Iterate on cross-platform support

.github/workflows/tests-macos.yml

@@ -25,10 +25,7 @@ jobs:
           python-version: '3.11'
 
       - name: Install uv
-        run: |
-          curl -LsSf https://astral.sh/uv/install.sh | sh
-          echo "$HOME/.cargo/bin" >> $GITHUB_PATH
-        shell: bash
+        uses: astral-sh/setup-uv@v3
 
       - name: Configure Git for private repos
         run: |

.github/workflows/tests-ubuntu.yml

@@ -24,11 +24,31 @@ jobs:
         with:
           python-version: '3.11'
 
-      - name: Install uv
+      - name: Install Qt system dependencies
         run: |
-          curl -LsSf https://astral.sh/uv/install.sh | sh
-          echo "$HOME/.cargo/bin" >> $GITHUB_PATH
-        shell: bash
+          sudo apt-get update
+          sudo apt-get install -y \
+            libegl1 \
+            libgl1 \
+            libglib2.0-0 \
+            libdbus-1-3 \
+            libxkbcommon0 \
+            libxcb-icccm4 \
+            libxcb-image0 \
+            libxcb-keysyms1 \
+            libxcb-randr0 \
+            libxcb-render-util0 \
+            libxcb-shape0 \
+            libxcb-sync1 \
+            libxcb-xfixes0 \
+            libxcb-xinerama0 \
+            libxcb-cursor0 \
+            libfontconfig1 \
+            libxkbcommon-x11-0 \
+            xvfb
+
+      - name: Install uv
+        uses: astral-sh/setup-uv@v3
 
       - name: Configure Git for private repos
         run: |
@@ -40,7 +60,7 @@ jobs:
 
       - name: Run tests
         run: |
-          uv run invoke run-tests
+          xvfb-run -a uv run invoke run-tests
 
       - name: Run linting
         run: |

.github/workflows/tests-windows.yml

@@ -25,10 +25,7 @@ jobs:
           python-version: '3.11'
 
       - name: Install uv
-        run: |
-          curl -LsSf https://astral.sh/uv/install.sh | sh
-          echo "$HOME/.cargo/bin" >> $GITHUB_PATH
-        shell: bash
+        uses: astral-sh/setup-uv@v3
 
       - name: Configure Git for private repos
         run: |

.gitignore

@@ -20,6 +20,8 @@ ENV/
 build/
 dist/
 downloads/
+installer/windows/output/
+tools/
 .eggs/
 .eggs/**
 *.egg-info/

.pre-commit-config.yaml

@@ -3,7 +3,7 @@ repos:
     hooks:
       - id: shredguard-check
         name: shredguard check
-        entry: shredguard check
+        entry: uv run --only-group dev shredguard check
         language: system
         types: [text]
 

installer/linux/VoxKit.desktop

@@ -0,0 +1,8 @@
+[Desktop Entry]
+Name=VoxKit
+Exec=VoxKit
+Icon=voxkit
+Type=Application
+Categories=Science;Audio;Education;
+Comment=AI/ML Speech Pathology Analysis
+Terminal=false

installer/windows/VoxKit.iss

@@ -0,0 +1,42 @@
+#define AppName "VoxKit"
+#define AppVersion "0.4.1"
+#define AppPublisher "Brain Behavior Analytics Lab"
+#define AppURL "https://github.com/BrainBehaviorAnalyticsLab/voxkit-desktop"
+#define AppExeName "VoxKit.exe"
+
+[Setup]
+AppId={{A3F2C1D4-8B7E-4F6A-9D2E-1C5B3A7E8F90}
+AppName={#AppName}
+AppVersion={#AppVersion}
+AppPublisher={#AppPublisher}
+AppPublisherURL={#AppURL}
+AppSupportURL={#AppURL}
+AppUpdatesURL={#AppURL}
+DefaultDirName={autopf}\{#AppName}
+DefaultGroupName={#AppName}
+DisableProgramGroupPage=yes
+OutputDir=output
+OutputBaseFilename=VoxKit-setup-windowsOS
+SetupIconFile=..\..\assets\voxkit.ico
+Compression=lzma
+SolidCompression=yes
+WizardStyle=modern
+PrivilegesRequired=lowest
+ArchitecturesAllowed=x64compatible
+ArchitecturesInstallIn64BitMode=x64compatible
+
+[Languages]
+Name: "english"; MessagesFile: "compiler:Default.isl"
+
+[Tasks]
+Name: "desktopicon"; Description: "{cm:CreateDesktopIcon}"; GroupDescription: "{cm:AdditionalIcons}"
+
+[Files]
+Source: "..\..\dist\{#AppExeName}"; DestDir: "{app}"; Flags: ignoreversion
+
+[Icons]
+Name: "{autoprograms}\{#AppName}"; Filename: "{app}\{#AppExeName}"; IconFilename: "{app}\{#AppExeName}"
+Name: "{autodesktop}\{#AppName}"; Filename: "{app}\{#AppExeName}"; IconFilename: "{app}\{#AppExeName}"; Tasks: desktopicon
+
+[Run]
+Filename: "{app}\{#AppExeName}"; Description: "{cm:LaunchProgram,{#StringChange(AppName, '&', '&&')}}"; Flags: nowait postinstall skipifsilent

main.py

@@ -1,8 +1,37 @@
 import sys
+import multiprocessing
+import os
+
+# CRITICAL: Must run before any heavy imports (voxkit, torch, PyQt6, etc.).
+# When PyTorch DataLoader workers spawn on Windows, each child re-launches the
+# frozen exe and re-imports this module. freeze_support() short-circuits the
+# child before it re-runs main() and tries to open another GUI window.
+if __name__ == "__main__":
+    multiprocessing.freeze_support()
+
+# In a --windowed PyInstaller build there is no console, so sys.stdout and
+# sys.stderr are None. Libraries like tqdm (used by transformers' Trainer)
+# crash with AttributeError: 'NoneType' object has no attribute 'write' when
+# they try to print progress. Redirect to devnull so writes silently no-op.
+if sys.stdout is None:
+    sys.stdout = open(os.devnull, "w", encoding="utf-8")
+if sys.stderr is None:
+    sys.stderr = open(os.devnull, "w", encoding="utf-8")
+
 import faulthandler
 import logging
-import os
-import multiprocessing
+
+# Windows: configure console and stdout/stderr for UTF-8 before any output.
+# Without this, rich's legacy renderer falls back to cp1252 and chokes on
+# Unicode characters (e.g., circled letters in pipeline config YAML).
+if sys.platform == 'win32':
+    try:
+        import ctypes
+        ctypes.windll.kernel32.SetConsoleOutputCP(65001)
+        sys.stdout.reconfigure(encoding='utf-8', errors='replace')
+        sys.stderr.reconfigure(encoding='utf-8', errors='replace')
+    except Exception:
+        pass
 
 # Apply patches for frozen (PyInstaller) environment BEFORE other imports
 if getattr(sys, 'frozen', False):
@@ -17,10 +46,6 @@
 logging.basicConfig(level=logging.INFO, format="%(asctime)s %(levelname)s %(name)s: %(message)s")
 log = logging.getLogger("voxkit.main")
 
-# CRITICAL: Must be at the top for frozen apps using multiprocessing
-if __name__ == "__main__":
-    multiprocessing.freeze_support()
-
 # Enable detailed crash reports
 faulthandler.enable()
 

pyproject.toml

@@ -16,8 +16,8 @@ dependencies = [
     "pyqt6>=6.9.1",
     "torch==2.8.0",
     "torchaudio==2.8.0",
-    "pypllrcomputer @ git+https://github.com/BrainBehaviorAnalyticsLab/PyPLLRComputer.git@48a0c934e75f73235ba5a002538b8588fd6697e7",
-    "wav2textgrid @ git+https://github.com/pkadambi/Wav2TextGrid.git@3862476185a7558d481f1ae1c9cbca2e48c825a0",
+    "pypllrcomputer",
+    "wav2textgrid",
     "datasets>=4.3.0",
     "accelerate>=1.11.0",
     "keyring>=25.6.0",
@@ -28,7 +28,8 @@ dependencies = [
     "pyyaml>=6.0.0",
     "rich>=14.2.0",
     "faster-whisper>=1.1.0",
-    "alignment-comparison-plots==0.1.1",
+    "alignment-comparison-plots",
+    "speechbrain>=1.0.3",
 ]
 
 [dependency-groups]
@@ -58,6 +59,12 @@ Homepage = "https://github.com/BrainBehaviorAnalyticsLab/voxkit-desktop"
 Repository = "https://github.com/BrainBehaviorAnalyticsLab/voxkit-desktop"
 Issues = "https://github.com/BrainBehaviorAnalyticsLab/voxkit-desktop/issues"
 
+[tool.uv.sources]
+pypllrcomputer = { git = "https://github.com/pkadambi/PyPhonemePronunciationScorer", branch = "voxkit-windows-variant" }
+wav2textgrid = { git = "https://github.com/pkadambi/Wav2TextGrid", branch = "voxkit-windows-variant" }
+alignment-comparison-plots = { git = "https://github.com/WISCLab/alignment-comparison-plots", branch = "voxkit-windows-variant" }
+speechbrain = { git = "https://github.com/BeckettFrey/speechbrain.git", branch = "fix/windows-lazy-import-inspect-path" }
+
 [tool.ruff]
 line-length = 100
 target-version = "py310"
@@ -120,7 +127,9 @@ exclude = [
     "build.py",
     "example_startup_script.py",
     "test_imports.py",
-    "tests/"
+    "tests/",
+    "build/",
+    "dist/",
 ]
 
 

src/voxkit/analyzers/__init__.py

@@ -14,6 +14,14 @@
     Extracts speaker count and audio file counts per speaker directory.
     Includes a bar chart visualization for quick dataset overview.
 
+**ClipDurationStatisticsAnalyzer**
+    Reads audio metadata to compute total, average, min, and max clip duration
+    per speaker. Includes a bar chart visualization of total duration.
+
+**AudioFormatProfileAnalyzer**
+    Reads audio metadata to surface dominant sample rate, channel count, and
+    flags files that deviate from the speaker's dominant format.
+
 Output Structure
 ----------------
 Analyzer output is stored alongside dataset metadata::
@@ -35,7 +43,9 @@
 
 from typing import List
 
+from .audio_format_profile import AudioFormatProfileAnalyzer
 from .base import DatasetAnalyzer
+from .clip_duration_statistics import ClipDurationStatisticsAnalyzer
 from .default_analyzer import DefaultAnalyzer
 
 
@@ -69,9 +79,18 @@ def get_analyzer(self, analyzer_id: str) -> DatasetAnalyzer:
             raise ValueError(f"No analyzer with id: {analyzer_id}")
 
 
-default_analyzer_instance = DefaultAnalyzer()
+_default = DefaultAnalyzer()
+_duration = ClipDurationStatisticsAnalyzer()
+_format = AudioFormatProfileAnalyzer()
+
 # Singleton instance for unified export/interface
-ManageAnalyzers = AnalyzerManager({default_analyzer_instance.name: default_analyzer_instance})
+ManageAnalyzers = AnalyzerManager(
+    {
+        _default.name: _default,
+        _duration.name: _duration,
+        _format.name: _format,
+    }
+)
 
 __all__ = [
     "ManageAnalyzers",

src/voxkit/analyzers/audio_format_profile.py

@@ -0,0 +1,84 @@
+"""Audio Format Profile Analyzer.
+
+Reads audio metadata (no decode) to surface sample rate and channel count
+per speaker. Flags speakers with inconsistent formats across their files.
+
+Output Columns
+--------------
+- **speaker_id**: Name of the speaker subdirectory
+- **file_count**: Number of audio files scanned
+- **dominant_sample_rate_hz**: Most common sample rate across the speaker's files
+- **dominant_channels**: Most common channel count (1 = mono, 2 = stereo)
+- **inconsistent_files**: Number of files that differ from the dominant sample rate
+"""
+
+import logging
+import os
+from collections import Counter
+from pathlib import Path
+from typing import Any, Dict, List
+
+from .base import DatasetAnalyzer
+
+logger = logging.getLogger(__name__)
+
+
+class AudioFormatProfileAnalyzer(DatasetAnalyzer):
+    """Per-speaker audio format profile: sample rate, channels, and consistency."""
+
+    @property
+    def name(self) -> str:
+        return "Audio Format Profile"
+
+    @property
+    def description(self) -> str:
+        return "Sample rate, channel count, and format consistency per speaker"
+
+    def analyze(self, dataset_path: str) -> List[Dict[str, Any]]:
+        import torchaudio
+
+        results = []
+        audio_extensions = {".wav", ".flac", ".mp3", ".ogg", ".m4a"}
+
+        try:
+            for entry in os.scandir(dataset_path):
+                if not entry.is_dir():
+                    continue
+
+                sample_rates: List[int] = []
+                channels: List[int] = []
+                for f in os.scandir(entry.path):
+                    if not f.is_file() or Path(f.name).suffix.lower() not in audio_extensions:
+                        continue
+                    try:
+                        info = torchaudio.info(f.path)
+                        if info.sample_rate > 0:
+                            sample_rates.append(info.sample_rate)
+                            channels.append(info.num_channels)
+                        else:
+                            waveform, sr = torchaudio.load(f.path)
+                            sample_rates.append(sr)
+                            channels.append(waveform.shape[0])
+                    except Exception as e:
+                        logger.warning("Skipping %s: %s", f.path, e)
+
+                if not sample_rates:
+                    continue
+
+                dominant_sr = Counter(sample_rates).most_common(1)[0][0]
+                dominant_ch = Counter(channels).most_common(1)[0][0]
+                inconsistent = sum(1 for sr in sample_rates if sr != dominant_sr)
+
+                results.append(
+                    {
+                        "speaker_id": entry.name,
+                        "file_count": len(sample_rates),
+                        "dominant_sample_rate_hz": dominant_sr,
+                        "dominant_channels": dominant_ch,
+                        "inconsistent_files": inconsistent,
+                    }
+                )
+        except Exception as e:
+            print(f"Error analyzing dataset: {e}")
+
+        return results