Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -2,6 +2,7 @@ import os
|
|
| 2 |
os.environ["IMAGE_MAGICK_BINARY"] = "/usr/bin/convert"
|
| 3 |
import gradio as gr
|
| 4 |
import pandas as pd
|
|
|
|
| 5 |
from faster_whisper import WhisperModel
|
| 6 |
from moviepy import VideoFileClip, TextClip, CompositeVideoClip
|
| 7 |
from arabic_reshaper import reshape
|
|
@@ -12,99 +13,77 @@ model = WhisperModel("large-v3", device="cpu", compute_type="int8")
|
|
| 12 |
def process_arabic_text(text):
|
| 13 |
return reshape(text) + "\n "
|
| 14 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 15 |
def step_1_extract_words(video_path, progress=gr.Progress()):
|
| 16 |
-
if not video_path:
|
| 17 |
-
return None, "الرجاء رفع فيديو أولاً."
|
| 18 |
-
|
| 19 |
-
progress(0, desc="جاري استخراج الكلمات...")
|
| 20 |
segments, _ = model.transcribe(video_path, word_timestamps=True, language="ar")
|
| 21 |
-
|
| 22 |
-
words_data =
|
| 23 |
-
for segment in segments:
|
| 24 |
-
for word in segment.words:
|
| 25 |
-
words_data.append([word.word.strip(), round(word.start, 2), round(word.end, 2)])
|
| 26 |
-
|
| 27 |
-
df = pd.DataFrame(words_data, columns=["الكلمة", "البداية", "النهاية"])
|
| 28 |
-
return df, "تم الاستخراج!"
|
| 29 |
|
| 30 |
def step_2_render_video(video_path, df_edited, font_selection, text_color, font_size, progress=gr.Progress()):
|
| 31 |
-
if video_path is None or df_edited is None
|
| 32 |
-
return None, "بيانات ناقصة."
|
| 33 |
|
| 34 |
-
#
|
| 35 |
-
|
|
|
|
|
|
|
|
|
|
| 36 |
|
| 37 |
-
output_path = "
|
| 38 |
video = VideoFileClip(video_path)
|
| 39 |
-
w, h =
|
| 40 |
clips = [video]
|
| 41 |
words_list = df_edited.values.tolist()
|
| 42 |
chunk_size = 3
|
| 43 |
|
| 44 |
-
progress(0.1, desc="جاري معالجة النصوص بالإعدادات الجديدة...")
|
| 45 |
-
|
| 46 |
for i in range(0, len(words_list), chunk_size):
|
| 47 |
-
|
| 48 |
-
sentence = " ".join([str(r[0]) for r in
|
| 49 |
clean_sentence = process_arabic_text(sentence)
|
| 50 |
|
| 51 |
-
|
| 52 |
-
c_end = float(current_chunk[-1][2])
|
| 53 |
-
duration = max(0.1, c_end - c_start)
|
| 54 |
|
| 55 |
-
# إنشاء الكليب مع التأكد من تمرير اللون والخط بشكل صريح
|
| 56 |
txt = TextClip(
|
| 57 |
text=clean_sentence,
|
| 58 |
font_size=int(font_size),
|
| 59 |
-
color=
|
| 60 |
stroke_color='black',
|
| 61 |
stroke_width=1.5,
|
| 62 |
font=actual_font,
|
| 63 |
method='caption',
|
| 64 |
-
size=(int(w * 0.
|
| 65 |
text_align='center'
|
| 66 |
-
).with_start(
|
| 67 |
|
| 68 |
clips.append(txt)
|
| 69 |
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
final_video.write_videofile(
|
| 74 |
-
output_path,
|
| 75 |
-
codec="libx264",
|
| 76 |
-
audio_codec="aac",
|
| 77 |
-
fps=video.fps,
|
| 78 |
-
logger='bar',
|
| 79 |
-
threads=4 # استخدام تعدد الخيوط لتسريع العملية
|
| 80 |
-
)
|
| 81 |
-
|
| 82 |
-
return output_path, f"تم الحفظ! اللون: {text_color}، الخط: {actual_font}"
|
| 83 |
|
| 84 |
# --- الواجهة ---
|
| 85 |
with gr.Blocks() as app:
|
| 86 |
-
gr.Markdown("##
|
| 87 |
-
|
| 88 |
with gr.Row():
|
| 89 |
-
v_in = gr.Video()
|
| 90 |
-
v_out = gr.Video()
|
| 91 |
-
|
| 92 |
with gr.Row():
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
value="arialbd.ttf",
|
| 97 |
-
label="نوع الخط"
|
| 98 |
-
)
|
| 99 |
-
color_opt = gr.ColorPicker(value="#FFFF00", label="لون النص")
|
| 100 |
-
size_opt = gr.Slider(30, 150, value=70, label="حجم الخط")
|
| 101 |
|
| 102 |
-
btn_1 = gr.Button("1. استخراج")
|
| 103 |
-
|
| 104 |
-
btn_2 = gr.Button("2. إنتاج الفيديو")
|
| 105 |
-
status = gr.Textbox(label="الحالة")
|
| 106 |
|
| 107 |
-
btn_1.click(step_1_extract_words,
|
| 108 |
-
btn_2.click(step_2_render_video,
|
| 109 |
|
| 110 |
app.launch()
|
|
|
|
| 2 |
os.environ["IMAGE_MAGICK_BINARY"] = "/usr/bin/convert"
|
| 3 |
import gradio as gr
|
| 4 |
import pandas as pd
|
| 5 |
+
import re # نحتاجه لتنظيف الألوان
|
| 6 |
from faster_whisper import WhisperModel
|
| 7 |
from moviepy import VideoFileClip, TextClip, CompositeVideoClip
|
| 8 |
from arabic_reshaper import reshape
|
|
|
|
| 13 |
def process_arabic_text(text):
|
| 14 |
return reshape(text) + "\n "
|
| 15 |
|
| 16 |
+
def clean_color(color_str):
|
| 17 |
+
"""دالة لتحويل أي صيغة لون غريبة إلى صيغة يفهمها البرنامج"""
|
| 18 |
+
if color_str.startswith('rgba'):
|
| 19 |
+
# استخراج الأرقام فقط وتحويلها لأرقام صحيحة
|
| 20 |
+
nums = re.findall(r"\d+\.?\d*", color_str)
|
| 21 |
+
if len(nums) >= 3:
|
| 22 |
+
r, g, b = int(float(nums[0])), int(float(nums[1])), int(float(nums[2]))
|
| 23 |
+
return f'rgb({r},{g},{b})'
|
| 24 |
+
return color_str
|
| 25 |
+
|
| 26 |
def step_1_extract_words(video_path, progress=gr.Progress()):
|
| 27 |
+
if not video_path: return None, "الرجاء رفع فيديو."
|
|
|
|
|
|
|
|
|
|
| 28 |
segments, _ = model.transcribe(video_path, word_timestamps=True, language="ar")
|
| 29 |
+
words_data = [[w.word.strip(), round(w.start, 2), round(w.end, 2)] for s in segments for w in s.words]
|
| 30 |
+
return pd.DataFrame(words_data, columns=["الكلمة", "البداية", "النهاية"]), "تم الاستخراج!"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
|
| 32 |
def step_2_render_video(video_path, df_edited, font_selection, text_color, font_size, progress=gr.Progress()):
|
| 33 |
+
if video_path is None or df_edited is None: return None, "بيانات ناقصة."
|
|
|
|
| 34 |
|
| 35 |
+
# تنظيف اللون قبل استخدامه لمنع خطأ ValueError
|
| 36 |
+
safe_color = clean_color(text_color)
|
| 37 |
+
|
| 38 |
+
# التأكد من وجود ملف الخط
|
| 39 |
+
actual_font = font_selection if os.path.exists(font_selection) else "DejaVu-Sans-Bold"
|
| 40 |
|
| 41 |
+
output_path = "final_fixed_video.mp4"
|
| 42 |
video = VideoFileClip(video_path)
|
| 43 |
+
w, h = video.size
|
| 44 |
clips = [video]
|
| 45 |
words_list = df_edited.values.tolist()
|
| 46 |
chunk_size = 3
|
| 47 |
|
|
|
|
|
|
|
| 48 |
for i in range(0, len(words_list), chunk_size):
|
| 49 |
+
chunk = words_list[i : i + chunk_size]
|
| 50 |
+
sentence = " ".join([str(r[0]) for r in chunk])
|
| 51 |
clean_sentence = process_arabic_text(sentence)
|
| 52 |
|
| 53 |
+
t_start, t_end = float(chunk[0][1]), float(chunk[-1][2])
|
|
|
|
|
|
|
| 54 |
|
|
|
|
| 55 |
txt = TextClip(
|
| 56 |
text=clean_sentence,
|
| 57 |
font_size=int(font_size),
|
| 58 |
+
color=safe_color, # اللون النظيف هنا
|
| 59 |
stroke_color='black',
|
| 60 |
stroke_width=1.5,
|
| 61 |
font=actual_font,
|
| 62 |
method='caption',
|
| 63 |
+
size=(int(w * 0.8), None),
|
| 64 |
text_align='center'
|
| 65 |
+
).with_start(t_start).with_duration(max(0.1, t_end - t_start)).with_position(('center', int(h * 0.65)))
|
| 66 |
|
| 67 |
clips.append(txt)
|
| 68 |
|
| 69 |
+
final = CompositeVideoClip(clips, size=(w, h))
|
| 70 |
+
final.write_videofile(output_path, codec="libx264", audio_codec="aac", fps=video.fps, logger='bar')
|
| 71 |
+
return output_path, f"تم بنجاح! اللون المستخدم: {safe_color}"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 72 |
|
| 73 |
# --- الواجهة ---
|
| 74 |
with gr.Blocks() as app:
|
| 75 |
+
gr.Markdown("## 🎬 Caption Pro: Stable Version")
|
|
|
|
| 76 |
with gr.Row():
|
| 77 |
+
v_in = gr.Video(); v_out = gr.Video()
|
|
|
|
|
|
|
| 78 |
with gr.Row():
|
| 79 |
+
font_opt = gr.Dropdown(choices=["arialbd.ttf", "Cairo-Bold.ttf", "Almarai-Bold.ttf"], value="arialbd.ttf", label="الخط")
|
| 80 |
+
color_opt = gr.ColorPicker(value="#FFFF00", label="اللون")
|
| 81 |
+
size_opt = gr.Slider(30, 150, value=70, label="الحجم")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 82 |
|
| 83 |
+
btn_1 = gr.Button("1. استخراج"); table = gr.Dataframe(interactive=True)
|
| 84 |
+
btn_2 = gr.Button("2. إنتاج الفيديو"); status = gr.Textbox()
|
|
|
|
|
|
|
| 85 |
|
| 86 |
+
btn_1.click(step_1_extract_words, [v_in], [table, status])
|
| 87 |
+
btn_2.click(step_2_render_video, [v_in, table, font_opt, color_opt, size_opt], [v_out, status])
|
| 88 |
|
| 89 |
app.launch()
|