import gradio as gr from faster_whisper import WhisperModel import os import subprocess import cv2 import asyncio import edge_tts import shutil import time import numpy as np import re import json import uuid # 👈 Session ID ထုတ်ရန်အတွက် Import လုပ်ခြင်း from google import genai from datetime import datetime, date from PIL import Image, ImageDraw, ImageFont # ===================================================================== # ⚙️ SECURITY SETTINGS & LOGIN CONFIG # ===================================================================== MAX_SUBTITLE_DURATION_SECONDS = 7.0 # 👥 LOGIN ACCOUNTS CONFIG (အကောင့်တစ်ခုတည်းကို လူအများ မျှသုံးနိုင်ပါသည်) ACCOUNTS = { "user1": "123123", # Free User (Session/Browser အလိုက် တစ်ရက် ၂ ပုဒ်) "admin": "admin789" # Admin User (Unlimited) } EXPIRY_DATE_STR = "2026-8-31" # 🛑 USER RATE LIMIT TRACKER (Session/Browser ID အလိုက် ရေတွက်မည်) USER_LIMIT_TRACKER = {} SYSTEM_INSTRUCTION = "You are a professional movie recap writer. Translate movie subtitle lines into natural, engaging, and thrilling Burmese movie recap style. Keep it concise." MODEL_NAME = "gemini-2.5-flash" def check_app_expiry(): try: expiry_date = datetime.strptime(EXPIRY_DATE_STR, "%Y-%m-%d").date() current_date = datetime.now().date() if current_date > expiry_date: return False, f"❌ ဤ App သည် သက်တမ်းကုန်ဆုံးသွားပါပြီ ({EXPIRY_DATE_STR})။" return True, "Active" except Exception as e: return False, f"လုံခြုံရေး စစ်ဆေးမှု မှားယွင်းနေပါသည်- {str(e)}" print("Loading Multilingual Faster-Whisper Base Model...") model = WhisperModel("base", device="cpu", compute_type="int8", cpu_threads=4) # ===================================================================== # 🖼️ REAL-TIME INTERACTIVE PREVIEW GENERATOR # ===================================================================== def update_preview_image(video_path, blur_y_percent, blur_strength): if not video_path: return None try: cap = cv2.VideoCapture(video_path) ret, frame = cap.read() cap.release() if not ret: return None h_o, w_o, _ = frame.shape b_h = int(h_o * 0.12) b_y = max(0, min(int(h_o * (blur_y_percent / 100)) - (b_h // 2), h_o - b_h)) k_size = int(blur_strength) if k_size % 2 == 0: k_size += 1 preview_frame = frame.copy() roi = preview_frame[b_y:b_y+b_h, 0:w_o] if roi.shape[0] > 0 and roi.shape[1] > 0: small_roi = cv2.resize(roi, (w_o // 4, b_h // 4), interpolation=cv2.INTER_LINEAR) blurred_small = cv2.GaussianBlur(small_roi, (k_size // 4 | 1, k_size // 4 | 1), 0) preview_frame[b_y:b_y+b_h, 0:w_o] = cv2.resize(blurred_small, (w_o, b_h), interpolation=cv2.INTER_LINEAR) cv2.rectangle(preview_frame, (0, b_y), (w_o, b_y+b_h), (0, 0, 255), 3) preview_rgb = cv2.cvtColor(preview_frame, cv2.COLOR_BGR2RGB) return Image.fromarray(preview_rgb) except Exception as e: print(f"Preview Error: {e}") return None # ===================================================================== # ⚡ DYNAMIC USER-PROVIDED API KEY TRANSLATION ENGINE # ===================================================================== def google_backup_translate(text, source_lang="en"): from urllib.parse import quote import urllib.request import json try: url = f"https://translate.googleapis.com/translate_a/single?client=gtx&sl={source_lang}&tl=my&dt=t&q={quote(text)}" req = urllib.request.Request(url, headers={'User-Agent': 'Mozilla/5.0'}) response = urllib.request.urlopen(req, timeout=5).read().decode('utf-8') result = json.loads(response) return result[0][0][0] or text except: return text def translate_segments_batch(segments, user_api_key, source_lang="en"): if not segments: return segments if not user_api_key or not user_api_key.strip(): print(f"⚠️ User က Gemini API Key မထည့်ထားပါ။ Google Translate Backup စနစ်ဖြင့် ဘာသာပြန်နေပါသည်။ (Source Lang: {source_lang})") for seg in segments: seg['mm_text'] = google_backup_translate(seg['text'], source_lang=source_lang) return segments payload_dict = {str(idx): seg['text'] for idx, seg in enumerate(segments)} large_prompt_text = json.dumps(payload_dict, ensure_ascii=False, indent=2) prompt = f""" You are an expert movie recap translator. Translate the following movie subtitle lines (which are originally in '{source_lang}' language) into thrilling, natural, and engaging Burmese movie recap style. CRITICAL: You MUST respond in valid JSON format only, keeping the exact same keys (0, 1, 2, etc.) as the input. The values should be the translated Burmese text. Do NOT include any markdown formatting like ```json or ``` in your response. Respond with pure JSON raw string only. Input Data: {large_prompt_text} """ translated_map = {} try: client = genai.Client(api_key=user_api_key.strip()) response = client.models.generate_content( model=MODEL_NAME, contents=prompt, config={ "system_instruction": SYSTEM_INSTRUCTION, "temperature": 0.3, } ) response_text = response.text.strip() if response_text.startswith("```"): response_text = response_text.split("\n", 1)[1].rsplit("\n", 1)[0].strip() if response_text.startswith("json"): response_text = response_text.split("\n", 1)[1].strip() translated_map = json.loads(response_text) except Exception as e: print(f"⚠️ User Gemini API Error သို့မဟုတ် Format Error: {e} -> Google Translate သို့ ပြောင်းလဲနေသည်။") for idx, seg in enumerate(segments): key_str = str(idx) if key_str in translated_map and translated_map[key_str]: seg['mm_text'] = translated_map[key_str] else: seg['mm_text'] = google_backup_translate(seg['text'], source_lang=source_lang) return segments # ===================================================================== # 🎬 TEXT WRAPPING & RENDER SYSTEMS # ===================================================================== def segment_myanmar_syllables(text): return re.findall(r'[a-zA-Z0-9\s\-\.,!\?]+|[\u1000-\u102a\u103f\u1040-\u1049]+[\u102b-\u103e\u1060-\u109f]*|[^\s]', text) def wrap_text_myanmar_smart(text, font, max_width, draw): cleaned_text = text.replace(" ြ", "ြ").replace("ြ ", "ြ").strip() tokens = segment_myanmar_syllables(cleaned_text) lines, current_line = [], "" for token in tokens: test_line = current_line + token bbox = draw.textbbox((0, 0), test_line, font=font) if (bbox[2] - bbox[0]) <= max_width: current_line = test_line else: if current_line: lines.append(current_line.strip()) current_line = token if current_line: lines.append(current_line.strip()) return lines def hex_to_rgb(hex_str): if not hex_str: return (255, 255, 0) hex_str = hex_str.lstrip('#') return tuple(int(hex_str[i:i+2], 16) for i in (0, 2, 4)) def draw_line_perfect_rendering(draw, position, text, font_primary, fill_color, stroke_w, stroke_c): x, y = position clean_text = text.replace(" ြ", "ြ").replace("ြ ", "ြ") draw.text((x, y), clean_text, font=font_primary, fill=fill_color, stroke_width=stroke_w, stroke_fill=stroke_c) def generate_voice_sync(text, voice_id, filename, desired_speed, target_duration_sec=None): rate_percentage = int((desired_speed - 1.0) * 100) rate_str = f"{'+' if rate_percentage >= 0 else ''}{rate_percentage}%" if target_duration_sec and target_duration_sec > 0: char_count = len(text) cps = char_count / target_duration_sec if cps > 15: rate_str = f"+{rate_percentage + 30}%" elif cps > 11: rate_str = f"+{rate_percentage + 15}%" elif cps > 7: rate_str = f"+{rate_percentage + 5}%" async def _async_gen(): communicate = edge_tts.Communicate(text, voice_id, rate=rate_str) await communicate.save(filename) try: loop = asyncio.get_event_loop() except RuntimeError: loop = asyncio.new_event_loop() asyncio.set_event_loop(loop) if loop.is_running(): import threading t = threading.Thread(target=lambda: asyncio.run(_async_gen())) t.start() t.join() else: loop.run_until_complete(_async_gen()) # ===================================================================== # 🎬 HIGH-SPEED PRODUCTION AUTOMATION ENGINE # ===================================================================== def process_magic_recap_video( video_path, user_api_key, ratio_select, background_fill, enable_zoom, zoom_level, logo_file, mirror_flip, filter_color, voice_gender, tone_style, text_color, stroke_color, blur_y_percent, blur_strength, sub_pos_percent, desired_speed, session_id, # 👈 Browser Session State request: gr.Request, progress=gr.Progress(track_tqdm=True) ): is_valid, msg = check_app_expiry() if not is_valid: raise gr.Error(msg) if not video_path: return None # 🛑 LOGIN ACCOUNT & SESSION ID ယူခြင်း login_username = request.username if (request and hasattr(request, 'username')) else "user1" if login_username == "admin": user_identifier = "admin" else: # Browser Session ID ဖြင့် လူတစ်ဦးစီကို သီးခြား ခွဲခြားမည် user_identifier = f"session_{session_id}" current_timestamp = time.time() print(f"👤 အသုံးပြုသူ Account: {login_username} | Identifier: {user_identifier}") # Admin မဟုတ်ပါက Session/Browser အလိုက် ၁ ရက် (၂) ပုဒ် Limit စစ်ဆေးမည် if user_identifier != "admin": if user_identifier in USER_LIMIT_TRACKER: user_data = USER_LIMIT_TRACKER[user_identifier] first_time = user_data["first_time"] count = user_data["count"] elapsed_time = current_timestamp - first_time if elapsed_time < 86400: # ၂၄ နာရီ အတွင်း if count >= 90: # ၂ ပုဒ် ပြည့်ပါက တားဆီးမည် remaining_seconds = 86400 - elapsed_time rem_hours = int(remaining_seconds // 3600) rem_mins = int((remaining_seconds % 3600) // 60) raise gr.Error(f"❌ ခွင့်ပြုချက်ကျော်လွန်နေပါသည်။ သင့် Browser မှ တစ်ရက်လျှင် ဗီဒီယို (၂) ပုဒ်သာ ထုတ်ယူခွင့်ရှိသည်။ ပြန်လည်စမ်းသပ်ရန် {rem_hours} နာရီ {rem_mins} မိနစ် လိုအပ်ပါသေးသည်။") else: USER_LIMIT_TRACKER[user_identifier] = {"count": 0, "first_time": current_timestamp} else: USER_LIMIT_TRACKER[user_identifier] = {"count": 0, "first_time": current_timestamp} temp_dir = "temp_space_workspace" if os.path.exists(temp_dir): shutil.rmtree(temp_dir) os.makedirs(temp_dir, exist_ok=True) try: progress(0.10, desc="🎙️ Faster-Whisper ဖြင့် စာသားဖတ်နေပါသည်...") segments_raw, info = model.transcribe(video_path, beam_size=1) raw_segments = [{"start": seg.start, "end": seg.end, "text": seg.text} for seg in segments_raw] cap = cv2.VideoCapture(video_path) fps = cap.get(cv2.CAP_PROP_FPS) or 30.0 orig_w = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)) orig_h = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)) video_duration = int(cap.get(cv2.CAP_PROP_FRAME_COUNT)) / fps if not raw_segments: raw_segments = [{'start': 0.0, 'end': min(6.0, video_duration), 'text': "Welcome to this movie recap."}] segments = [] for seg in raw_segments: s_start = seg['start'] s_end = seg['end'] s_text = seg['text'].strip() if not s_text: continue dur = s_end - s_start if dur > MAX_SUBTITLE_DURATION_SECONDS and MAX_SUBTITLE_DURATION_SECONDS > 0: words = s_text.split() chunks_count = int(np.ceil(dur / MAX_SUBTITLE_DURATION_SECONDS)) words_per_chunk = int(np.ceil(len(words) / chunks_count)) for i in range(chunks_count): w_sub = words[i*words_per_chunk : (i+1)*words_per_chunk] if not w_sub: continue segments.append({'start': s_start + (i * (dur / chunks_count)), 'end': min(s_end, s_start + ((i+1) * (dur / chunks_count))), 'text': " ".join(w_sub)}) else: segments.append({'start': s_start, 'end': s_end, 'text': s_text}) detected_lang = info.language if user_api_key and user_api_key.strip(): progress(0.30, desc=f"⚡ ထည့်သွင်းထားသော Gemini API ({detected_lang}) စနစ်ဖြင့် ဘာသာပြန်နေပါသည်...") else: progress(0.30, desc=f"🌐 Google Translate Backup ({detected_lang}) စနစ်ဖြင့် ဘာသာပြန်နေပါသည်...") segments = translate_segments_batch(segments, user_api_key, source_lang=detected_lang) if ratio_select == "9:16 (Tiktok/Reels)": target_w, target_h = 720, 1280 else: target_w, target_h = 1280, 720 logo_img = None if logo_file: try: logo_cv = cv2.imread(logo_file.name, cv2.IMREAD_UNCHANGED) if logo_cv is not None: l_w = int(target_w * 0.16) logo_img = cv2.resize(logo_cv, (l_w, int(l_w * (logo_cv.shape[0] / logo_cv.shape[1])))) except: pass progress(0.50, desc="🎙️ AI အသံများ စတင်ဖန်တီးနေပါသည်...") audio_segments = [] python_srt_segments = [] voice_id = "my-MM-NilarNeural" if "မိန်းကလေး" in voice_gender else "my-MM-ThihaNeural" v_segments_time_map = [] total_adjusted_duration = 0.0 for idx, seg in enumerate(segments): mm_text = seg.get('mm_text', seg['text']) mm_text = mm_text.replace(" ြ", "ြ").replace("ြ ", "ြ").strip() orig_start, orig_end = float(seg.get('start', 0.0)), float(seg.get('end', 0.0)) orig_dur = orig_end - orig_start if (orig_end - orig_start) > 0 else 2.0 raw_seg_filename = os.path.join(temp_dir, f"raw_{idx}.mp3") fixed_seg_filename = os.path.join(temp_dir, f"fixed_{idx}.mp3") generate_voice_sync(mm_text, voice_id, raw_seg_filename, 1.15, orig_dur) if os.path.exists(raw_seg_filename) and os.path.getsize(raw_seg_filename) > 0: subprocess.run([ 'ffmpeg', '-y', '-i', raw_seg_filename, '-filter:a', f"atempo={desired_speed}", fixed_seg_filename ], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL) if os.path.exists(fixed_seg_filename) and os.path.getsize(fixed_seg_filename) > 0: probe_res = subprocess.run(['ffprobe', '-v', 'error', '-show_entries', 'format=duration', '-of', 'default=noprint_wrappers=1:nokey=1', fixed_seg_filename], stdout=subprocess.PIPE, text=True) try: audio_dur = float(probe_res.stdout.strip()) except: audio_dur = orig_dur / desired_speed python_srt_segments.append({'start': total_adjusted_duration, 'end': total_adjusted_duration + audio_dur, 'text': mm_text}) audio_segments.append(fixed_seg_filename) v_segments_time_map.append({'orig_start': orig_start, 'orig_end': orig_end, 'new_start': total_adjusted_duration, 'new_end': total_adjusted_duration + audio_dur, 'pts_ratio': audio_dur / orig_dur}) total_adjusted_duration += audio_dur progress(0.70, desc="⚡ Render ဗီဒီယိုနှင့် စာတန်းထိုးများ ပေါင်းစပ်နေပါသည်...") output_video_path = os.path.abspath("magic_recap_output.mp4") if os.path.exists(output_video_path): os.remove(output_video_path) final_burn_temp = os.path.join(temp_dir, "final_burn_temp.mp4") video_writer = cv2.VideoWriter(final_burn_temp, cv2.VideoWriter_fourcc(*'mp4v'), fps, (target_w, target_h)) font_size = int(target_h * 0.038) font_path = "Myanmar font.ttf" font_primary = ImageFont.truetype(font_path, font_size) if os.path.exists(font_path) else ImageFont.load_default() t_color, s_color = hex_to_rgb(text_color), hex_to_rgb(stroke_color) b_h = int(target_h * 0.12) b_y = max(0, min(int(target_h * (blur_y_percent / 100)) - (b_h // 2), target_h - b_h)) m_w, s_w = int(target_w * 0.90), max(2, int(font_size * 0.12)) k_size = int(blur_strength) | 1 total_output_frames = int(total_adjusted_duration * fps) for f_out_idx in range(total_output_frames): c_sec = f_out_idx / fps target_orig_sec = 0.0 for mapping in v_segments_time_map: if mapping['new_start'] <= c_sec <= mapping['new_end']: target_orig_sec = mapping['orig_start'] + ((c_sec - mapping['new_start']) / mapping['pts_ratio']) break else: if v_segments_time_map: target_orig_sec = v_segments_time_map[-1]['orig_end'] target_frame_idx = int(target_orig_sec * fps) cap.set(cv2.CAP_PROP_POS_FRAMES, target_frame_idx) ret, orig_frame = cap.read() if not ret or orig_frame is None: orig_frame = np.zeros((orig_h, orig_w, 3), dtype=np.uint8) if background_fill == "Blur Background (အနောက်ခံ ဝါးမည်)": small_bg = cv2.resize(orig_frame, (target_w // 4, target_h // 4), interpolation=cv2.INTER_LINEAR) blurred_small_bg = cv2.blur(small_bg, (11, 11)) bg_layer = cv2.resize(blurred_small_bg, (target_w, target_h), interpolation=cv2.INTER_LINEAR) else: bg_layer = np.zeros((target_h, target_w, 3), dtype=np.uint8) if enable_zoom and zoom_level > 1.0: fg_cropped = orig_frame[int((orig_h - orig_h/zoom_level)//2):int((orig_h + orig_h/zoom_level)//2), int((orig_w - orig_w/zoom_level)//2):int((orig_w + orig_w/zoom_level)//2)] else: fg_cropped = orig_frame fg_w = target_w fg_h = int(fg_w / (fg_cropped.shape[1] / fg_cropped.shape[0])) if fg_h > target_h: fg_h = target_h fg_w = int(fg_h * (fg_cropped.shape[1] / fg_cropped.shape[0])) fg_resized = cv2.flip(cv2.resize(fg_cropped, (fg_w, fg_h)), 1) if mirror_flip else cv2.resize(fg_cropped, (fg_w, fg_h)) if filter_color == "Chrome Cool": fg_resized = cv2.convertScaleAbs(fg_resized, alpha=1.0, beta=15) elif filter_color == "Warm Cinema": fg_resized = cv2.convertScaleAbs(fg_resized, alpha=1.05, beta=5) else: fg_resized = cv2.convertScaleAbs(fg_resized, alpha=0.99, beta=2) bg_layer[(target_h - fg_h)//2:(target_h - fg_h)//2+fg_h, (target_w - fg_w)//2:(target_w - fg_w)//2+fg_w] = fg_resized frame = bg_layer if logo_img is not None: ly, lx = 25, target_w - logo_img.shape[1] - 25 if logo_img.shape[2] == 4: alpha_l = logo_img[:, :, 3] / 255.0 for c in range(3): frame[ly:ly+logo_img.shape[0], lx:lx+logo_img.shape[1], c] = alpha_l * logo_img[:, :, c] + (1.0 - alpha_l) * frame[ly:ly+logo_img.shape[0], lx:lx+logo_img.shape[1], c] else: frame[ly:ly+logo_img.shape[0], lx:lx+logo_img.shape[1]] = logo_img[:, :, :3] if b_h > 0 and (b_y + b_h) <= target_h: roi = frame[b_y:b_y+b_h, 0:target_w] if roi.shape[0] > 0 and roi.shape[1] > 0: roi_small = cv2.resize(roi, (target_w // 4, b_h // 4), interpolation=cv2.INTER_LINEAR) roi_blur = cv2.GaussianBlur(roi_small, (k_size // 4 | 1, k_size // 4 | 1), 0) frame[b_y:b_y+b_h, 0:target_w] = cv2.resize(roi_blur, (target_w, b_h), interpolation=cv2.INTER_LINEAR) text_str = "" for s in python_srt_segments: if s['start'] <= c_sec <= s['end']: text_str = s['text'] break if text_str and isinstance(font_primary, ImageFont.FreeTypeFont): pil_img = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)) draw = ImageDraw.Draw(pil_img) sub_lines = wrap_text_myanmar_smart(text_str, font_primary, m_w, draw) total_text_height = sum([draw.textbbox((0, 0), l, font=font_primary)[3] - draw.textbbox((0, 0), l, font=font_primary)[1] for l in sub_lines]) curr_y = int(target_h - total_text_height - (target_h * (sub_pos_percent / 100))) for line in sub_lines: clean_line = line.replace(" ြ", "ြ").replace("ြ ", "ြ") bbox = draw.textbbox((0, 0), clean_line, font=font_primary) draw_line_perfect_rendering(draw, ((target_w - (bbox[2] - bbox[0])) // 2, curr_y), clean_line, font_primary, t_color, s_w, s_color) curr_y += (bbox[3] - bbox[1]) + 10 frame = cv2.cvtColor(np.array(pil_img), cv2.COLOR_RGB2BGR) video_writer.write(frame) cap.release() video_writer.release() merged_audio_path = os.path.join(temp_dir, "final_speech_track.mp3") if audio_segments: inputs_cmd = [] for idx, audio_file in enumerate(audio_segments): inputs_cmd.extend(['-i', audio_file]) subprocess.run(['ffmpeg', '-y'] + inputs_cmd + ['-filter_complex', f"concat=n={len(audio_segments)}:v=0:a=1[outa]", '-map', '[outa]', '-c:a', 'libmp3lame', '-b:a', '192k', merged_audio_path], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL) has_audio = os.path.exists(merged_audio_path) and os.path.getsize(merged_audio_path) > 0 else: has_audio = False progress(0.90, desc="⚡ ရုပ်သံနှင့် အသံလှိုင်းများကို အပြီးသတ် ပေါင်းစပ်နေပါသည်...") if has_audio: subprocess.run(['ffmpeg', '-y', '-i', final_burn_temp, '-i', merged_audio_path, '-c:v', 'libx264', '-preset', 'ultrafast', '-threads', '0', '-pix_fmt', 'yuv420p', '-c:a', 'aac', '-b:a', '192k', '-map', '0:v:0', '-map', '1:a:0', '-shortest', output_video_path], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL) else: subprocess.run(['ffmpeg', '-y', '-i', final_burn_temp, '-c:v', 'libx264', '-preset', 'ultrafast', '-threads', '0', '-pix_fmt', 'yuv420p', output_video_path], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL) if os.path.exists(temp_dir): shutil.rmtree(temp_dir) # 📈 ဗီဒီယို အောင်မြင်စွာ ထုတ်ယူပြီးမှသာ ထို Session ၏ Count ကို ၁ တိုးမည် if user_identifier != "admin": USER_LIMIT_TRACKER[user_identifier]["count"] += 1 print(f"✅ [{user_identifier}] ဗီဒီယိုထုတ်ယူမှု အောင်မြင်ပါသည်။ ယနေ့ ထုတ်ယူပြီးစီးမှု: {USER_LIMIT_TRACKER[user_identifier]['count']} ကြိမ်") return output_video_path except Exception as e: if os.path.exists(temp_dir): shutil.rmtree(temp_dir) raise gr.Error(f"❌ အမှားအယွင်း တစ်ခု ဖြစ်ပွားခဲ့သည်- {str(e)}") # ===================================================================== # 🎨 GRADIO INTERFACE # ===================================================================== def create_main_app_interface(): is_valid, msg = check_app_expiry() with gr.Blocks() as main_app: # 🔑 အသုံးပြုသူတိုင်းအတွက် သီးသန့် Session ID ထုတ်ပေးခြင်း session_id_state = gr.State(lambda: str(uuid.uuid4())) gr.Markdown("

✨ Video Auto Recap (Multi-Account) ✨

") if not is_valid: gr.Markdown(f"### {msg}") return main_app gr.Markdown("
📢 Free App Rules: တစ်ရက်လျှင် Video (၂) ပုဒ် အခမဲ့ ထုတ်ယူခွင့် ရှိပါသည်
") with gr.Row(): with gr.Column(scale=2): video_input = gr.Video(label="🎥 ဗီဒီယို ထည့်ရန်", sources=["upload"]) preview_image = gr.Image(label="🖼️ Interactive Blur Preview", interactive=False) with gr.Column(scale=1): user_api_key = gr.Textbox(label="🔑 သင်၏ Gemini API Key ထည့်ရန် (မထည့်ပါက Google Translate ဖြင့် လုပ်ဆောင်မည်)", placeholder="AIZAcy... (Optional)", type="password") ratio_select = gr.Dropdown(choices=["9:16 (Tiktok/Reels)", "16:9 (Landscape)"], value="9:16 (Tiktok/Reels)", label="ဗီဒီယိုအမျိုးအစား") background_fill = gr.Radio(choices=["Blur Background (အနောက်ခံ ဝါးမည်)", "Black Background (အမည်းရောင်ထားမည်)"], value="Blur Background (အနောက်ခံ ဝါးမည်)", label="နောက်ခံ ဖြည့်စွက်မှု") with gr.Column(variant="panel"): gr.Markdown("### 🗣️ အသံနှင့် စာတန်းထိုး ဆက်တင်များ") voice_select = gr.Radio(choices=["🧕 မိန်းကလေး (Female)", "👨 ယောကျာ်လေး (Male)"], value="🧕 မိန်းကလေး (Female)", label="AI အသံ အမျိုးအစား") tone_style = gr.Dropdown(choices=["Thriller", "Comedy", "Dramatic", "Action/Epic"], value="Thriller", label="🎬 Narrative Tone") text_color_input = gr.ColorPicker(label="စာလုံးအရောင်", value="#FFFF00") stroke_color_input = gr.ColorPicker(label="အနားသတ်အရောင်", value="#000000") sub_pos_percent = gr.Slider(minimum=0, maximum=100, value=15, step=1, label="မြန်မာစာတန်း တည်နေရာ %") blur_y_percent = gr.Slider(minimum=50, maximum=100, value=75, step=1, label="📍 မူရင်းစာတန်းဖျောက်မည့်နေရာ %") blur_strength = gr.Slider(minimum=5, maximum=151, value=51, step=2, label="🌫️ မူရင်းစာတန်း ဝါးမည့်ပမာဏ") desired_speed = gr.Slider(minimum=1.0, maximum=1.6, value=1.35, step=0.05, label="🎙️🎬 အသံနှင့် ဗီဒီယို အရှိန်မြှင့်နှုန်း") with gr.Accordion("⚙️ အဆင့်မြင့် ဆက်တင်များ", open=False): enable_zoom = gr.Checkbox(label="Zoom & Crop သုံးရန်", value=False) zoom_level = gr.Slider(minimum=1.0, maximum=3.0, value=1.0, step=0.1, label="Zoom Level") logo_file = gr.File(label="လိုဂို ထည့်ရန် (Optional)") mirror_flip = gr.Checkbox(label="ဘယ်ညာ ပြောင်းရန် (Mandatory Auto-Active)", value=True) filter_color = gr.Dropdown(choices=["None (အလိုအလျောက်ကုဒ်ပြောင်းမည်)", "Chrome Cool", "Warm Cinema"], value="None (အလိုအလျောက်ကုဒ်ပြောင်းမည်)", label="ဗီဒီယို Filter") submit_btn = gr.Button("🚀 Generate Video", variant="primary") with gr.Column(variant="panel"): output_video = gr.Video(label="✅ ပြီးပြည့်စုံသော ဗီဒီယို") video_input.change(fn=update_preview_image, inputs=[video_input, blur_y_percent, blur_strength], outputs=preview_image) blur_y_percent.change(fn=update_preview_image, inputs=[video_input, blur_y_percent, blur_strength], outputs=preview_image) blur_strength.change(fn=update_preview_image, inputs=[video_input, blur_y_percent, blur_strength], outputs=preview_image) submit_btn.click( fn=process_magic_recap_video, inputs=[ video_input, user_api_key, ratio_select, background_fill, enable_zoom, zoom_level, logo_file, mirror_flip, filter_color, voice_select, tone_style, text_color_input, stroke_color_input, blur_y_percent, blur_strength, sub_pos_percent, desired_speed, session_id_state # 👈 session_id ကို Function ထဲ ပို့ပေးခြင်း ], outputs=[output_video] ) return main_app def verify_member(username, password): if username in ACCOUNTS and ACCOUNTS[username] == password: return True return False demo_app = create_main_app_interface() if __name__ == "__main__": demo_app.launch(auth=verify_member, theme=gr.themes.Default())