Update utils/render.py

utils/render.py

@@ -1,75 +1,98 @@
-import uuid
-from moviepy.editor import (
-    VideoFileClip,
-    TextClip,
-    CompositeVideoClip
-)
+import os, uuid
+# Force ImageMagick path for Linux
+os.environ["IMAGEMAGICK_BINARY"] = "/usr/bin/convert"
+
+from moviepy import VideoFileClip, TextClip, CompositeVideoClip
+from moviepy.video.fx import Resize
+from .logger import logger
 
 # ---------------------------------------------------
 # FONT CONFIG
 # ---------------------------------------------------
-FONT_PATH = "fonts/TikTok-Bold.ttf"
-
+FONT_PATH = os.path.abspath(os.path.join(os.path.dirname(__file__), "..", "fonts", "TikTok-Bold.ttf"))
 
 # ---------------------------------------------------
-# CREATE TIKTOK STYLE SUBTITLE
+# MAIN RENDER FUNCTION
 # ---------------------------------------------------
-def create_text_clip(word, start, end, video_w, video_h):
+def burn_subtitles(video_path, highlight_frames):
+    """
+    Renders TikTok-style captions. 
+    highlight_frames: List of dicts with 'full_frame_text', 'active_word_index', 'start', 'end'
+    """
+    original_cwd = os.getcwd()
+    os.chdir("/tmp") # Permission fix for HF Spaces
+    
+    try:
+        output_path = os.path.abspath(f"/tmp/render_{uuid.uuid4().hex}.mp4")
+        video = VideoFileClip(video_path)
+        w, h = int(video.w), int(video.h)
+        
+        # Verify font exists
+        current_font = FONT_PATH if os.path.exists(FONT_PATH) else "sans-serif"
+        
+        subtitle_clips = []
 
-    txt_clip = (
-        TextClip(
-            txt=word.upper(),
-            fontsize=72,
-            font=FONT_PATH,
-            color="white",
-            stroke_color="black",
-            stroke_width=6,
-            method="caption",   # REQUIRED
-            size=(int(video_w * 0.85), None)
-        )
-        .set_start(start)
-        .set_end(end)
-        .set_position(("center", int(video_h * 0.75)))
-    )
-
-    return txt_clip
+        for frame in highlight_frames:
+            words = frame["full_frame_text"]
+            active_idx = frame["active_word_index"]
+            duration = frame["end"] - frame["start"]
+            
+            if duration <= 0: continue
 
+            # Build Pango Markup for single-word yellow highlight
+            markup_parts = [
+                f"<span foreground='yellow'>{word}</span>" if i == active_idx else word 
+                for i, word in enumerate(words)
+            ]
+            pango_markup = " ".join(markup_parts)
 
-# ---------------------------------------------------
-# MAIN RENDER FUNCTION
-# ---------------------------------------------------
-def burn_subtitles(video_path, words):
+            # Create the TextClip with Size 30
+            txt_clip = (
+                TextClip(
+                    text=pango_markup,
+                    font=current_font,
+                    font_size=30,      # Set to 30 as requested
+                    color="white",
+                    stroke_color="black",
+                    stroke_width=1.0,   # Scaled down for size 30
+                    method="pango",     # Changed to pango for the highlight logic
+                    size=(int(w * 0.80), None),
+                    text_align="center"
+                )
+                .with_start(frame["start"])
+                .with_duration(duration)
+                .with_position(("center", int(h * 0.80))) # Lowered for minimalist look
+            )
 
-    output_path = f"/tmp/render_{uuid.uuid4().hex}.mp4"
+            # Rhythmic 'Pop' Effect
+            txt_clip = Resize(lambda t: 1.10 if t < 0.08 else 1.0).apply(txt_clip)
+            subtitle_clips.append(txt_clip)
 
-    video = VideoFileClip(video_path)
+        # Composite and Export
+        final = CompositeVideoClip([video, *subtitle_clips], size=(w, h))
 
-    subtitle_clips = []
+        # Explicit Temp Audio Config to prevent "Invalid Stream" errors
+        temp_audio = os.path.join("/tmp", f"audio_tmp_{uuid.uuid4().hex}.m4a")
 
-    for w in words:
-        subtitle_clips.append(
-            create_text_clip(
-                w["word"],
-                w["start"],
-                w["end"],
-                video.w,
-                video.h
-            )
+        final.write_videofile(
+            output_path,
+            codec="libx264",
+            audio_codec="aac",
+            fps=video.fps,
+            preset="ultrafast", # Faster for HF CPU
+            threads=4,
+            temp_audiofile=temp_audio,
+            remove_temp=True,
+            logger=None
         )
 
-    final = CompositeVideoClip([video, *subtitle_clips])
-
-    final.write_videofile(
-        output_path,
-        codec="libx264",
-        audio_codec="aac",
-        fps=video.fps,
-        preset="veryfast",
-        threads=2,
-        logger=None
-    )
+        video.close()
+        final.close()
 
-    video.close()
-    final.close()
+        return output_path
 
-    return output_path
+    except Exception as e:
+        logger.error(f"Render Engine Error: {e}")
+        raise e
+    finally:
+        os.chdir(original_cwd)