Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -12,7 +12,9 @@ model = WhisperModel("large-v3", device="cpu", compute_type="int8")
|
|
| 12 |
|
| 13 |
def process_arabic_text(text):
|
| 14 |
if not text: return ""
|
|
|
|
| 15 |
reshaped = reshape(text)
|
|
|
|
| 16 |
return reshaped + "\n "
|
| 17 |
|
| 18 |
def clean_color(color_str):
|
|
@@ -23,32 +25,16 @@ def clean_color(color_str):
|
|
| 23 |
return f'rgb({r},{g},{b})'
|
| 24 |
return color_str
|
| 25 |
|
| 26 |
-
def
|
| 27 |
if not video_path: return None, "الرجاء رفع فيديو."
|
|
|
|
| 28 |
|
| 29 |
-
# استخراج النص العربي الأصلي مع التوقيت
|
| 30 |
-
segments_ar, _ = model.transcribe(video_path, word_timestamps=True, language="ar")
|
| 31 |
-
|
| 32 |
-
# طلب الترجمة إلى الإنجليزية من النموذج مباشرة
|
| 33 |
-
segments_en, _ = model.transcribe(video_path, task="translate")
|
| 34 |
-
|
| 35 |
-
# سنقوم بدمج الترجمة الإنجليزية مع الكلمات العربية بناءً على التوقيت تقريبياً
|
| 36 |
words_data = []
|
| 37 |
-
|
| 38 |
-
en_texts = list(segments_en)
|
| 39 |
-
|
| 40 |
-
for segment in segments_ar:
|
| 41 |
-
# البحث عن أقرب ترجمة إنجليزية لهذا الجزء
|
| 42 |
-
en_translation = ""
|
| 43 |
-
for en_seg in en_texts:
|
| 44 |
-
if (en_seg.start <= segment.start <= en_seg.end) or (segment.start <= en_seg.start <= segment.end):
|
| 45 |
-
en_translation = en_seg.text.strip()
|
| 46 |
-
break
|
| 47 |
-
|
| 48 |
for word in segment.words:
|
| 49 |
-
words_data.append([word.word.strip(),
|
| 50 |
|
| 51 |
-
return pd.DataFrame(words_data, columns=["ال
|
| 52 |
|
| 53 |
def step_2_render_video(video_path, df_edited, font_selection, text_color, font_size, progress=gr.Progress()):
|
| 54 |
if video_path is None or df_edited is None: return None, "بيانات ناقصة."
|
|
@@ -56,66 +42,53 @@ def step_2_render_video(video_path, df_edited, font_selection, text_color, font_
|
|
| 56 |
safe_color = clean_color(text_color)
|
| 57 |
actual_font = font_selection if os.path.exists(font_selection) else "DejaVu-Sans-Bold"
|
| 58 |
|
| 59 |
-
output_path = "
|
| 60 |
video = VideoFileClip(video_path)
|
| 61 |
w, h = video.size
|
| 62 |
clips = [video]
|
| 63 |
|
| 64 |
-
|
| 65 |
|
| 66 |
-
for row in
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
t_end = float(row[3])
|
| 71 |
|
| 72 |
-
if not
|
| 73 |
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
text=
|
| 78 |
font_size=int(font_size),
|
| 79 |
color=safe_color,
|
| 80 |
stroke_color='black',
|
| 81 |
-
stroke_width=2,
|
| 82 |
font=actual_font,
|
| 83 |
method='label'
|
| 84 |
-
).with_start(t_start).with_duration(max(0.1, t_end - t_start)).with_position(('center', int(h * 0.
|
| 85 |
-
|
| 86 |
-
# 2. النص الإنجليزي (صغير) في الأسفل
|
| 87 |
-
en_clip = TextClip(
|
| 88 |
-
text=en_text,
|
| 89 |
-
font_size=int(font_size * 0.4), # تصغير الترجمة لـ 40% من حجم العربي
|
| 90 |
-
color="white",
|
| 91 |
-
stroke_color='black',
|
| 92 |
-
stroke_width=1,
|
| 93 |
-
font="Arial", # خط بسيط للإنجليزي
|
| 94 |
-
method='caption',
|
| 95 |
-
size=(int(w * 0.8), None)
|
| 96 |
-
).with_start(t_start).with_duration(max(0.1, t_end - t_start)).with_position(('center', int(h * 0.6)))
|
| 97 |
|
| 98 |
-
clips.append(
|
| 99 |
-
clips.append(en_clip)
|
| 100 |
|
| 101 |
final = CompositeVideoClip(clips, size=(w, h))
|
| 102 |
final.write_videofile(output_path, codec="libx264", audio_codec="aac", fps=video.fps, logger='bar')
|
| 103 |
-
return output_path, "تم إنتاج الفيديو
|
| 104 |
|
| 105 |
# --- الواجهة ---
|
| 106 |
with gr.Blocks() as app:
|
| 107 |
-
gr.Markdown("## 🎬 محرر الفيديو
|
| 108 |
with gr.Row():
|
| 109 |
v_in = gr.Video(); v_out = gr.Video()
|
| 110 |
with gr.Row():
|
| 111 |
-
font_opt = gr.Dropdown(choices=["arialbd.ttf"], value="arialbd.ttf", label="الخط
|
| 112 |
-
color_opt = gr.ColorPicker(value="#FF8C00", label="لون ال
|
| 113 |
-
|
|
|
|
| 114 |
|
| 115 |
-
btn_1 = gr.Button("1. تحليل
|
| 116 |
btn_2 = gr.Button("2. إنتاج الفيديو"); status = gr.Textbox()
|
| 117 |
|
| 118 |
-
btn_1.click(
|
| 119 |
btn_2.click(step_2_render_video, [v_in, table, font_opt, color_opt, size_opt], [v_out, status])
|
| 120 |
|
| 121 |
app.launch()
|
|
|
|
| 12 |
|
| 13 |
def process_arabic_text(text):
|
| 14 |
if not text: return ""
|
| 15 |
+
# 1. إعادة تشكيل الحروف العربية لتظهر متصلة وصحيحة (بدون النقاط الإضافية •)
|
| 16 |
reshaped = reshape(text)
|
| 17 |
+
# 2. إضافة سطر فارغ في الأسفل لمنع قص النقاط السفلية للأحرف
|
| 18 |
return reshaped + "\n "
|
| 19 |
|
| 20 |
def clean_color(color_str):
|
|
|
|
| 25 |
return f'rgb({r},{g},{b})'
|
| 26 |
return color_str
|
| 27 |
|
| 28 |
+
def step_1_extract_words(video_path, progress=gr.Progress()):
|
| 29 |
if not video_path: return None, "الرجاء رفع فيديو."
|
| 30 |
+
segments, _ = model.transcribe(video_path, word_timestamps=True, language="ar")
|
| 31 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 32 |
words_data = []
|
| 33 |
+
for segment in segments:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 34 |
for word in segment.words:
|
| 35 |
+
words_data.append([word.word.strip(), round(word.start, 2), round(word.end, 2)])
|
| 36 |
|
| 37 |
+
return pd.DataFrame(words_data, columns=["الكلمة", "البداية", "النهاية"]), "تم استخراج الكلمات!"
|
| 38 |
|
| 39 |
def step_2_render_video(video_path, df_edited, font_selection, text_color, font_size, progress=gr.Progress()):
|
| 40 |
if video_path is None or df_edited is None: return None, "بيانات ناقصة."
|
|
|
|
| 42 |
safe_color = clean_color(text_color)
|
| 43 |
actual_font = font_selection if os.path.exists(font_selection) else "DejaVu-Sans-Bold"
|
| 44 |
|
| 45 |
+
output_path = "final_clean_text_video.mp4"
|
| 46 |
video = VideoFileClip(video_path)
|
| 47 |
w, h = video.size
|
| 48 |
clips = [video]
|
| 49 |
|
| 50 |
+
words_list = df_edited.values.tolist()
|
| 51 |
|
| 52 |
+
for row in words_list:
|
| 53 |
+
word_text = str(row[0])
|
| 54 |
+
t_start = float(row[1])
|
| 55 |
+
t_end = float(row[2])
|
|
|
|
| 56 |
|
| 57 |
+
if not word_text.strip(): continue
|
| 58 |
|
| 59 |
+
clean_word = process_arabic_text(word_text)
|
| 60 |
+
|
| 61 |
+
txt = TextClip(
|
| 62 |
+
text=clean_word,
|
| 63 |
font_size=int(font_size),
|
| 64 |
color=safe_color,
|
| 65 |
stroke_color='black',
|
| 66 |
+
stroke_width=2.0, # تقليل سمك التحديد قليلاً ليتناسب مع النص الأصغر
|
| 67 |
font=actual_font,
|
| 68 |
method='label'
|
| 69 |
+
).with_start(t_start).with_duration(max(0.1, t_end - t_start)).with_position(('center', int(h * 0.5)))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 70 |
|
| 71 |
+
clips.append(txt)
|
|
|
|
| 72 |
|
| 73 |
final = CompositeVideoClip(clips, size=(w, h))
|
| 74 |
final.write_videofile(output_path, codec="libx264", audio_codec="aac", fps=video.fps, logger='bar')
|
| 75 |
+
return output_path, "تم إنتاج الفيديو بنجاح!"
|
| 76 |
|
| 77 |
# --- الواجهة ---
|
| 78 |
with gr.Blocks() as app:
|
| 79 |
+
gr.Markdown("## 🎬 محرر الفيديو: نصوص نظيفة")
|
| 80 |
with gr.Row():
|
| 81 |
v_in = gr.Video(); v_out = gr.Video()
|
| 82 |
with gr.Row():
|
| 83 |
+
font_opt = gr.Dropdown(choices=["arialbd.ttf"], value="arialbd.ttf", label="الخط")
|
| 84 |
+
color_opt = gr.ColorPicker(value="#FF8C00", label="لون ذهبي برتقالي")
|
| 85 |
+
# تم تصغير الحجم الافتراضي من 130 إلى 90
|
| 86 |
+
size_opt = gr.Slider(30, 200, value=90, label="حجم النص")
|
| 87 |
|
| 88 |
+
btn_1 = gr.Button("1. تحليل الكلمات"); table = gr.Dataframe(interactive=True)
|
| 89 |
btn_2 = gr.Button("2. إنتاج الفيديو"); status = gr.Textbox()
|
| 90 |
|
| 91 |
+
btn_1.click(step_1_extract_words, [v_in], [table, status])
|
| 92 |
btn_2.click(step_2_render_video, [v_in, table, font_opt, color_opt, size_opt], [v_out, status])
|
| 93 |
|
| 94 |
app.launch()
|