feat: add heavy GPU dependencies and Wav2Lip auto-downloader for Colab deployment

feat: unify React Dubbing Studio UI with Gradio Manga AI in single Docker container

remove: delete 'Soft & Seductive (Maya)' voice tone from all dropdowns

fix: replace localhost iframe in dubbing tab with native Gradio UI that works on HF Spaces

revert: switch HF Space back to Gradio SDK running app.py (Maya Immersive Tamil Manga AI)

fix: remove root GET route blocking React frontend on HF Spaces

fix: upgrade Node.js 18 to 20 to satisfy Vite 8 build requirements

fix: switch HF Space SDK from gradio to docker to use unified Dockerfile

fix: ensure static files are mounted last to prevent API route interception

fix: unified Docker deployment for Hugging Face to resolve local network connection errors

feat: further refine Tanglish vocabulary and common spoken verb forms

feat: enhance translation with colloquial (spoken) Tamil style

feat: implement detailed pipeline diagnostics, memory optimization for transcription, and fix import issues

Implement Cinematic Panel Narration with TTS splicing

Implement 2-stage PaddleOCR + Mistral pipeline

Update OCR prompt for comic books

OCR: Integrated Master Comic AI Instruction Set for panel-aware dialogue extraction

OCR: Upgrade to High-Fidelity Junk Filtering and Multi-Stage Vision Scanning to eliminate gibberish

UI: Fix audio player timeline and timestamp visibility for better clarity

Fix: Remove JS pre-processor and add global auto-play shield to isolate final UI Error cause

Privacy: Remove Private Gallery and Admin Monitoring sections as requested

UI: Remove Maya Chat column as requested and stabilize TTS audio buffer format

Fix: Bypass Gradio 4 frontend Image deserialization bug by reading the PDF page directly in the backend process

Fix: Resolve Gradio 4 JS payload corruption causing UI Error states by ensuring triggerHaptic returns input arguments

Fix: Add global try-except shield to completely prevent Gradio UI crashes and expose hidden process errors

Fix: Resolve 4 critical hidden exceptions causing Gradio UI 'Error' boxes (Qwen timeout, Paddle kwargs, TTS empty strings, Autoplay TypeErrors)

Feature: Integrate Qwen2.5-VL as the Ultimate Primary OCR Vision Engine

Feature: Integrate Tesseract as the ultimate tertiary OCR engine for maximum text extraction power

Fix: Implement Dual-OCR fallback and UI error hardening for stable comic reading

Fix: Implement 'Neural Silencer' to permanently stop asyncio log spam

Build: Force server refresh with final stabilized Maya AI engine (v1.0.1)

Ultimate: Definite AI Intelligence and UI stability fix (Phi-3 Integration)

Fix: Ultimate stabilization of asyncio event loop for clean server logs

Stability: Deploy Qwen-2.5-1.5B with Smart Context Fallback for Maya

Final: Stabilize Maya AI with non-streaming direct responses and 10s timeout

Stability: Switch to Llama-3.2-1B for guaranteed AI responses and expanded Tanglish fallback

Fix: Enable high-priority AI brain with Qwen-2.5 and HF_TOKEN support

Feature: Add 'Share with Friends' button with clipboard sync

Fix: Stabilize asyncio event loop in TTS to avoid Invalid File Descriptor errors

Language: Switch Maya's entire conversation engine to Tanglish (Latinized Tamil)

Hardening: Full Uncensored Bypass and Lustful Fallback for Maya Chatbot

Upgrade: Transform Maya into a true LLM-powered Chatbot (Llama-3 Integration)

UI: Remove visual GIF gallery for a cleaner interface

Sync: Full project update including premium UI and advanced chat engine

Stability: Activate Ultimate Stability Mode to resolve server hangs and memory issues

Optimization: Restore 5-page background prefetch for zero-latency reading

Optimization: Strip down heavy UI effects and disable experimental SSR for maximum performance

Optimization: Implement aggressive cache cleanup and reduce prefetch for low-memory environments

Fix: Resolve Gradio 6.0 API mismatches for Chatbot and CSS

Feature: Add dedicated 'Chat with Maya' tab with full-screen chatbot interface