VideoTools/docs/ROADMAP.md

VideoTools Roadmap

This roadmap is intentionally lightweight. It captures the next few high-priority goals without locking the project into a rigid plan.

How We Use This

• The roadmap is a short list, not a full backlog.
• Items can move between buckets as priorities change.
• We update this at the start of each dev cycle.

Current State

• dev21 focused on stylistic filters and enhancement module planning.
• Filters module now includes decade-based authentic effects (8mm, 16mm, B&W Film, Silent Film, VHS, Webcam).
• Player stability identified as critical blocker for enhancement development.

Now (dev22 focus)

• Rock-solid video player implementation - CRITICAL PRIORITY

  • Fix fundamental A/V synchronization issues
  • Implement frame-accurate seeking without restarts
  • Add hardware acceleration (CUDA/VA-API/VideoToolbox)
  • Integrate chapter detection from Author module
  • Build foundation for frame extraction and keyframing
  • Eliminate seeking glitches and desync issues

• Enhancement module foundation - DEPENDS ON PLAYER

  • Unified Filters + Upscale workflow
  • Content-type aware processing (general/anime/film)
  • AI model management system (extensible for future models)
  • Multi-pass processing pipeline
  • Before/after preview system
  • Real-time enhancement feedback

Next (dev23+)

• Enhancement module completion - DEPENDS ON PLAYER

  • Open-source AI model integration (BasicVSR, RIFE, RealCUGan)
  • Model registry system for easy addition of new models
  • Content-aware model selection
  • Advanced restoration (SVFR, SeedVR2, diffusion-based)
  • Quality-aware enhancement strategies

• Trim module with timeline interface - DEPENDS ON PLAYER

  • Frame-accurate trimming and cutting
  • Manual chapter support with keyframing
  • Visual timeline with chapter markers
  • Preview-based trimming with exact frame selection
  • Import chapter detection from Author module

• Professional workflow integration

  • Seamless module communication (Player ↔ Enhancement ↔ Trim)
  • Batch enhancement processing through queue
  • Cross-platform frame extraction
  • Hardware-accelerated enhancement pipeline

Later

• Advanced AI features

  • AI-powered scene detection
  • Intelligent upscaling model selection
  • Temporal consistency algorithms
  • Custom model training framework
  • Cloud processing options

• Module expansion

  • Audio enhancement and restoration
  • Subtitle processing and burning
  • Multi-track management
  • Advanced metadata editing

Versioning Note

We keep continuous dev numbering. After v0.1.1 release, the next dev tag becomes v0.1.1-dev22 (or whatever the next number is).

Technical Debt and Architecture

Player Module Critical Issues Identified

The current video player has fundamental architectural problems preventing stable playback:

1. Separate A/V Processes - No synchronization, guaranteed drift
2. Command-Line Interface Limitations - VLC/MPV controllers use basic CLI, not proper IPC
3. Frame-Accurate Seeking - Seeking restarts processes with full re-decoding
4. No Frame Extraction - Critical for enhancement and chapter functionality
5. Poor Buffer Management - Small audio buffers cause stuttering
6. No Hardware Acceleration - Software decoding causes high CPU usage

Proposed Go-Based Solution

Unified FFmpeg Player Architecture:

• Single FFmpeg process with multiplexed A/V output
• Proper PTS-based synchronization with drift correction
• Frame buffer pooling and memory management
• Hardware acceleration through FFmpeg's native support
• Frame extraction via pipe without restarts

Key Implementation Strategies:

• Ring buffers for audio/video to eliminate stuttering
• Master clock reference for A/V sync
• Adaptive frame timing with drift correction
• Zero-copy frame operations where possible
• Hardware backend detection and utilization

This player enhancement is the foundation requirement for all advanced features including enhancement module and all other features that depend on reliable video playback.