Update generate.py

generate.py

@@ -11,41 +11,58 @@ import upload
 
 WIDTH, HEIGHT = 1080, 1920
 FPS = 30
-DURATION_PER_LINE = 3
-PEXELS_API_KEY = os.environ.get('PEXELS_API_KEY', '')
+DURATION_PER_LINE = 4
+HF_TOKEN = os.environ.get('HF_TOKEN', '')
 
-def get_background_image(keyword):
+def generate_ai_image(prompt):
+    """Generate cinematic image using SDXL on HF"""
     try:
-        headers = {'Authorization': PEXELS_API_KEY}
-        r = requests.get(
-            f'https://api.pexels.com/v1/search?query={keyword}&per_page=5&orientation=portrait',
-            headers=headers, timeout=10
-        )
-        data = r.json()
-        if data.get('photos'):
-            photo_url = data['photos'][0]['src']['portrait']
-            img_response = requests.get(photo_url, timeout=15)
-            img = Image.open(BytesIO(img_response.content)).convert('RGB')
+        API_URL = "https://api-inference.huggingface.co/models/stabilityai/stable-diffusion-xl-base-1.0"
+        headers = {"Authorization": f"Bearer {HF_TOKEN}"}
+        payload = {
+            "inputs": f"cinematic shot, {prompt}, 9:16 vertical, high quality, dramatic lighting, photorealistic",
+            "parameters": {
+                "width": 576,
+                "height": 1024,
+                "num_inference_steps": 20
+            }
+        }
+        print(f"Generating AI image for: {prompt[:50]}")
+        response = requests.post(API_URL, headers=headers, json=payload, timeout=60)
+        if response.status_code == 200:
+            img = Image.open(BytesIO(response.content)).convert('RGB')
             img = img.resize((WIDTH, HEIGHT))
-            overlay = Image.new('RGB', (WIDTH, HEIGHT), (0, 0, 0))
-            img = Image.blend(img, overlay, 0.55)
             return img
+        else:
+            print(f"SDXL error: {response.status_code} - {response.text[:100]}")
+            return None
     except Exception as e:
-        print(f"Pexels error: {e}")
-    return None
+        print(f"AI image error: {e}")
+        return None
 
 def make_text_frame(text, bg_image=None):
     if bg_image:
         img = bg_image.copy()
     else:
         img = Image.new('RGB', (WIDTH, HEIGHT), color=(15, 15, 25))
+
+    # Dark overlay for text readability
+    overlay = Image.new('RGBA', (WIDTH, HEIGHT), (0, 0, 0, 160))
+    img = img.convert('RGBA')
+    img = Image.alpha_composite(img, overlay)
+    img = img.convert('RGB')
+
     draw = ImageDraw.Draw(img)
     try:
         font = ImageFont.truetype("/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf", 65)
     except:
         font = ImageFont.load_default()
-    draw.rectangle([60, 180, WIDTH-60, 193], fill=(255, 80, 80))
-    draw.rectangle([60, HEIGHT-193, WIDTH-60, HEIGHT-180], fill=(255, 80, 80))
+
+    # Accent lines
+    draw.rectangle([60, 180, WIDTH-60, 195], fill=(255, 80, 80))
+    draw.rectangle([60, HEIGHT-195, WIDTH-60, HEIGHT-180], fill=(255, 80, 80))
+
+    # Text
     lines = textwrap.wrap(text, width=22)
     y = HEIGHT // 2 - (len(lines) * 85) // 2
     for line in lines:
@@ -55,6 +72,7 @@ def make_text_frame(text, bg_image=None):
         draw.text((x+4, y+4), line, font=font, fill=(0, 0, 0))
         draw.text((x, y), line, font=font, fill=(255, 255, 255))
         y += 90
+
     return np.array(img)
 
 def generate_tts(text, output_path):
@@ -68,24 +86,35 @@ def generate_tts(text, output_path):
         return False
 
 def generate_video(script, title, description):
-    print("Starting video generation...")
+    print("Starting cinematic video generation...")
     print(f"Script received: {script[:100]}")
+
     sentences = re.split(r'[.!?\n]', script)
     sentences = [s.strip() for s in sentences if len(s.strip()) > 5]
+
     if not sentences:
         words = script.split()
         sentences = [' '.join(words[i:i+8]) for i in range(0, len(words), 8)]
-    sentences = sentences[:15]
+
+    sentences = sentences[:10]
     print(f"Total clips: {len(sentences)}")
-    keyword = title.replace('#shorts', '').replace('Facts', '').strip()
-    print(f"Fetching background for: {keyword}")
-    bg_image = get_background_image(keyword)
+
     clips = []
+
     for i, sentence in enumerate(sentences):
         print(f"Creating clip {i+1}/{len(sentences)}")
+
+        # Generate AI image for this sentence
+        bg_image = generate_ai_image(sentence)
+
+        # Generate TTS
         audio_path = f'/app/audio_{i}.mp3'
         has_audio = generate_tts(sentence, audio_path)
+
+        # Create frame
         frame = make_text_frame(sentence, bg_image)
+
+        # Create clip
         if has_audio and os.path.exists(audio_path):
             audio = AudioFileClip(audio_path)
             duration = max(audio.duration + 0.5, DURATION_PER_LINE)
@@ -93,13 +122,17 @@ def generate_video(script, title, description):
             clip = clip.set_audio(audio)
         else:
             clip = ImageClip(frame, duration=DURATION_PER_LINE)
-        clip = clip.fadein(0.3).fadeout(0.3)
+
+        clip = clip.fadein(0.5).fadeout(0.5)
         clips.append(clip)
+
     if not clips:
         print("No clips generated!")
         return False
+
     print("Combining clips...")
     final = concatenate_videoclips(clips, method="compose")
+
     output_path = '/app/video.mp4'
     print("Writing video...")
     final.write_videofile(
@@ -110,12 +143,14 @@ def generate_video(script, title, description):
         verbose=False,
         logger=None
     )
+
     for i in range(len(sentences)):
         try:
             os.remove(f'/app/audio_{i}.mp3')
         except:
             pass
-    print("Video generated successfully!")
+
+    print("Cinematic video generated successfully!")
     return True
 
 if __name__ == '__main__':