AI-powered local transcription system with cross-platform production builds.

Developed by Panagiotis Zois
MSc in Information & Communication Technology
Focus Areas: AI Integration • Business Process Automation • Systems Architecture • ERP-Oriented Thinking

Panos AI Transcriber is a fully local AI transcription desktop application designed to demonstrate how artificial intelligence can be integrated into structured, production-oriented workflows.

The project goes beyond a simple Whisper wrapper.
It implements:

• Deterministic build pipelines
• Cross-platform distribution
• Chunk-based processing orchestration
• Long-running job monitoring
• Enterprise-style execution control

Latest Production Release: v1.1.0 (Cross-Platform)
Available for macOS and Windows (x64).

This project was designed as a portfolio-grade demonstration of:

• AI model integration into operational workflows
• Process orchestration & job lifecycle management
• Desktop system engineering
• Reproducible build pipelines
• Cross-platform packaging discipline

It reflects how AI tools can be embedded inside real business environments rather than used as standalone scripts.

• Fully local speech-to-text processing (no cloud dependency)
• Automatic audio chunking for large files
• Real-time progress monitoring
• Estimated time calculation (ETA logic)
• Cancellation & lifecycle control
• Cross-platform standalone builds (macOS / Windows x64)

The system follows a layered architecture:

• PySide6 desktop interface
• User interaction control
• Job state visibility

• Whisper Python API integration
• FFmpeg-based segmentation
• Time-based performance estimation
• Controlled execution environment

• Chunk scheduling
• Job lifecycle management
• Error handling & safe termination
• Deterministic execution flow

This structure mirrors real-world enterprise patterns such as:

• Batch processing systems
• Orchestration engines
• Workflow automation pipelines
• Long-running background task monitoring

• Whisper invoked via Python API (not CLI subprocess)
• CPU-first execution model
• Fully local execution (no external API calls)
• Bundled ffmpeg / ffprobe binaries
• Deterministic macOS build script
• Windows CI-based reproducible build pipeline
• Semantic versioning for releases

Versioned cross-platform builds are available under Releases.

Current stable release:

Panos AI Transcriber – v1.1.0

Included assets:

• macOS standalone application
• Windows standalone application (x64)

No Python installation is required for end users.

• Python 3.11
• PySide6 (Qt GUI framework)
• OpenAI Whisper
• FFmpeg
• PyInstaller
• GitHub Actions (Windows CI pipeline)

Performance scales with hardware capability.

Tested successfully on:

• Apple Silicon macOS
• Legacy Intel i5 (4th Gen) systems (CPU-only execution)

This demonstrates portability across hardware tiers and predictable scaling behavior.

This project demonstrates competencies aligned with:

• Business Process Automation
• AI-enabled workflow integration
• System lifecycle management
• Enterprise-style orchestration design
• Operational tooling development
• Cross-platform deployment discipline

It reflects the intersection of:

AI + Systems Engineering + Automation + Business Workflow Design

• Multi-file batch processing
• Structured output formats (JSON / CSV)
• ERP / CRM integration hooks
• API-based version
• Performance benchmarking module
• Model selection options
• Linux build pipeline

MIT License