Your Journey with Cleft
What Cleft Is (And Isn’t)
Input Method Comparison
What Cleft Is
Personal thought processor for your ideas and insights- Voice-first note-taking with intelligent formatting
- Privacy-focused with local Whisper transcription
- Complementary tool that works alongside your existing workflow
- Built for verbal thinkers who process ideas out loud
What Cleft Is Not
- ❌ Meeting transcription service (use your meeting platform’s built-in transcription)
- ❌ AI assistant replacement (like ChatGPT)
- ❌ Team collaboration tool
- ❌ Generic transcription service
- ❌ “Do everything” productivity app
Our One Thing: Transforming Verbal Thoughts Into Usable Notes
We do one thing exceptionally well: take your longwinded, messy verbal thoughts and transform them into organized, usable notes. Not transcription. Not AI assistance. Thought transformation. You talk. Cleft listens, understands, and formats your ideas exactly how you want them - from custom instructions to future expansions we’re planning. This is what you pay for: an amazing experience that turns voice into value.We Work WITH Your Tools, Not Against Them
Integration Philosophy:- Use Cleft IN Google Meet for your personal notes
- Use Cleft WITH Microsoft Teams for your follow-up thoughts
- Export TO Notion, Obsidian, or wherever you work
- We complement your workflow, we do not compete with it
- A replacement for meeting transcription (use your favorite dedicated tool like Google Meet transcription)
- An AI assistant competitor (ChatGPT does that better)
- A team collaboration tool (Slack handles that)
- Your personal thought processor
- A bridge between your verbal mind and your digital workflow
- Software that charges for experience, not storage
For Verbal Thinkers: Like having skilled scribes running behind you on your morning walk or while packing the kids into the car - capturing and shaping your ideas as they flow naturally.
The Philosophy Behind Cleft
Traditional note-taking forces your brain into rigid structures before ideas are fully formed. Cleft reverses this:Traditional Approach
Think → Structure → Write Forces premature organization and loses the natural flow of ideasCleft Approach
Think → Speak → AI Structures Preserves natural thought flow while creating organized, actionable outputWhy Voice-First Works
Your brain generates ideas faster than you can type or organize them. Voice capture eliminates the bottleneck between thought and capture, while AI handles the organization you would struggle with in real-time.The Neuroscience: Speaking engages different neural pathways than writing, often leading to more natural and complete expression of ideas. The act of speaking out loud also helps clarify and develop thoughts.
What We’re Opinionated About
We have strong opinions about how software should work:Voice Is First-Class
Voice is not an afterthought or accessibility feature - it’s how many people naturally process complex ideas. We built for verbal thinkers first.Privacy Matters
Your thoughts should not train someone else’s AI. Local Whisper processing means your audio never leaves your device.Custom Instructions Are Essential
Generic formatting doesn’t work for personal productivity. You need control over how your thoughts get shaped.Integration > Competition
We’re not trying to replace your entire workflow. We work WITH Google Meet, Teams, Notion, Obsidian - wherever you already work.What We’re Unopinionated About
We stay flexible where it matters:- Organization: Use folders, tags, or chaos - whatever works for you
- Export destinations: Send your notes anywhere you want
- Processing levels: Dial AI formatting up or down based on your needs
- Workflow integration: Fit Cleft into YOUR process, not ours
Data Philosophy: Your Thoughts, Your Control
Cleft’s Approach
Privacy First:- Audio processed locally with Whisper
- Only text sent to OpenAI for formatting
- You control all your data
- No audio leaves your device
- No training data harvesting
- Export anywhere, anytime
- No vendor lock-in
- Standard formats (text, markdown)
- Full ownership of your notes
Alternative Approaches
Cloud-First Services:- Audio uploaded to external servers
- May be used for model training
- Require trust in third-party data handling
- Often proprietary formats
- Better collaboration features
- More processing power
- Advanced AI capabilities
- Dependency on internet and services
Understanding Data Approaches
Local Processing (Cleft):- Your audio never leaves your device
- No risk of data breaches with your personal thoughts
- Works offline (for basic recording)
- You control your data completely
- Audio/text uploaded to external servers
- May be used to train AI models (varies by service)
- Better features but less privacy
- Depends on company’s data policies
- Privacy matters: Your personal thoughts should not train someone else’s AI
- Local Whisper: Audio transcription happens on your device
- Minimal cloud usage: Only formatted text goes to OpenAI for processing
- Cost savings: We do not monetize your data, so we can charge less
The Technical Architecture
Cleft uses a sophisticated multi-stage processing pipeline designed for privacy, speed, and accuracy:Stage 1: Voice Capture (On-Device)
Local Audio Processing- High-quality audio recording optimized for speech
- Noise reduction and audio enhancement
- Secure local storage with encryption
- No audio data ever transmitted off-device
Stage 2: Speech-to-Text (On-Device)
Whisper Model Integration- OpenAI’s state-of-the-art Whisper model runs locally
- Industry-leading accuracy for natural speech patterns
- Support for technical terminology and proper names
- Real-time transcription during recording
Privacy by Design: Your voice never leaves your device. Only the text transcript is sent for AI processing.
Stage 3: AI Intelligence (Cloud Processing)
Text-Only Processing- Only the transcript (not audio) sent to OpenAI’s models
- Advanced language models understand context and intent
- Custom instructions guide output formatting
- Structured output returned as organized notes
Stage 4: Note Generation
Intelligent Structuring- AI identifies key concepts, action items, and relationships
- Creates logical hierarchy and organization
- Preserves important context and nuance
- Formats for readability and actionability
How AI Understands Your Thoughts
Natural Language Processing
Cleft’s AI doesn’t just transcribe—it comprehends: Context Awareness- Understands implicit connections between ideas
- Recognizes when you’re brainstorming vs. planning vs. reflecting
- Identifies action items, decisions, and open questions
- Maintains thematic consistency across longer recordings
- Distinguishes between facts, opinions, and hypotheses
- Recognizes temporal relationships (“before we do X, we need Y”)
- Identifies priority levels and urgency indicators
- Understands conditional statements and dependencies
- Creates logical hierarchies from stream-of-consciousness input
- Groups related concepts automatically
- Generates appropriate headings and subheadings
- Maintains narrative flow while adding organization
Example: “I think we should probably maybe look into that new framework, but first we definitely need to finish the current sprint, and oh I forgot we also need to update the documentation” becomes a structured action plan with clear priorities and dependencies.
Custom Instructions: Personalizing Your AI Assistant
Tailoring Output to Your Needs
Custom Instructions allow you to shape how Cleft processes your voice notes, creating consistent output that matches your workflow and communication style. How It Works- Your instructions are combined with the transcript before AI processing
- The AI considers your preferences while maintaining its core understanding capabilities
- Instructions persist across all notes until you modify them
- Executive Summary Style: “Always include a brief executive summary at the top”
- Action-Oriented: “Focus on extracting actionable items and decisions”
- Technical Format: “Use technical terminology appropriate for software development”
- Meeting Structure: “Organize as: Key Points, Decisions, Action Items, Follow-ups”
The Processing Pipeline with Custom Instructions
Data Flow and Security Architecture
Privacy-First Processing Pipeline
Cleft’s architecture ensures your voice data never leaves your device while still providing powerful AI capabilities: Local Processing (100% Private)- Voice recording stored with AES-256 encryption
- Whisper model runs entirely on-device (no network required)
- Audio files remain locally accessible only to you
- Transcription happens in real-time during recording
- Only the text transcript is transmitted (never audio)
- HTTPS/TLS encryption for all network communications
- Processing through OpenAI’s API with enterprise-grade security
- No voice data stored on external servers
Performance Optimization
Intelligent Caching- Frequently used AI processing patterns cached locally
- Faster response times for similar note types
- Reduced API calls and improved offline experience
- Short notes processed immediately
- Longer recordings processed in optimized chunks
- Background processing doesn’t block user interface
- Compressed text transmission
- Retry logic for poor network conditions
- Offline queuing for complete network outages
🎨 Make Cleft Yours with Custom Instructions (Plus)
Transform how Cleft organizes your thoughts. With Custom Instructions, you control exactly how your notes are structured—perfect for your unique thinking style.
- 📝 Choose your preferred format (bullets, paragraphs, action items)
- 🎯 Add context about your work or thinking style
- 🔧 Create templates for different types of notes
- ✨ Get consistently formatted notes every time
Learn about Custom Instructions →
Upgrade to Plus for Custom Instructions →
Complete Workflow Engine
Multi-Modal Input Processing
Cleft handles various input types seamlessly: Voice Recording- Real-time waveform visualization during capture
- Automatic pause detection for natural speech breaks
- Background noise filtering and audio enhancement
- Support for multiple microphone configurations
- Direct text entry with same AI processing capabilities
- Markdown formatting preservation
- Mixed voice/text notes with unified processing
- Image recognition and OCR processing
- Document parsing for meeting preparation
- Context-aware file organization
Advanced AI Capabilities
Contextual Understanding- Cross-reference previous notes for consistency
- Maintain context across multi-part recordings
- Understand implicit references and pronouns
- Preserve speaker intent even with unclear speech
- Automatic heading hierarchy generation
- Bullet point organization based on content flow
- Table creation for structured data
- Code block detection and syntax highlighting
- Multi-language support with automatic detection
- Technical terminology recognition
- Proper noun and acronym handling
- Emotional tone and urgency detection
🔒 Privacy by Design
Your voice recordings never leave your device. Only the text transcript is sent to AI models for processing.
- ✅ Voice transcription happens 100% locally using Whisper
- ✅ Only text (transcript + your instructions) is sent to OpenAI
- ✅ Your actual voice audio stays completely private
- ✅ You get beautiful, organized notes without compromising privacy
Platform Integration and Scalability
Cross-Platform Synchronization
Real-Time Sync- Notes synchronized instantly across iOS, iPadOS, and macOS
- Conflict resolution for simultaneous edits
- Offline changes merged when connectivity returns
- Version history maintained for all edits
- iOS: Optimized for touch interface and mobile workflows
- macOS: Keyboard shortcuts and desktop productivity features
- Universal: Consistent experience across all Apple devices
Enterprise and Scale Considerations
Performance at Scale- Efficiently handles thousands of notes with instant search
- Intelligent indexing for fast content discovery
- Background processing optimized for battery life
- Memory management for large voice recording libraries
- RESTful API for third-party integrations
- Webhook support for automation workflows
- Export capabilities in multiple formats (Markdown, PDF, HTML)
- Import tools for migration from other note-taking systems
- Modular architecture allows for feature expansion
- AI model updates deployed seamlessly
- New voice recognition languages added dynamically
- Custom integration development supported through SDK
Technical Foundation: Built on modern Swift and SwiftUI frameworks with CoreData for local storage, CloudKit for synchronization, and advanced machine learning frameworks for on-device processing.