Merge pull request #210 from modelscope/codex/fix-subtitle-color-135

fix: preserve selected subtitle colors

Author: LauraGPT

funclip/subtitle_renderer.py

@@ -0,0 +1,59 @@
+"""Pillow-backed subtitle text rendering for MoviePy 1.x."""
+
+from pathlib import Path
+from numbers import Real
+
+import numpy as np
+from moviepy.editor import ImageClip
+from PIL import Image, ImageColor, ImageDraw, ImageFont
+
+
+DEFAULT_FONT_PATH = (
+    Path(__file__).resolve().parents[1] / "font" / "STHeitiMedium.ttc"
+)
+
+
+def make_text_clip(
+    text,
+    font_path=DEFAULT_FONT_PATH,
+    font_size=32,
+    color="white",
+):
+    """Render transparent subtitle text without ImageMagick."""
+    if isinstance(font_size, bool) or not isinstance(font_size, Real):
+        raise TypeError("font_size must be a number")
+    if font_size <= 0 or not float(font_size).is_integer():
+        raise ValueError("font_size must be a positive integer")
+    font_size = int(font_size)
+
+    font_path = Path(font_path)
+    if not font_path.is_file():
+        raise FileNotFoundError(f"subtitle font not found: {font_path}")
+
+    try:
+        fill = ImageColor.getrgb(str(color))[:3]
+    except ValueError as error:
+        raise ValueError(f"unsupported subtitle color: {color!r}") from error
+
+    text = str(text)
+    font = ImageFont.truetype(str(font_path), font_size)
+    spacing = max(1, font_size // 5)
+    probe = Image.new("RGBA", (1, 1), (0, 0, 0, 0))
+    draw = ImageDraw.Draw(probe)
+    left, top, right, bottom = draw.multiline_textbbox(
+        (0, 0), text, font=font, spacing=spacing
+    )
+    padding = max(2, font_size // 12)
+    width = max(1, right - left + 2 * padding)
+    height = max(1, bottom - top + 2 * padding)
+
+    image = Image.new("RGBA", (width, height), (0, 0, 0, 0))
+    draw = ImageDraw.Draw(image)
+    draw.multiline_text(
+        (padding - left, padding - top),
+        text,
+        font=font,
+        fill=(*fill, 255),
+        spacing=spacing,
+    )
+    return ImageClip(np.asarray(image), transparent=True)

funclip/videoclipper.py

@@ -14,9 +14,13 @@
 import soundfile as sf
 from moviepy.editor import *
 import moviepy.editor as mpy
-from moviepy.video.tools.subtitles import SubtitlesClip, TextClip
+from moviepy.video.tools.subtitles import SubtitlesClip
 from moviepy.editor import VideoFileClip, concatenate_videoclips
 from moviepy.video.compositing.CompositeVideoClip import CompositeVideoClip
+try:
+    from .subtitle_renderer import make_text_clip
+except ImportError:
+    from subtitle_renderer import make_text_clip
 from utils.subtitle_utils import generate_srt, generate_srt_clip, str2list
 from utils.argparse_tools import ArgumentParser, get_commandline_args
 from utils.trans_utils import pre_proc, proc, write_state, load_state, proc_spk, convert_pcm_to_float
@@ -346,7 +350,9 @@ def video_clip(self,
             start_end_info = "from {} to {}".format(start, end)
             clip_srt += srt_clip
             if add_sub:
-                generator = lambda txt: TextClip(txt, font='./font/STHeitiMedium.ttc', fontsize=font_size, color=font_color)
+                generator = lambda txt: make_text_clip(
+                    txt, font_size=font_size, color=font_color
+                )
                 subtitles = SubtitlesClip(subs, generator)
                 video_clip = CompositeVideoClip([video_clip, subtitles.set_pos(('center','bottom'))])
             concate_clip = [video_clip]
@@ -365,7 +371,9 @@ def video_clip(self,
                 start_end_info += ", from {} to {}".format(str(start)[:5], str(end)[:5])
                 clip_srt += srt_clip
                 if add_sub:
-                    generator = lambda txt: TextClip(txt, font='./font/STHeitiMedium.ttc', fontsize=font_size, color=font_color)
+                    generator = lambda txt: make_text_clip(
+                        txt, font_size=font_size, color=font_color
+                    )
                     subtitles = SubtitlesClip(chi_subs, generator)
                     _video_clip = CompositeVideoClip([_video_clip, subtitles.set_pos(('center','bottom'))])
                     # _video_clip.write_videofile("debug.mp4", audio_codec="aac")

requirements.txt

@@ -5,6 +5,7 @@ funasr>=1.4.9
 transformers>=4.32.0,<5.0
 huggingface_hub>=0.19.3,<1.0
 moviepy==1.0.3
+pillow
 numpy==1.26.4
 gradio>=4.31.3,<5.0
 starlette<1.0

tests/test_subtitle_renderer.py

@@ -0,0 +1,70 @@
+from pathlib import Path
+
+import numpy as np
+from moviepy.editor import ColorClip, CompositeVideoClip
+from moviepy.video.tools.subtitles import SubtitlesClip
+
+from funclip.subtitle_renderer import make_text_clip
+
+
+ROOT = Path(__file__).resolve().parents[1]
+FONT_PATH = ROOT / "font" / "STHeitiMedium.ttc"
+
+
+def _foreground_rgb(color):
+    clip = make_text_clip("字幕 Test", FONT_PATH, 48, color)
+    frame = clip.get_frame(0)
+    mask = clip.mask.get_frame(0)
+
+    assert mask.min() == 0
+    assert mask.max() > 0.9
+    return frame[mask > 0.5]
+
+
+def test_subtitle_renderer_preserves_selected_colors():
+    red = _foreground_rgb("red").mean(axis=0)
+    green = _foreground_rgb("green").mean(axis=0)
+    black = _foreground_rgb("black").mean(axis=0)
+    white = _foreground_rgb("white").mean(axis=0)
+
+    assert red[0] > 200 and red[1] < 40 and red[2] < 40
+    assert green[1] > 100 and green[0] < 40 and green[2] < 40
+    assert black.max() < 10
+    assert white.min() > 240
+
+
+def test_subtitle_renderer_rejects_invalid_font_size():
+    assert make_text_clip("subtitle", FONT_PATH, 48.0, "white").size[0] > 0
+
+    for value in (0, -1, 48.5, "48", True):
+        try:
+            make_text_clip("subtitle", FONT_PATH, value, "white")
+        except (TypeError, ValueError):
+            continue
+        raise AssertionError(f"font size {value!r} should be rejected")
+
+
+def test_selected_color_reaches_composited_video_frame():
+    background = ColorClip((320, 120), color=(20, 20, 20), duration=1)
+    subtitles = SubtitlesClip(
+        [((0, 1), "字幕")],
+        lambda text: make_text_clip(text, FONT_PATH, 48, "red"),
+    ).set_pos(("center", "bottom"))
+    frame = CompositeVideoClip([background, subtitles]).get_frame(0.5)
+
+    red_pixels = (
+        (frame[:, :, 0] > 180)
+        & (frame[:, :, 1] < 60)
+        & (frame[:, :, 2] < 60)
+    )
+    assert red_pixels.sum() > 100
+
+
+def test_pillow_is_an_explicit_runtime_dependency():
+    requirements = {
+        line.strip().lower()
+        for line in (ROOT / "requirements.txt").read_text(encoding="utf-8").splitlines()
+        if line.strip() and not line.lstrip().startswith("#")
+    }
+
+    assert "pillow" in requirements