[Patch v11.7] Notch filters (50/100/150Hz) for electrical hum removal + enhanced denoise (nr=20)
[Patch v11.6] Echo removal (afftdn om=f) + smart cleaning + single ASR + mandatory sample_text
[Patch v8.3] urlopen(): UTF-8 safe wrapper applied
[Patch v9.1] urlopen(): debug logging added for GPT-SoVITS requests
[Patch v0.6.1] _save_json() and _save_json_nolock(): atomic write with backup
[Patch v11.6] train_voice(): echo removal + sample_text storage + single ASR
[Patch v11.8] train(): COMPLETELY disable auto voice generation during training
[Patch v10.2] synthesize(): direct GPT-SoVITS API call (UTF-8 safe)
[Patch v11.6] train_voice_with_type(): echo cleaning + sample_text/ref_text extraction
[Patch v8.0] do_POST(): reseed + voice_type extraction
[Patch v8.3] _do_generate(): lip_sync_mode routed via engine._lip_sync_mode
[Patch v8.0] _handle_voice_preview(): base64 inline + 100% cloning
[Patch v8.0] _handle_voice_synthesize(): 100% cloning enabled
[Patch v8.1] _transcribe_audio(): safe wrapper with UTF-8 error handling
[Patch v11.6] All patches applied successfully!
[Patch v9.1] _load_voices(): debug logging added
[Patch v8.5] Original directory writable: D:\AI-Ad-Agent\digital_human_v2, keeping native paths
[Patch v9.0] check_available(): TCP socket check + HTTP fallback
[Patch v10] Running voice file integrity check...
[Patch v10] All 3 voices pass integrity check
[Patch v11.6] Digital human training: no auto voice cloning (use voice library instead)
[Patch v8.5] Starting server...
[Patch v8.5] Starting server...
[服务] 正方元智能广告工作台 - 数字人 v2 (v8.4)
[服务] 地址: http://192.168.5.250:8766