Skip to content

roadmap.md

Latest commit

 

History

History
237 lines (180 loc) · 8.17 KB

roadmap.md

File metadata and controls

237 lines (180 loc) · 8.17 KB

mojovoice Project Roadmap

Status: Active Development | Last Updated: 2026-02-08

Project roadmap and feature planning.

Version Roadmap

v0.6.0 - UI/UX Overhaul & Platform Maturity

Theme: Production-ready design and best-in-class user experience

Priority Order:

  1. UI/UX Overhaul (Critical - Comprehensive redesign)

    Settings Reorganization:

    • Collapse advanced options (4 primary settings visible vs current 12)
    • Voice Recognition section (model + language dropdown)
    • Recording section (duration + audio device selection)
    • Behavior section (text options)
    • Appearance section (UI scale)
    • Advanced section (collapsed by default for power users)
    • Smart restart notifications (only show when daemon settings change)

    Models Tab (New Navigation Item):

    • Separate model management from Settings
    • Model library browser with download/cancel/delete
    • Visual download progress with real-time stats
    • Model cards with speed/quality indicators

    Dev Coffee Brand Consistency:

    • Apply bracket frames, subtle glows, technical grids across all components
    • Typography hierarchy using JetBrains Mono for headings
    • Purposeful emerald green accents (not decorative)
    • Near-black backgrounds with glass layers
    • Scanlines and retro-tech aesthetic refinements

    Dashboard Improvements:

    • Better visual hierarchy and information density
    • Enhanced status visualization
    • Recording feedback with waveform visualization
    • Clearer daemon/GPU/model status display

    Model Setup Wizard (First-Run Experience):

    • Hardware detection (GPU type, VRAM, disk space)
    • Model recommendation based on capabilities
    • Guided download with progress feedback
    • Test recording/transcription in wizard
    • Auto-configure environment

    Reference Documents:

    • docs/project/2026-01-14-settings-ui-redesign.md
    • docs/context/dev-coffee-brand-style-guide.md

  2. Linux Automated Testing (Critical - Reliability)

    • CI testing on Ubuntu, Fedora, Arch
    • Validate CUDA installation across distros
    • Test audio capture on different audio servers (PipeWire, PulseAudio, ALSA)
    • Integration tests for Waybar/systemd
    • Distro-specific packaging validation

  3. Cross-Platform Dictation Indicator (High - Consistency)

    • System tray icon with recording state (all platforms)
    • Visual feedback during recording (waveform/pulsing indicator)
    • Status overlay (OSD) support
    • Consistent behavior on Linux/macOS/Windows

  4. CUDA Multi-Distro Support (High - Remove friction)

    • Automated CUDA detection and fallback
    • Distro-specific CUDA path handling
    • Better error messages for missing CUDA
    • Documentation for each major distro

  5. macOS Unsigned .dmg (Medium - Good enough for v0.6.0)

    • Improve .dmg with better first-run experience
    • Handle permission prompts gracefully
    • Clear instructions for Accessibility/Microphone permissions

Deferred to v0.6.1+:

  • macOS signed installer with code signing/notarization (requires $99/year Apple Developer account)

v0.7.0 - Platform Expansion

Theme: AMD and Windows support

Features:

  • AMD ROCm Support (Critical)

    • ROCm GPU acceleration for AMD graphics cards
    • Automated ROCm detection and fallback
    • Testing on popular AMD GPUs

  • Windows Support (Critical)

    • Full Windows 10/11 support
    • Audio capture validation
    • Text injection via Windows APIs
    • Installer with proper permissions

  • Polybar Integration (High)

    • Status module for X11/i3/bspwm users
    • Mirrors Waybar integration pattern

v0.8.0 - Intelligence & Performance

Theme: Smarter transcription and quality-of-life improvements

Features:

  • Context-Aware Vocabulary

    • Detect active file type (.rs, .py, .ts, .md)
    • Auto-bias toward relevant technical terms
    • Project-specific vocabulary learning

  • Speculative Decoding

    • Draft model integration for 30-50% speedup
    • Research correct Whisper API implementation

  • Mojo-Audio 0.1.2

    • Performance improvements
    • Better audio preprocessing

  • Dynamic Prompt Biasing

    • UI for adjusting biasing prompts
    • Preset vocabularies (Rust dev, Python dev, DevOps, etc.)
    • Learn from correction history

  • General QoL Improvements

    • Voice commands ("undo last", "new line", "format code")
    • Better error recovery
    • Performance optimizations

Completed Features (v0.5.x)

  • ✅ Desktop App CI Builds - Automated Tauri builds for Linux/macOS
  • ✅ Documentation Overhaul - Complete README rewrite
  • ✅ Performance Benchmarking - mojovoice benchmark command
  • ✅ Community Infrastructure - LICENSE, CONTRIBUTING.md
  • ✅ Model Management UI - Download, switch, delete with visual meters
  • ✅ Transcription History - Searchable persistent history
  • ✅ Audio Device Selection - UI-based device picker
  • ✅ Daemon Subcommands - up, down, restart, status, logs

Future Features Under Consideration

Features requiring further research, evaluation, or design decisions before commitment.

Speaker Diarization Integration

Status: Under Review Research Date: 2026-01-23

Overview: Microsoft's VibeVoice-ASR offers unified ASR + speaker diarization + timestamps, but has significant integration barriers:

VibeVoice-ASR Analysis:

  • ✅ Unified pipeline (ASR + diarization + timestamps in one pass)
  • ✅ Long-form support (up to 60 minutes continuous)
  • ✅ Hotword support for domain-specific vocabulary
  • ❌ 9B parameters (vs Whisper Large-v3 Turbo ~1.5B) - significantly higher resource requirements
  • ❌ Python-only implementation (no Rust bindings)
  • ❌ Incompatible architecture (Qwen2.5-based vs Whisper encoder-decoder)
  • ❌ Uses 7.5 Hz speech tokenizers (incompatible with mojo-audio mel spectrogram pipeline)
  • ❌ Limited language support (English/Chinese only vs Whisper's 99 languages)

Alternative Approaches:

  • pyannote-audio (Rust-friendly via ONNX export)
  • WhisperX (adds diarization to Whisper, similar integration pattern)
  • Keep Whisper for ASR, add separate lightweight diarization module

Decision Needed:

  • Is speaker diarization a core use case for dev-voice?
  • What performance/resource trade-offs are acceptable?
  • Should this wait for better Rust-native solutions?

Reference: https://huggingface.co/microsoft/VibeVoice-ASR

Text-to-Speech Integration (Qwen3 TTS)

Status: Under Review Research Date: 2026-01-23

Overview: Qwen's newly released Qwen3-TTS models offer text-to-speech capabilities that could enable voice feedback and accessibility features in mojovoice.

Qwen3-TTS Model Family:

  • ✅ Multiple model sizes (0.6B, 1.7B parameters)
  • ✅ Custom voice support (CustomVoice variants)
  • ✅ Voice design capabilities (VoiceDesign variants)
  • ✅ 12Hz sampling rate
  • ✅ Natural-sounding speech synthesis
  • ⚠️ Very new release (hours old - maturity unknown)
  • ⚠️ Python-first implementation (Rust integration TBD)
  • ⚠️ Resource requirements unclear for real-time use

Potential Use Cases:

  • Audio feedback (read back transcriptions for verification)
  • Accessibility features (voice confirmation of commands)
  • Testing infrastructure (generate synthetic audio for STT pipeline testing)
  • Future voice assistant features (bidirectional voice interaction)

Decision Needed:

  • Is TTS a core feature for dev-voice or scope creep?
  • What's the priority vs existing STT improvements?
  • Should we wait for more mature Rust bindings?
  • Which variant/size is optimal for developer workflow use?

Reference: https://huggingface.co/collections/Qwen/qwen3-tts

Ideas & Future Exploration

Advanced Features (Post v0.8.0)

  • Custom wake word for hands-free mode
  • Multi-language support testing (Spanish, French, German, etc.)
  • Noise cancellation preprocessing (DeepFilterNet)
  • IDE plugins (VSCode, Neovim, JetBrains)
  • Mobile companion app (trigger from phone)

Community Infrastructure (Ongoing)

  • LICENSE file (MIT) ✅
  • CONTRIBUTING.md ✅
  • CODE_OF_CONDUCT.md
  • Issue templates (bug report, feature request)
  • PR template with checklist
  • Discord/Matrix community channel

🤖 Generated with Claude Code