Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -27,24 +27,19 @@ def get_font_path():
|
|
| 27 |
return None # Fallback: FFmpeg soll selbst suchen (klappt manchmal nicht)
|
| 28 |
|
| 29 |
def save_temp_audio(audio_file):
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
if ext.lower() not in allowed_audios:
|
| 33 |
-
ext = ".mp3"
|
| 34 |
-
temp_audio = Path(tempfile.mkdtemp()) / f"input{ext}"
|
| 35 |
-
with open(temp_audio, "wb") as f:
|
| 36 |
-
f.write(audio_file.encode())
|
| 37 |
-
return temp_audio
|
| 38 |
-
elif hasattr(audio_file, 'name'):
|
| 39 |
ext = Path(audio_file.name).suffix
|
| 40 |
if ext.lower() not in allowed_audios:
|
| 41 |
ext = ".mp3"
|
| 42 |
-
|
|
|
|
| 43 |
audio_file.seek(0)
|
| 44 |
with open(temp_audio, "wb") as f:
|
| 45 |
shutil.copyfileobj(audio_file, f)
|
| 46 |
-
|
| 47 |
-
|
|
|
|
| 48 |
|
| 49 |
def create_timed_drawtext(word, start_time, duration, font_option, font_size, y_pos):
|
| 50 |
"""Erstellt einen FFmpeg drawtext Filter, der ein Wort mit weichen Übergängen (Alpha-Kanal) einblendet."""
|
|
@@ -62,10 +57,7 @@ def create_timed_drawtext(word, start_time, duration, font_option, font_size, y_
|
|
| 62 |
fade_out_start = end_time - WORD_FADE_DURATION
|
| 63 |
|
| 64 |
# Alpha-Ausdruck für smooth Fade-In und Fade-Out
|
| 65 |
-
#
|
| 66 |
-
# if(lt(t, fade_in_end), (t-start_time)/WORD_FADE_DURATION, ...) -> Fade-In
|
| 67 |
-
# if(lt(t, fade_out_start), 1, ...) -> Volle Deckkraft
|
| 68 |
-
# if(lt(t, end_time), (end_time-t)/WORD_FADE_DURATION, 0) -> Fade-Out
|
| 69 |
alpha_expression = (
|
| 70 |
f"if(lt(t,{start_time}), 0, "
|
| 71 |
f"if(lt(t,{fade_in_end}), (t-{start_time})/{WORD_FADE_DURATION}, "
|
|
@@ -83,102 +75,80 @@ def create_timed_drawtext(word, start_time, duration, font_option, font_size, y_
|
|
| 83 |
|
| 84 |
|
| 85 |
def generate_slideshow_with_audio(images, input_text, duration_per_word, duration_per_image, fade_duration, font_size, y_pos, audio_file):
|
| 86 |
-
|
| 87 |
-
print(f"DEBUG: Font Size: {font_size}, Y-Pos: {y_pos}, Duration/Word: {duration_per_word}, Fade: {fade_duration}")
|
| 88 |
-
|
| 89 |
if not images:
|
| 90 |
return None, "❌ Keine Bilder ausgewählt"
|
| 91 |
|
| 92 |
temp_dir = tempfile.mkdtemp()
|
| 93 |
-
|
| 94 |
-
|
| 95 |
# Text in Wörter aufteilen
|
| 96 |
words = input_text.split() if input_text else []
|
|
|
|
|
|
|
| 97 |
|
| 98 |
-
#
|
| 99 |
-
|
|
|
|
| 100 |
|
|
|
|
|
|
|
|
|
|
| 101 |
# Schriftart finden
|
| 102 |
font_path = get_font_path()
|
| 103 |
font_option = f":fontfile='{font_path}'" if font_path else ""
|
| 104 |
|
| 105 |
# Audio verarbeiten
|
| 106 |
-
temp_audio_file = None
|
| 107 |
-
if audio_file:
|
| 108 |
-
temp_audio_file = save_temp_audio(audio_file)
|
| 109 |
|
| 110 |
-
# --- 1. ERSTES BILD (Sequenzieller Text) ---
|
| 111 |
-
|
| 112 |
-
# Dauer des ersten Clips ist mindestens so lang wie der Text
|
| 113 |
-
duration_clip_1 = max(duration_per_image, total_text_duration)
|
| 114 |
-
|
| 115 |
-
# **KORREKTUR für FFmpeg:** Fade-Out Startzeit in Python berechnen
|
| 116 |
-
fade_out_start_1 = duration_clip_1 - fade_duration
|
| 117 |
-
if fade_out_start_1 < 0: fade_out_start_1 = 0
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
# Generiere die sequentiellen Drawtext-Filter
|
| 121 |
-
drawtext_filters = []
|
| 122 |
-
current_time = 0.0
|
| 123 |
-
for word in words:
|
| 124 |
-
filter_str = create_timed_drawtext(word, current_time, duration_per_word, font_option, font_size, y_pos)
|
| 125 |
-
drawtext_filters.append(filter_str)
|
| 126 |
-
current_time += duration_per_word
|
| 127 |
-
|
| 128 |
-
# 1. Basisanpassungen
|
| 129 |
-
base_filters = (
|
| 130 |
-
"scale=w=1280:h=720:force_original_aspect_ratio=decrease,"
|
| 131 |
-
"pad=1280:720:(ow-iw)/2:(oh-ih)/2:color=black,"
|
| 132 |
-
"fps=25,format=yuv420p"
|
| 133 |
-
)
|
| 134 |
-
|
| 135 |
-
# 2. Fade-Filter (jetzt mit korrekter berechneter Startzeit)
|
| 136 |
-
fade_img_filter_1 = f"fade=t=in:st=0:d={fade_duration},fade=t=out:st={fade_out_start_1}:d={fade_duration}"
|
| 137 |
|
| 138 |
-
|
| 139 |
-
|
| 140 |
-
all_drawtext_filters = ",".join(drawtext_filters)
|
| 141 |
-
vf_filters_clip1 = f"{base_filters},{all_drawtext_filters},{fade_img_filter_1}"
|
| 142 |
-
else:
|
| 143 |
-
vf_filters_clip1 = f"{base_filters},{fade_img_filter_1}"
|
| 144 |
-
|
| 145 |
-
# Erstelle Clip 1
|
| 146 |
-
img_path_1 = Path(images[0].name)
|
| 147 |
-
clip_path_1 = Path(temp_dir) / "clip_with_text_0.mp4"
|
| 148 |
-
|
| 149 |
-
cmd_1 = [
|
| 150 |
-
"ffmpeg", "-y", "-loop", "1", "-i", str(img_path_1),
|
| 151 |
-
"-t", str(duration_clip_1),
|
| 152 |
-
"-vf", vf_filters_clip1,
|
| 153 |
-
str(clip_path_1)
|
| 154 |
-
]
|
| 155 |
-
|
| 156 |
-
try:
|
| 157 |
-
subprocess.run(cmd_1, check=True, capture_output=True, text=True)
|
| 158 |
-
clips_with_text.append(clip_path_1)
|
| 159 |
-
except subprocess.CalledProcessError as e:
|
| 160 |
-
return None, f"❌ FFmpeg Fehler bei Bild 1 (mit Text):\n{e.stderr}"
|
| 161 |
-
|
| 162 |
-
# --- 2. FOLGE-BILDER (Nur Bild mit Fade) ---
|
| 163 |
-
for i in range(1, len(images)):
|
| 164 |
img_path = Path(images[i].name)
|
| 165 |
-
clip_path = Path(temp_dir) / f"
|
| 166 |
-
|
| 167 |
-
# **KORREKTUR für FFmpeg:** Fade-Out Startzeit in Python berechnen
|
| 168 |
-
fade_out_start_n = duration_per_image - fade_duration
|
| 169 |
-
if fade_out_start_n < 0: fade_out_start_n = 0
|
| 170 |
|
| 171 |
-
#
|
| 172 |
-
|
| 173 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 174 |
"scale=w=1280:h=720:force_original_aspect_ratio=decrease,"
|
| 175 |
"pad=1280:720:(ow-iw)/2:(oh-ih)/2:color=black,"
|
| 176 |
-
|
| 177 |
)
|
| 178 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 179 |
cmd = [
|
| 180 |
"ffmpeg", "-y", "-loop", "1", "-i", str(img_path),
|
| 181 |
-
"-t", str(
|
| 182 |
"-vf", vf_filters_clip,
|
| 183 |
str(clip_path)
|
| 184 |
]
|
|
@@ -187,9 +157,12 @@ def generate_slideshow_with_audio(images, input_text, duration_per_word, duratio
|
|
| 187 |
subprocess.run(cmd, check=True, capture_output=True, text=True)
|
| 188 |
clips_with_text.append(clip_path)
|
| 189 |
except subprocess.CalledProcessError as e:
|
| 190 |
-
|
|
|
|
|
|
|
|
|
|
| 191 |
|
| 192 |
-
#
|
| 193 |
filelist_path = Path(temp_dir) / "filelist.txt"
|
| 194 |
with open(filelist_path, "w") as f:
|
| 195 |
for clip in clips_with_text:
|
|
@@ -205,11 +178,14 @@ def generate_slideshow_with_audio(images, input_text, duration_per_word, duratio
|
|
| 205 |
]
|
| 206 |
|
| 207 |
try:
|
| 208 |
-
subprocess.run(cmd_concat, check=True)
|
| 209 |
except subprocess.CalledProcessError as e:
|
|
|
|
|
|
|
| 210 |
return None, f"❌ FFmpeg Fehler beim Zusammenfügen:\n{e.stderr}"
|
| 211 |
|
| 212 |
-
#
|
|
|
|
| 213 |
if temp_audio_file:
|
| 214 |
final_output = Path(temp_dir) / f"final_{uuid.uuid4().hex}.mp4"
|
| 215 |
cmd_audio = [
|
|
@@ -218,12 +194,19 @@ def generate_slideshow_with_audio(images, input_text, duration_per_word, duratio
|
|
| 218 |
str(final_output)
|
| 219 |
]
|
| 220 |
try:
|
| 221 |
-
subprocess.run(cmd_audio, check=True)
|
| 222 |
except subprocess.CalledProcessError as e:
|
|
|
|
|
|
|
| 223 |
return None, f"❌ FFmpeg Fehler beim Hinzufügen von Audio:\n{e.stderr}"
|
|
|
|
|
|
|
|
|
|
|
|
|
| 224 |
return str(final_output), "✅ Video mit Audio erstellt!"
|
| 225 |
|
| 226 |
-
|
|
|
|
| 227 |
|
| 228 |
# Gradio UI
|
| 229 |
with gr.Blocks() as demo:
|
|
@@ -231,12 +214,13 @@ with gr.Blocks() as demo:
|
|
| 231 |
|
| 232 |
with gr.Row():
|
| 233 |
img_input = gr.Files(label="Bilder", file_types=allowed_medias)
|
| 234 |
-
|
|
|
|
| 235 |
|
| 236 |
with gr.Row():
|
| 237 |
-
duration_image_input = gr.Number(value=3, label="Dauer pro BILD (s)
|
| 238 |
duration_word_input = gr.Number(value=1.0, label="Dauer pro WORT (s) [bestimmt Geschwindigkeit der Text-Anzeige]")
|
| 239 |
-
fade_input = gr.Number(value=0.5, label="Bild-Fade Dauer (s)")
|
| 240 |
font_size_input = gr.Number(value=80, label="Schriftgröße (px)")
|
| 241 |
ypos_input = gr.Slider(0.0, 1.0, value=0.9, label="Y-Position (0=Oben, 1=Unten)")
|
| 242 |
|
|
|
|
| 27 |
return None # Fallback: FFmpeg soll selbst suchen (klappt manchmal nicht)
|
| 28 |
|
| 29 |
def save_temp_audio(audio_file):
|
| 30 |
+
"""Speichert die hochgeladene Audio-Datei in einem temporären Verzeichnis."""
|
| 31 |
+
if hasattr(audio_file, 'name'):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 32 |
ext = Path(audio_file.name).suffix
|
| 33 |
if ext.lower() not in allowed_audios:
|
| 34 |
ext = ".mp3"
|
| 35 |
+
temp_audio_dir = Path(tempfile.mkdtemp())
|
| 36 |
+
temp_audio = temp_audio_dir / f"input{ext}"
|
| 37 |
audio_file.seek(0)
|
| 38 |
with open(temp_audio, "wb") as f:
|
| 39 |
shutil.copyfileobj(audio_file, f)
|
| 40 |
+
# Rückgabe des Verzeichnisses, das später gelöscht werden kann, und des Dateipfads
|
| 41 |
+
return temp_audio_dir, temp_audio
|
| 42 |
+
return None, None
|
| 43 |
|
| 44 |
def create_timed_drawtext(word, start_time, duration, font_option, font_size, y_pos):
|
| 45 |
"""Erstellt einen FFmpeg drawtext Filter, der ein Wort mit weichen Übergängen (Alpha-Kanal) einblendet."""
|
|
|
|
| 57 |
fade_out_start = end_time - WORD_FADE_DURATION
|
| 58 |
|
| 59 |
# Alpha-Ausdruck für smooth Fade-In und Fade-Out
|
| 60 |
+
# Steuert die Deckkraft basierend auf der Zeit t (relativ zum Clip-Start)
|
|
|
|
|
|
|
|
|
|
| 61 |
alpha_expression = (
|
| 62 |
f"if(lt(t,{start_time}), 0, "
|
| 63 |
f"if(lt(t,{fade_in_end}), (t-{start_time})/{WORD_FADE_DURATION}, "
|
|
|
|
| 75 |
|
| 76 |
|
| 77 |
def generate_slideshow_with_audio(images, input_text, duration_per_word, duration_per_image, fade_duration, font_size, y_pos, audio_file):
|
| 78 |
+
|
|
|
|
|
|
|
| 79 |
if not images:
|
| 80 |
return None, "❌ Keine Bilder ausgewählt"
|
| 81 |
|
| 82 |
temp_dir = tempfile.mkdtemp()
|
| 83 |
+
|
|
|
|
| 84 |
# Text in Wörter aufteilen
|
| 85 |
words = input_text.split() if input_text else []
|
| 86 |
+
total_words = len(words)
|
| 87 |
+
num_images = len(images)
|
| 88 |
|
| 89 |
+
# Berechnung der gleichmäßigen Verteilung der Wörter auf die Bilder
|
| 90 |
+
base_words_per_clip = total_words // num_images
|
| 91 |
+
remainder = total_words % num_images
|
| 92 |
|
| 93 |
+
current_word_index = 0
|
| 94 |
+
clips_with_text = [] # Paths der generierten MP4-Clips
|
| 95 |
+
|
| 96 |
# Schriftart finden
|
| 97 |
font_path = get_font_path()
|
| 98 |
font_option = f":fontfile='{font_path}'" if font_path else ""
|
| 99 |
|
| 100 |
# Audio verarbeiten
|
| 101 |
+
audio_temp_dir, temp_audio_file = save_temp_audio(audio_file) if audio_file else (None, None)
|
|
|
|
|
|
|
| 102 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 103 |
|
| 104 |
+
# --- 1. SCHLEIFE: Erstelle jeden Clip mit seinem Textsegment ---
|
| 105 |
+
for i in range(num_images):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 106 |
img_path = Path(images[i].name)
|
| 107 |
+
clip_path = Path(temp_dir) / f"clip_with_text_{i}.mp4"
|
|
|
|
|
|
|
|
|
|
|
|
|
| 108 |
|
| 109 |
+
# 1. Bestimme das Wortsegment für diesen Clip
|
| 110 |
+
words_on_this_clip = base_words_per_clip + (1 if i < remainder else 0)
|
| 111 |
+
|
| 112 |
+
# Extrahieren des Segments aus der Gesamtliste der Wörter
|
| 113 |
+
word_segment = words[current_word_index : current_word_index + words_on_this_clip]
|
| 114 |
+
current_word_index += len(word_segment)
|
| 115 |
+
|
| 116 |
+
# 2. Berechne die Clip-Dauer
|
| 117 |
+
text_duration = len(word_segment) * duration_per_word
|
| 118 |
+
# Die Dauer ist das Maximum aus der gewünschten Bilddauer und der benötigten Textdauer
|
| 119 |
+
duration_clip = max(duration_per_image, text_duration)
|
| 120 |
+
|
| 121 |
+
# 3. Generiere Drawtext Filter (Startzeit ist relativ zum Clip-Start, also 0)
|
| 122 |
+
drawtext_filters = []
|
| 123 |
+
word_start_time = 0.0
|
| 124 |
+
for word in word_segment:
|
| 125 |
+
filter_str = create_timed_drawtext(word, word_start_time, duration_per_word, font_option, font_size, y_pos)
|
| 126 |
+
drawtext_filters.append(filter_str)
|
| 127 |
+
word_start_time += duration_per_word
|
| 128 |
+
|
| 129 |
+
# 4. Basis- und Fade-Filter
|
| 130 |
+
base_filters = (
|
| 131 |
"scale=w=1280:h=720:force_original_aspect_ratio=decrease,"
|
| 132 |
"pad=1280:720:(ow-iw)/2:(oh-ih)/2:color=black,"
|
| 133 |
+
"fps=25,format=yuv420p"
|
| 134 |
)
|
| 135 |
|
| 136 |
+
fade_out_start = duration_clip - fade_duration
|
| 137 |
+
if fade_out_start < 0: fade_out_start = 0
|
| 138 |
+
fade_img_filter = f"fade=t=in:st=0:d={fade_duration},fade=t=out:st={fade_out_start}:d={fade_duration}"
|
| 139 |
+
|
| 140 |
+
# 5. Kombiniere alle Filter
|
| 141 |
+
if drawtext_filters:
|
| 142 |
+
all_drawtext_filters = ",".join(drawtext_filters)
|
| 143 |
+
vf_filters_clip = f"{base_filters},{all_drawtext_filters},{fade_img_filter}"
|
| 144 |
+
else:
|
| 145 |
+
# Kein Text mehr: Nur Bild mit Fade
|
| 146 |
+
vf_filters_clip = f"{base_filters},{fade_img_filter}"
|
| 147 |
+
|
| 148 |
+
# 6. FFmpeg Command zum Erstellen des Clips
|
| 149 |
cmd = [
|
| 150 |
"ffmpeg", "-y", "-loop", "1", "-i", str(img_path),
|
| 151 |
+
"-t", str(duration_clip),
|
| 152 |
"-vf", vf_filters_clip,
|
| 153 |
str(clip_path)
|
| 154 |
]
|
|
|
|
| 157 |
subprocess.run(cmd, check=True, capture_output=True, text=True)
|
| 158 |
clips_with_text.append(clip_path)
|
| 159 |
except subprocess.CalledProcessError as e:
|
| 160 |
+
# Bereinigung bei Fehler
|
| 161 |
+
shutil.rmtree(temp_dir)
|
| 162 |
+
if audio_temp_dir: shutil.rmtree(audio_temp_dir)
|
| 163 |
+
return None, f"❌ FFmpeg Fehler bei Bild {i+1}:\n{e.stderr}"
|
| 164 |
|
| 165 |
+
# --- 2. ZUSAMMENFÜGEN ---
|
| 166 |
filelist_path = Path(temp_dir) / "filelist.txt"
|
| 167 |
with open(filelist_path, "w") as f:
|
| 168 |
for clip in clips_with_text:
|
|
|
|
| 178 |
]
|
| 179 |
|
| 180 |
try:
|
| 181 |
+
subprocess.run(cmd_concat, check=True, capture_output=True, text=True)
|
| 182 |
except subprocess.CalledProcessError as e:
|
| 183 |
+
shutil.rmtree(temp_dir)
|
| 184 |
+
if audio_temp_dir: shutil.rmtree(audio_temp_dir)
|
| 185 |
return None, f"❌ FFmpeg Fehler beim Zusammenfügen:\n{e.stderr}"
|
| 186 |
|
| 187 |
+
# --- 3. AUDIO HINZUFÜGEN (falls vorhanden) ---
|
| 188 |
+
final_output = output_video
|
| 189 |
if temp_audio_file:
|
| 190 |
final_output = Path(temp_dir) / f"final_{uuid.uuid4().hex}.mp4"
|
| 191 |
cmd_audio = [
|
|
|
|
| 194 |
str(final_output)
|
| 195 |
]
|
| 196 |
try:
|
| 197 |
+
subprocess.run(cmd_audio, check=True, capture_output=True, text=True)
|
| 198 |
except subprocess.CalledProcessError as e:
|
| 199 |
+
shutil.rmtree(temp_dir)
|
| 200 |
+
if audio_temp_dir: shutil.rmtree(audio_temp_dir)
|
| 201 |
return None, f"❌ FFmpeg Fehler beim Hinzufügen von Audio:\n{e.stderr}"
|
| 202 |
+
|
| 203 |
+
# Bereinige das separate Audio-Temp-Verzeichnis
|
| 204 |
+
if audio_temp_dir: shutil.rmtree(audio_temp_dir)
|
| 205 |
+
|
| 206 |
return str(final_output), "✅ Video mit Audio erstellt!"
|
| 207 |
|
| 208 |
+
# Nur Video-Pfad zurückgeben
|
| 209 |
+
return str(final_output), "✅ Video erstellt (ohne Audio)"
|
| 210 |
|
| 211 |
# Gradio UI
|
| 212 |
with gr.Blocks() as demo:
|
|
|
|
| 214 |
|
| 215 |
with gr.Row():
|
| 216 |
img_input = gr.Files(label="Bilder", file_types=allowed_medias)
|
| 217 |
+
# TEXT WURDE GEÄNDERT: Neue Beschreibung für Textverteilung
|
| 218 |
+
text_input = gr.Textbox(label="Text (Wörter werden gleichmäßig auf alle Bilder verteilt)", lines=5, placeholder="Jedes Wort wird für 'Dauer pro Wort' angezeigt.")
|
| 219 |
|
| 220 |
with gr.Row():
|
| 221 |
+
duration_image_input = gr.Number(value=3, label="Mindest-Dauer pro BILD (s)")
|
| 222 |
duration_word_input = gr.Number(value=1.0, label="Dauer pro WORT (s) [bestimmt Geschwindigkeit der Text-Anzeige]")
|
| 223 |
+
fade_input = gr.Number(value=0.5, label="Bild-Fade Dauer (s)")
|
| 224 |
font_size_input = gr.Number(value=80, label="Schriftgröße (px)")
|
| 225 |
ypos_input = gr.Slider(0.0, 1.0, value=0.9, label="Y-Position (0=Oben, 1=Unten)")
|
| 226 |
|