Update video2.py

video2.py

@@ -256,47 +256,68 @@ def audio_func(id, lines, lang):
     return asyncio.run(generate_tts(id, lines, lang))
 #-----------------------------
 #---------------------------------
-def video_func(id, lines,lang):
-    duration, audio_path = audio_func(id,lines,lang)
+import os
+import subprocess
+import shlex
+import time
+import math
+import numpy as np
+import cv2
+from moviepy.editor import VideoFileClip, AudioFileClip
+from moviepy.video.fx.speedx import speedx
+
+def video_func(id, lines, lang):
+    duration, audio_path = audio_func(id, lines, lang)
     if not duration or not audio_path:
         print("Failed to generate audio.")
         return None
-    #listf = lines.split("&&&")
-    #TEXT = listf[0].strip()
-    TEXT=lines[id]
+    TEXT = lines[id]
     print("-----------------------------------------------------------------------------")
     print(TEXT)
     SKIP_SPACES = False
    
-    FPS = 30 # Increased for smoother animation
-    ANIMATION_FRAMES_PER_CHAR = 3 # Number of sub-frames for pen movement per character
-    WIDTH, HEIGHT = 1280, 720 # Keep as is
+    FPS = 30  # Keep for smoothness, but can reduce to 24 if needed for speed
+    ANIMATION_FRAMES_PER_CHAR = 2  # Reduced from 3 for faster rendering (less frames per char)
+    WIDTH, HEIGHT = 1280, 720
     MARGIN_X, MARGIN_Y = 40, 60
-    LINE_SPACING = 8 # additional px between lines
+    LINE_SPACING = 8
     FONT = cv2.FONT_HERSHEY_SIMPLEX
-    FONT_SCALE = 1.5 # tweak for desired size
-    THICKNESS = 2
-    TEXT_COLOR = (0, 0, 0) # BGR
-    BG_COLOR = (255, 255, 255) # BGR
+    DEFAULT_FONT_SCALE = 1.5
+    HEADER_FONT_SCALE = 2.0  # Increased size for headers
+    DEFAULT_THICKNESS = 2
+    HEADER_THICKNESS = 3  # Bolder for headers
+    DEFAULT_TEXT_COLOR = (0, 0, 0)  # BGR Black
+    HEADER_TEXT_COLOR = (255, 0, 0)  # BGR Blue
+    BG_COLOR = (255, 255, 255)  # BGR White
     silent_video_name = f"silent_video{id}.mp4"
     silent_video_path = os.path.join(CLIPS_DIR, silent_video_name)
-    FFMPEG_PRESET = "ultrafast" # fastest encode
-    CRF = 23 # For faster encoding
+    FFMPEG_PRESET = "ultrafast"
+    CRF = 28  # Increased CRF for faster encoding (lower quality, but quicker)
     # Pen settings
-    PEN_COLOR = (0, 0, 255) # Red pen for visibility (BGR)
-    PEN_TIP_RADIUS = 5 # Size of pen tip circle
-    PEN_LENGTH = 20 # Length of pen line
-    PEN_THICKNESS = 2 # Thickness of pen line
-    PEN_BASE_ANGLE = 45 # Base angle of pen (degrees)
-    PEN_MOVEMENT_AMPLITUDE = 10 # How much the pen moves up/down (pixels)
+    PEN_COLOR = (0, 0, 255)  # Red pen (BGR)
+    PEN_TIP_RADIUS = 5
+    PEN_LENGTH = 20
+    PEN_THICKNESS = 2
+    PEN_BASE_ANGLE = 45
+    PEN_MOVEMENT_AMPLITUDE = 10
     # ===================================
    
-    # Helper: wrap text by pixel width using cv2.getTextSize
-    def wrap_text_cv(text, font, font_scale, thickness, max_width):
+    # Helper: wrap text by pixel width using cv2.getTextSize, now with per-line styles
+    def wrap_text_cv(text, font, default_font_scale, default_thickness, max_width):
         wrapped_lines = []
+        styles = []  # List of (is_header) for each wrapped line
         for para in text.splitlines():
+            is_header = para.strip().startswith("###")
+            if is_header:
+                para = para.strip()[3:].strip()  # Remove "### " or "###"
+                font_scale = HEADER_FONT_SCALE
+                thickness = HEADER_THICKNESS
+            else:
+                font_scale = default_font_scale
+                thickness = default_thickness
             if para == "":
-                wrapped_lines.append("") # preserve blank line
+                wrapped_lines.append("")
+                styles.append(False)  # Not header
                 continue
             words = para.split(" ")
             cur = ""
@@ -308,6 +329,7 @@ def video_func(id, lines,lang):
                 else:
                     if cur != "":
                         wrapped_lines.append(cur)
+                        styles.append(is_header)
                     (single_w, _), _ = cv2.getTextSize(w, font, font_scale, thickness)
                     if single_w > max_width:
                         chunk = ""
@@ -318,6 +340,7 @@ def video_func(id, lines,lang):
                                 chunk = cand2
                             else:
                                 wrapped_lines.append(chunk)
+                                styles.append(is_header)
                                 chunk = ch
                         if chunk:
                             cur = chunk
@@ -327,10 +350,12 @@ def video_func(id, lines,lang):
                         cur = w
             if cur != "":
                 wrapped_lines.append(cur)
-        return wrapped_lines
-    # Pre-wrap text
+                styles.append(is_header)
+        return wrapped_lines, styles
+   
+    # Pre-wrap text with styles
     text_area_width = WIDTH - 2 * MARGIN_X
-    wrapped_lines = wrap_text_cv(TEXT, FONT, FONT_SCALE, THICKNESS, text_area_width)
+    wrapped_lines, line_styles = wrap_text_cv(TEXT, FONT, DEFAULT_FONT_SCALE, DEFAULT_THICKNESS, text_area_width)
     full_text = "\n".join(wrapped_lines)
     if not full_text:
         full_text = ""
@@ -345,20 +370,23 @@ def video_func(id, lines,lang):
     if total_glyphs == 0:
         print("No text to animate.")
         return None
-    # Always render the minimal animation frames for the full text (no repeats or padding during rendering)
+    # Minimal frames
     min_frames = total_glyphs * ANIMATION_FRAMES_PER_CHAR
     print(f"Rendering {min_frames} minimal frames for full text animation.")
-    # Pre-calc line heights and y_positions
+    # Pre-calc line heights and y_positions with per-line styles
     line_heights = []
-    for line in wrapped_lines:
-        if line == "":
-            (w, h), baseline = cv2.getTextSize("Ay", FONT, FONT_SCALE, THICKNESS)
-        else:
-            (w, h), baseline = cv2.getTextSize(line, FONT, FONT_SCALE, THICKNESS)
-        line_heights.append(h + baseline + LINE_SPACING)
     y_positions = []
     y = MARGIN_Y
-    for lh in line_heights:
+    for i, line in enumerate(wrapped_lines):
+        is_header = line_styles[i]
+        font_scale = HEADER_FONT_SCALE if is_header else DEFAULT_FONT_SCALE
+        thickness = HEADER_THICKNESS if is_header else DEFAULT_THICKNESS
+        if line == "":
+            (w, h), baseline = cv2.getTextSize("Ay", FONT, font_scale, thickness)
+        else:
+            (w, h), baseline = cv2.getTextSize(line, FONT, font_scale, thickness)
+        lh = h + baseline + LINE_SPACING
+        line_heights.append(lh)
         y_positions.append(y)
         y += lh
     # Prepare ffmpeg
@@ -372,18 +400,22 @@ def video_func(id, lines,lang):
     print("FFMPEG CMD:", ffmpeg_cmd)
    
     proc = subprocess.Popen(shlex.split(ffmpeg_cmd), stdin=subprocess.PIPE, bufsize=10**8)
-    # Render function, modified: if pen_x <= 0, no pen
+    # Render function, now with per-line colors and styles
     def render_frame(visible_text, pen_x, pen_y, anim_offset):
         img = np.full((HEIGHT, WIDTH, 3), BG_COLOR, dtype=np.uint8)
         lines = visible_text.split("\n")
         for idx, line in enumerate(lines):
+            is_header = line_styles[idx]
+            font_scale = HEADER_FONT_SCALE if is_header else DEFAULT_FONT_SCALE
+            thickness = HEADER_THICKNESS if is_header else DEFAULT_THICKNESS
+            color = HEADER_TEXT_COLOR if is_header else DEFAULT_TEXT_COLOR
             x = MARGIN_X
             y = y_positions[idx]
-            (w, h), baseline = cv2.getTextSize(line, FONT, FONT_SCALE, THICKNESS)
+            (w, h), baseline = cv2.getTextSize(line, FONT, font_scale, thickness)
             y_draw = y + h
             if line != "":
-                cv2.putText(img, line, (x, y_draw), FONT, FONT_SCALE, TEXT_COLOR, THICKNESS, lineType=cv2.LINE_AA)
-        if pen_x > 0: # Only draw pen if pen_x > 0
+                cv2.putText(img, line, (x, y_draw), FONT, font_scale, color, thickness, lineType=cv2.LINE_AA)
+        if pen_x > 0:
             offset_y = int(PEN_MOVEMENT_AMPLITUDE * math.sin(anim_offset * math.pi))
             pen_tip_y = pen_y + offset_y
             angle_rad = math.radians(PEN_BASE_ANGLE)
@@ -404,7 +436,10 @@ def video_func(id, lines,lang):
             lines = visible_sub.split("\n")
             last_line = lines[-1]
             line_idx = len(lines) - 1
-            (w, h), baseline = cv2.getTextSize(last_line, FONT, FONT_SCALE, THICKNESS)
+            is_header = line_styles[line_idx]
+            font_scale = HEADER_FONT_SCALE if is_header else DEFAULT_FONT_SCALE
+            thickness = HEADER_THICKNESS if is_header else DEFAULT_THICKNESS
+            (w, h), baseline = cv2.getTextSize(last_line, FONT, font_scale, thickness)
             pen_x = MARGIN_X + w + 5
             pen_y = y_positions[line_idx] + h // 2
             last_pen_x = pen_x
@@ -414,7 +449,6 @@ def video_func(id, lines,lang):
                 proc.stdin.write(frame_img.tobytes())
                 frames_sent += 1
             prev_visible_sub = visible_sub
-    # No repeat or remaining frames added during rendering - full minimal animation only
     proc.stdin.close()
     proc.wait()
     elapsed = time.time() - t0
@@ -422,7 +456,7 @@ def video_func(id, lines,lang):
     if not os.path.exists(silent_video_path):
         print("Silent video generation failed.")
         return None
-    # Now combine with audio using MoviePy: always render full text animation, then adjust speed to match audio duration
+    # Combine with audio using MoviePy
     final_video_name = f"clip{id}.mp4"
     final_video_path = os.path.join(CLIPS_DIR, final_video_name)
     video_clip = VideoFileClip(silent_video_path)
@@ -433,15 +467,9 @@ def video_func(id, lines,lang):
         print(f"Adjusting video speed by factor: {speed_factor:.3f}")
         video_clip = video_clip.fx(speedx, speed_factor)
     final_clip = video_clip.set_audio(AudioFileClip(audio_path))
-    # Write final video
-    final_clip.write_videofile(final_video_path, codec='libx264', audio_codec='aac', preset='ultrafast', verbose=False, logger=None)
-    # Print the final video file name
+    # Write final video with faster settings
+    final_clip.write_videofile(final_video_path, codec='libx264', audio_codec='aac', preset='ultrafast', verbose=False, logger=None, threads=4)  # Added threads for multi-threading
     print(f"Final video saved at: {final_video_path}")
-    # For notebook display (comment out if not needed in HF Spaces)
-    # if os.path.exists(final_video_path):
-    # display(Video(final_video_path, embed=True, width=WIDTH, height=HEIGHT))
-    # Clean up silent video if not needed
+    # Clean up
     os.remove(silent_video_path)
-    return final_video_path
-
-    #video
+    return final_video_path