Added ASS rendering fontsdir config with test, to make fonts work in different environments

Author: beveradb

lyrics_transcriber/output/video.py

@@ -31,6 +31,7 @@ def __init__(
         self.output_dir = output_dir
         self.cache_dir = cache_dir
         self.video_resolution = video_resolution
+        self.styles = styles
         self.logger = logger or logging.getLogger(__name__)
 
         # Get background settings from styles, with defaults
@@ -199,6 +200,21 @@ def _resize_background_image(self, input_path: str) -> str:
             self.logger.error(f"Failed to resize background image: {e.output}")
             raise
 
+    def _build_ass_filter(self, ass_path: str) -> str:
+        """Build ASS filter with font directory support."""
+        ass_filter = f"ass={ass_path}"
+        
+        # Get font path from styles configuration
+        karaoke_styles = self.styles.get("karaoke", {})
+        font_path = karaoke_styles.get("font_path")
+        
+        if font_path and os.path.isfile(font_path):
+            font_dir = os.path.dirname(font_path)
+            ass_filter += f":fontsdir={font_dir}"
+            self.logger.info(f"Returning ASS filter with fonts dir: {ass_filter}")
+        
+        return ass_filter
+
     def _build_ffmpeg_command(self, ass_path: str, audio_path: str, output_path: str) -> List[str]:
         """Build FFmpeg command for video generation with optimized settings."""
         width, height = self.video_resolution
@@ -230,11 +246,10 @@ def _build_ffmpeg_command(self, ass_path: str, audio_path: str, output_path: str
                 "-i", f"color=c={self.background_color}:s={width}x{height}:r=30"
             ])
 
-        # Add audio input and subtitle overlay
         cmd.extend([
             "-i", audio_path,
             "-c:a", "flac",  # Re-encode audio as FLAC
-            "-vf", f"ass={ass_path}",  # Add subtitles
+            "-vf", self._build_ass_filter(ass_path),  # Add subtitles with font directories
             "-c:v", self._get_video_codec(),
             # Video quality settings
             "-preset", "fast",  # Better compression efficiency
@@ -284,12 +299,11 @@ def _build_preview_ffmpeg_command(self, ass_path: str, audio_path: str, output_p
                 "-i", f"color=c={self.background_color}:s={width}x{height}:r=30"
             ])
 
-        # Add audio input and subtitle overlay
         cmd.extend([
             "-i", audio_path,
             "-c:a", "aac",  # Use AAC for audio
             "-b:a", "128k",  # Audio bitrate
-            "-vf", f"ass={ass_path}",  # Add subtitles
+            "-vf", self._build_ass_filter(ass_path),  # Add subtitles with font directories
             "-c:v", "libx264",  # Use H.264 codec
             "-profile:v", "baseline",  # Most compatible H.264 profile
             "-level", "3.0",  # Compatibility level

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "lyrics-transcriber"
-version = "0.57.1"
+version = "0.58.0"
 description = "Automatically create synchronised lyrics files in ASS and MidiCo LRC formats with word-level timestamps, using Whisper and lyrics from Genius and Spotify"
 authors = ["Andrew Beveridge <[email protected]>"]
 license = "MIT"

tests/unit/output/test_video.py

@@ -4,6 +4,7 @@
 import json
 import subprocess
 from unittest.mock import patch, Mock, call
+import logging
 
 from lyrics_transcriber.output.video import VideoGenerator
 
@@ -203,3 +204,94 @@ def test_invalid_video_resolution():
     styles = {"karaoke": {"background_color": "black"}}
     with pytest.raises(ValueError):
         VideoGenerator(output_dir="test", cache_dir="test", video_resolution=(0, 0), styles=styles)
+
+
+def test_build_ass_filter_no_font_path(video_generator):
+    """Test _build_ass_filter with no font path configured."""
+    ass_path = "test.ass"
+    result = video_generator._build_ass_filter(ass_path)
+    assert result == "ass=test.ass"
+
+
+def test_build_ass_filter_with_valid_font_path(video_generator, tmp_path):
+    """Test _build_ass_filter with valid font path."""
+    # Create a test font file
+    font_dir = tmp_path / "fonts"
+    font_dir.mkdir()
+    font_file = font_dir / "test.ttf"
+    font_file.touch()
+    
+    # Update styles to include font path
+    video_generator.styles["karaoke"]["font_path"] = str(font_file)
+    
+    ass_path = "test.ass"
+    result = video_generator._build_ass_filter(ass_path)
+    expected = f"ass=test.ass:fontsdir={str(font_dir)}"
+    assert result == expected
+
+
+def test_build_ass_filter_with_invalid_font_path(video_generator):
+    """Test _build_ass_filter with invalid font path (file doesn't exist)."""
+    # Set non-existent font path
+    video_generator.styles["karaoke"]["font_path"] = "/nonexistent/font.ttf"
+    
+    ass_path = "test.ass"
+    result = video_generator._build_ass_filter(ass_path)
+    assert result == "ass=test.ass"
+
+
+def test_build_ass_filter_with_empty_font_path(video_generator):
+    """Test _build_ass_filter with empty font path."""
+    # Set empty font path
+    video_generator.styles["karaoke"]["font_path"] = ""
+    
+    ass_path = "test.ass"
+    result = video_generator._build_ass_filter(ass_path)
+    assert result == "ass=test.ass"
+
+
+def test_build_ass_filter_with_none_font_path(video_generator):
+    """Test _build_ass_filter with None font path."""
+    # Set None font path
+    video_generator.styles["karaoke"]["font_path"] = None
+    
+    ass_path = "test.ass"
+    result = video_generator._build_ass_filter(ass_path)
+    assert result == "ass=test.ass"
+
+
+def test_build_ass_filter_no_karaoke_section(tmp_path):
+    """Test _build_ass_filter with no karaoke section in styles."""
+    styles = {}  # No karaoke section
+    generator = VideoGenerator(
+        output_dir=str(tmp_path / "output"),
+        cache_dir=str(tmp_path / "cache"),
+        video_resolution=(1920, 1080),
+        styles=styles,
+    )
+    
+    ass_path = "test.ass"
+    result = generator._build_ass_filter(ass_path)
+    assert result == "ass=test.ass"
+
+
+def test_build_ass_filter_logging(video_generator, tmp_path, caplog):
+    """Test _build_ass_filter logging behavior."""
+    # Set log level to INFO to capture the logs
+    caplog.set_level(logging.INFO)
+    
+    # Create a test font file
+    font_dir = tmp_path / "fonts"
+    font_dir.mkdir()
+    font_file = font_dir / "test.ttf"
+    font_file.touch()
+    
+    # Update styles to include font path
+    video_generator.styles["karaoke"]["font_path"] = str(font_file)
+    
+    ass_path = "test.ass"
+    result = video_generator._build_ass_filter(ass_path)
+    
+    # Check that info log was generated
+    assert "Returning ASS filter with fonts dir:" in caplog.text
+    assert str(font_dir) in caplog.text