et à la construction du monde de cette œuvre.
Relationship Evolution
Track Character Relationships
Automatically tracks sentiment and trust between characters as the story progresses. AI will reference relationship history in responses.
Personnages du scénario
0 / 5Persona
Définissez qui vous êtes dans cette histoire. L'IA fera référence au nom, à l'apparence et à la personnalité de votre personnage.
Aucun
Image Generation
Adds image generation buttons to chat
Miniapps AI models cost credits per image. Higher quality, more reliable.
AI analyzes the scene and creates an optimized image prompt before generating
Default uses free RolePlay V1. Switch to a paid model for higher quality prompts (costs credits per image).
Save Pollinations images to CDN permanently (prevents expired links)
When off, Pollinations images use temporary URLs that may expire. When on, images are uploaded to CDN for permanent storage.
Provide your own images instead of AI generation
When enabled, clicking the Image button shows options: Embed URL, Upload Image, View Image Prompt, and Recommended Providers.
Cost Display
Show estimated cost on the Image button
Show actual cost badges on image generations
Known Characters Tracker Experimental
AI automatically tracks NPCs you meet in the story
⚠️ Experimental feature — results may vary. Piggybacks on chapter summaries, so no extra AI credits needed. NPCs are extracted when summaries are generated.
No NPCs tracked yet. They'll appear here as you meet characters in the story.
Export / Backup
Download this scenario as a JSON file including all messages, settings, characters, journal, and story memory. Re-import it anytime from New → Import → PocketAI Export.
Password-protect your export file
⚠️ Remember this password — it cannot be recovered!
Appearance
Customize the app's color theme.
All generated images start blurred. Tap "Unblur" to reveal them. Applies across all modes.
When enabled, images in Standard chat, PulsePlay, D&D, Afterlife, and galleries will be blurred until you tap the Unblur button on each image.
Privacy & Security
Storage usage is healthy
Export all data including scenarios, personas, and settings
Last backup: Never
Store data only on this device (no cloud sync)
⚠️ When enabled, your scenarios will NOT sync across devices. Data is stored in session storage and cleared when you close the tab.
Delete all scenarios when closing the app
🔒 Maximum privacy. All scenarios and messages will be permanently deleted when you close this tab. This cannot be undone.
Data synced to cloud
Your scenarios are backed up and available on other devices
Session Statistics
Track your roleplay activity and achievements.
Quality metrics for AI responses (coherence, consistency, relevance)
Informations sur le scénario
Narration
Automatically read AI responses aloud after generation
Uses your device's built-in speech engine. Tap "Narrate" on any AI message to read it aloud.
Lower = deeper voice, higher = lighter voice. Effect depends on your device's TTS engine.
Highlights words as they're spoken (visual follow-along)
⚠️ Limited compatibility. Requires a TTS engine that reports word boundaries. Google TTS works best. Samsung Voice Engine, Brave browser, and some other engines may not support this feature — the narration will still play, but words won't highlight.
Platform Narration Credits
Higher quality voices via Miniapps AI (costs credits per narration)
⚠️ Costs credits. Platform TTS uses Miniapps AI voices. Each narration costs credits (varies by length). Browser TTS remains free and is the default.
🔒 Disabled in offline mode. Platform TTS requires internet connection to Miniapps AI. When Ollama is active, only browser TTS is available.
Tap to browse voices with search, previews, and details.
Memory Context Limits
Control how much story history the AI sees. More context = better memory but higher cost.
Recent complete message exchanges sent to AI in full detail.
Older chapter summaries sent as long-term memory.
Regenerate embeddings for all chapters. Use if semantic retrieval seems inaccurate or after manual chapter edits.
Smart Compression
AI-Powered Context Compression
Uses AI to intelligently compress long conversations while preserving key story beats, character developments, and important decisions. Reduces context size for better performance and lower costs.
Context Caching
Cache AI Context
Caches the built message context for faster responses. Automatically rebuilds when you edit scenario details, characters, or settings.
Smart Auto-Mode
Automatically disables caching for short conversations or small contexts where rebuilding is faster than cache lookup.
Presets
Save and load configuration profiles. Built-in presets cover common use cases; create custom presets from your current settings.
Built-in Presets
Your Presets
Lore Books Manager
Lore Book System
Keyword-triggered context injection with hierarchical entries. Enable to manage and use lore books.
Lore Semantic Retrieval
Semantic retrieval for lore book activation (AI-powered)
Lore Book Chaining
Automatic injection of referenced lore entries
Context Optimization
💰 Credit-Saving Optimizations
These features reduce context size and save credits. Enabled by default.
System Prompt Optimization
Compact system prompts for free models
Narrative Tone Optimization
Compact tone directives on free models
NPC Description Optimization
Optimized NPC description format (30% reduction)
Anchor Message Detection
Preserve foundational context (character intros, world-building)
Dynamic Relevance Windows
Attention-weighted message retention based on narrative importance
Semantic Memory (Hash-based)
Hash-based semantic memory (no AI cost)
Character Isolation
Separate context windows per character for multi-character scenarios
Fact Verification
Automatic contradiction detection and flagging
⚡ Advanced Features (Cost Credits)
These features enhance quality but increase credit usage. Disabled by default.
Hierarchical Memory System
5-layer progressive memory with AI compression (~1 credit/layer)
AI-Powered Compression
Compress journal, middle sections (~1 credit each)
Semantic Memory (AI-powered)
Embedding-based semantic memory retrieval (AI-powered)
Narrative Beat Injection
Structural guidance for consistent pacing
Style Envelopes
Per-scene style adaptation (intense, intimate, horror, etc.)
Sensory State Tracking
Environmental state tracking for sensory continuity
Speech Pattern Tracking
Character voice fingerprinting and speech pattern tracking
Anti-Refusal Conditioning
Model-specific jailbreak frames for stubborn models
Content-Aware Routing
Content-type-aware detail routing (violent, intimate, horror, action)
Estimated Cost Impact
À propos de cette application
Bienvenue sur Unlimited PocketAI — une application de narration par IA conçue pour la liberté créative. Contrairement aux chatbots grand public qui refusent, réécrivent ou censurent votre contenu, cette application vous donne un contrôle total sur les thèmes, le ton et la direction. Votre histoire va là où vous le souhaitez. Ce qui fait la différence : • Aucun filtre de contenu ni censure — écrivez n'importe quel genre, n'importe quel thème • Modèles d'IA non censurés conçus pour la fiction créative sans restriction • Importez des cartes de personnages depuis chub.ai, botbooru et d'autres plateformes • Plusieurs styles de réponse, perspectives et niveaux de détail • Mémoire complète de l'histoire avec résumés de chapitres et suivi des PNJ Tout le contenu est entièrement fictif. Vous êtes seul responsable de ce que vous créez. En utilisant cette application, vous attestez que vous avez 18 ans ou plus, ou que vous avez l'âge légal dans votre juridiction selon vos lois locales, et que vous acceptez cette clause de non-responsabilité.
Refine Image
Create New
New Scenario
Quick Start
Pick a preset and start your story in seconds. Each creates a ready-to-play scenario with an immersive opening.
Importer
Gestionnaire d'histoire
Récapitulatif de l'histoire
Générez un résumé de votre histoire jusqu'à présent.
Chapitres de l'histoire
Journal de l'histoire
Épinglez les faits clés, les lieux et les points de l'intrigue dont l'IA doit toujours se souvenir.
Aucune entrée de journal pour le moment. Ajoutez des notes pour aider l'IA à se souvenir des détails importants de l'histoire.
Nouvelle entrée
L'IA compresse les entrées détaillées pour économiser le budget de contexte (coûte 1 crédit)
Version compressée :
Wizards
Choose a guided wizard to craft your story. Each wizard asks questions and AI creates a personalized scenario.
Histoire d'origine
Origin Story AI Model
Choose the AI model for your Origin Story.
Constructeur de monde
World Builder AI Model
Choose the AI model for your World Builder.
Et si ?
What If? AI Model
Choose the AI model for your What If? story. Sorted by cost (lowest first).
Your Story AI Model
Choose the AI model for your Your Story wizard.
Personas
Aucun persona pour le moment
Créez un persona pour définir qui vous êtes dans vos histoires. L'IA connaîtra le nom, l'apparence et la personnalité de votre personnage.
Votre histoire
Format de réponse
Choose how the AI structures its replies.
Niveau de détail
Control how long and detailed responses are.
Ton narratif
Set the emotional atmosphere of the narrative.
AI Prompt
Describe your story idea and AI will craft a complete, ready-to-play scenario with rich lore, characters, and an immersive opening.
Sequential 4-pass generation for richer, more detailed scenarios. Each section is crafted individually with full context from previous passes.
Each pass builds on the previous for maximum coherence and detail.
AI Models
Tap a model to select it. Sorted by cost (lowest first).
Each gold coin shown is the estimated cost per message sent to the AI. The number represents Miniapps credits (e.g., 1 = 1 credit per request). Costs vary by model — Free models cost nothing, while premium models cost more. Longer conversations may use slightly more credits due to context tokens.
Image Models
Tap a model to select it. Sorted by cost (lowest first).
Each gold coin is the estimated cost per image generated. Higher quality models cost more. ComfyUI local generation is free.
Some image models have content filters that may block certain prompts. If an image fails with a content filter error, try a different model or adjust your prompt.
Snap Composer
Platform Voices
Create New
Create
Tools
New Scenario
Quick Start
Wizards
Standard
Interactive
Importer
Search Scenarios
Start typing to search your scenarios
🚀 Developer Console
⚠️ Developer Tools
These tools are for advanced users. Export the source code, build a standalone desktop app, or connect to a local Ollama instance for offline AI. Use at your own risk.
1A · Export Source Code
Download all app files as a ZIP archive. Use this to backup the source code or host it yourself.
1B · Export Desktop App (Electron)
Download a complete Electron project. After extracting, run npm install then npm run build:win to create an .exe. Requires Node.js and Ollama installed locally.
1C · Export AppImage (Linux / Steam Deck)
Download a Linux-optimized Electron project with an automated build.sh script. Builds an .AppImage compatible with Steam Deck. Requires Node.js and Ollama installed locally.
1D · Setup Guide
Complete instructions for browser, Electron, AppImage, Ollama, and ComfyUI setup.
1E · Local Ollama AI Experimental
⚪ Inactive — using platform API
Danger: Only connect to trusted local networks. Do not expose this IP to the public internet.
Pull the model in terminal first, then refresh. Tap to select.
Some local models take longer. Increase if you get timeout errors. Range: 30s–10min.
Pollinations image generation timeout. Increase if images time out on slow connections. Range: 30s–5min.
Store data only on this device (no cloud sync)
Delete all data when closing the app
Manage locally saved Pollinations images (Electron mode only)
Images older than this will be deleted when you click "Cleanup Now"
⚠️ When enabled, all AI requests route to your local Ollama instance. Requires Ollama installed and running. If Ollama is unreachable, you will see a retry prompt.
Testing Tools
Preview the empty state (no scenarios) screen without deleting your data. Toggle off to restore your scenarios.
This is a safe preview mode — your scenarios are hidden, not deleted. Toggle again to restore them.
Downloads the app logo SVG for manual upload to miniapp settings.
Opens the full architecture documentation — file structure, features, technical debt, and maintenance notes.
Copy the full adversarial audit to clipboard in two parts — paste each into your AI reviewer separately to avoid truncation.
Unified test dashboard with 12 test categories, coverage metrics, and one-click test execution. Consolidates all test harnesses into a single interface.
⚪ Disabled — minimal logging
Enable verbose console logging for debugging. Off by default to reduce noise. Useful for diagnosing navigation, ViewStack, and MenuShell issues.
1G · Storage Encryption Eject Only
AES-256 encryption for exported app data
⚠️ Eject mode only. Encryption protects your data when running as a standalone Electron app. You'll need to enter your password each time you open the app. Not available in miniapps.ai cloud mode.
Password is stored securely. If you forget it, you'll lose access to encrypted data.
1F · ComfyUI Image Generation Experimental
⚪ Inactive — using Pollinations
Requires ComfyUI running locally. Download from github.com/comfyanonymous/ComfyUI. Start with python main.py. Default URL is http://127.0.0.1:8188.
Danger: Only connect to trusted local networks. Do not expose this IP to the public internet.
⚠️ When enabled, image generation uses your local ComfyUI instance instead of Pollinations. Supports any checkpoint model loaded in ComfyUI. Images are generated locally — no internet needed for generation. ComfyUI works independently of Ollama — you can use platform chat with local images, or both together for full offline.
ComfyUI generation timeout. Increase for slow GPUs or complex checkpoints. Range: 30s–10min.
dnd.wizard.title
D&D 5E AI Model
Choose the AI model for your D&D 5E Dungeon Master.
🎲 Dice Roller
Character Sheet
⚔️ Combat Tracker
Galerie
0Filtrer les images
Roleplay model detected. This model may prioritize creativity over factual accuracy. Switch to DeepSeek or Gemini for reliable answers.
PocketAI