feat: add backend dubbing and translation services for audio processing

services/dubbing_service.py

@@ -445,7 +445,7 @@ class VideoDubbingService:
 
                         # 2. Smooth pitch-preserving speed adjustment capped at max 1.15x to preserve natural cadence
                         if seg_duration > max_allowed_duration:
-                            speed_factor = max(1.0, min(1.15, seg_duration / max_allowed_duration))
+                            speed_factor = max(1.0, min(1.35, seg_duration / max_allowed_duration))
                             speed_path = os.path.join(temp_dir, f"seg_{idx}_speed.mp3")
                             adjusted_path = adjust_audio_speed_ffmpeg(seg_audio_path, speed_path, speed_factor)
                             if os.path.exists(adjusted_path) and os.path.getsize(adjusted_path) > 0 and adjusted_path != seg_audio_path:
@@ -458,7 +458,7 @@ class VideoDubbingService:
 
                         # 3. STRICT NON-OVERLAPPING SEQUENTIAL ANCHORING:
                         target_start_ms = int(orig_start * 1000)
-                        play_pos_ms = max(target_start_ms, last_end_ms + 40) if last_end_ms > 0 else target_start_ms
+                        play_pos_ms = max(target_start_ms, last_end_ms + 10) if last_end_ms > 0 else target_start_ms
 
                         needed_duration_ms = play_pos_ms + len(seg_audio)
                         if len(master_audio) < needed_duration_ms:

services/translation_service.py

@@ -216,7 +216,7 @@ class TranslationService:
             is_large_gap = False
             if idx < len(segments) - 1:
                 next_start = segments[idx + 1].get("start", 0)
-                if next_start - seg.get("end", 0) > 0.8:
+                if next_start - seg.get("end", 0) > 1.2:
                     is_large_gap = True
 
             if (is_sentence_end and current_duration >= 2.5) or current_duration >= 4.5 or is_large_gap or len(current_group) >= 8 or idx == len(segments) - 1: