Update app.py

app.py

@@ -3,38 +3,72 @@ import tempfile
 from pathlib import Path
 import uuid
 import subprocess
+import requests
+import base64
+import math
 
 allowed_medias = [".png", ".jpg", ".jpeg", ".bmp", ".gif", ".tiff"]
-
-def generate_slideshow(images, duration, texts=None, y_pos=0.5, fade_duration=0.7, font_size=60):
+API_URL = "https://text.pollinations.ai/openai"
+
+def transcribe_audio(audio_path):
+    with open(audio_path, "rb") as f:
+        audio_data = base64.b64encode(f.read()).decode()
+    payload = {
+        "model": "openai-audio",
+        "messages": [{
+            "role": "user",
+            "content": [
+                {"type": "text", "text": "Transcribe this audio:"},
+                {"type": "input_audio", "input_audio": {"data": audio_data, "format": "wav"}}
+            ]
+        }]
+    }
+    response = requests.post(API_URL, json=payload)
+    response.raise_for_status()
+    result = response.json()
+    # API liefert Text in choices[0].message.content
+    text = result['choices'][0]['message']['content']
+    return text
+
+def generate_slideshow_with_audio(images, audio_file, duration_per_image=3, y_pos=0.5, fade_duration=0.7, font_size=60):
     if not images:
         return None, "❌ Keine Bilder ausgewählt"
-
-    y_pos = min(max(0.0, y_pos), 0.9)  # 0=oben, 1=unten, max 0.9
+    
+    y_pos = min(max(0.0, y_pos), 0.9)
     temp_dir = tempfile.mkdtemp()
     clips = []
 
-    if texts is None:
+    # Transkription, falls Audio vorhanden
+    if audio_file:
+        transcript = transcribe_audio(audio_file.name)
+        # Einfach in gleiche Länge wie Bilder aufteilen
+        words = transcript.split()
+        total_words = len(words)
+        segments_per_image = math.ceil(total_words / len(images))
+        texts = []
+        for i in range(len(images)):
+            start = i*segments_per_image
+            end = min((i+1)*segments_per_image, total_words)
+            texts.append(" ".join(words[start:end]))
+    else:
         texts = [""] * len(images)
 
     for i, img_path in enumerate(images):
         clip_path = Path(temp_dir) / f"clip_{i}.mp4"
         text = texts[i] if i < len(texts) else ""
 
-        # Scale + Pad für proportionalen Fit in 1280x720
         vf_filters = (
             "scale=w=1280:h=720:force_original_aspect_ratio=decrease,"
             "pad=1280:720:(ow-iw)/2:(oh-ih)/2:color=black,"
             "fps=25,format=yuv420p"
         )
 
-        # Text Overlay Filter
         if text:
             safe_text = text.replace(":", "\\:").replace("'", "\\'")
             drawtext_filter = (
                 f",drawtext=text='{safe_text}':fontcolor=white:fontsize={font_size}:borderw=2:"
                 f"x=(w-text_w)/2:y=(h-text_h)*{y_pos}:"
-                f"alpha='if(lt(t,{fade_duration}), t/{fade_duration}, if(lt(t,{duration}-{fade_duration}), 1, ({duration}-t)/{fade_duration}))'"
+                f"alpha='if(lt(t,{fade_duration}), t/{fade_duration}, if(lt(t,{duration_per_image}-{fade_duration}), 1, ({duration_per_image}-t)/{fade_duration}))'"
             )
             vf_filters += drawtext_filter
 
@@ -43,7 +77,7 @@ def generate_slideshow(images, duration, texts=None, y_pos=0.5, fade_duration=0.
             "-y",
             "-loop", "1",
             "-i", str(img_path),
-            "-t", str(duration),
+            "-t", str(duration_per_image),
             "-vf", vf_filters,
             str(clip_path)
         ]
@@ -54,7 +88,7 @@ def generate_slideshow(images, duration, texts=None, y_pos=0.5, fade_duration=0.
 
         clips.append(clip_path)
 
-    # Alle Clips zu einem Video zusammenfügen
+    # Clips zusammenfügen
     filelist_path = Path(temp_dir) / "filelist.txt"
     with open(filelist_path, "w") as f:
         for clip in clips:
@@ -76,32 +110,44 @@ def generate_slideshow(images, duration, texts=None, y_pos=0.5, fade_duration=0.
     except subprocess.CalledProcessError as e:
         return None, f"❌ FFmpeg Concat Fehler:\n{e.stderr}"
 
-    if output_file.exists():
-        return str(output_file), f"✅ Slideshow erstellt ({len(images)} Bilder, Text fixiert, Bildproportionen erhalten)"
-    else:
-        return None, "❌ Video konnte nicht erstellt werden"
+    # Audio hinzufügen, falls vorhanden
+    if audio_file:
+        final_output = Path(temp_dir) / f"slideshow_audio_{uuid.uuid4().hex}.mp4"
+        cmd_audio = [
+            "ffmpeg",
+            "-y",
+            "-i", str(output_file),
+            "-i", audio_file.name,
+            "-c:v", "copy",
+            "-c:a", "aac",
+            "-shortest",
+            str(final_output)
+        ]
+        try:
+            subprocess.run(cmd_audio, check=True, capture_output=True, text=True)
+            return str(final_output), "✅ Slideshow mit Audio und automatischen Untertiteln erstellt"
+        except subprocess.CalledProcessError as e:
+            return None, f"❌ FFmpeg Audio Merge Fehler:\n{e.stderr}"
+
+    return str(output_file), "✅ Slideshow erstellt (ohne Audio)"
 
 # Gradio UI
 with gr.Blocks() as demo:
-    gr.Markdown("# Slideshow mit Text-Overlay & schwarzen Rändern")
+    gr.Markdown("# Slideshow mit Audio & automatischen Untertiteln")
 
     img_input = gr.Files(label="Bilder auswählen (mehrere)", file_types=allowed_medias)
+    audio_input = gr.File(label="Audio hinzufügen (optional, WAV)", file_types=[".wav"])
     duration_input = gr.Number(value=3, label="Dauer pro Bild in Sekunden", precision=1)
     fade_input = gr.Number(value=0.7, label="Fade Dauer in Sekunden", precision=1)
-    text_input = gr.Textbox(label="Texte pro Bild (mit Komma trennen)", placeholder="Text1, Text2, Text3 ...")
     ypos_input = gr.Slider(minimum=0.0, maximum=0.9, step=0.01, value=0.5, label="Y-Position für alle Texte (0=oben, 0.5=mitte, 0.9=unten)")
     font_size_input = gr.Number(value=60, label="Textgröße (px)")
     out_video = gr.Video(interactive=False, label="Generiertes Video")
     status = gr.Textbox(interactive=False, label="Status")
 
-    def wrapper(images, duration, fade_duration, text_str, y_pos, font_size):
-        texts = [t.strip() for t in text_str.split(",")] if text_str else None
-        return generate_slideshow(images, duration, texts, y_pos, fade_duration, font_size)
-
     btn = gr.Button("Video erstellen")
     btn.click(
-        fn=wrapper,
-        inputs=[img_input, duration_input, fade_input, text_input, ypos_input, font_size_input],
+        fn=generate_slideshow_with_audio,
+        inputs=[img_input, audio_input, duration_input, ypos_input, fade_input, font_size_input],
         outputs=[out_video, status]
     )