Revert to GLM-4.6V (stable)

README.md

@@ -9,7 +9,7 @@ app_file: app.py
 pinned: false
 hardware: t4-small
 tags:
-  - qwen2-vl
+  - glm-4.6v
   - lama
   - manga
   - comic
@@ -22,16 +22,16 @@ tags:
   - inpainting
   - speech-bubble
   - webtoon
-short_description: AI manga & comic translator powered by Qwen2-VL + LaMa
+short_description: AI manga & comic translator powered by GLM-4.6V + LaMa
 ---
 
 # ✍️ BubbleScribe
 
-Translate manga and comics from Japanese/Korean/Chinese to any language using Qwen2-VL for OCR and LaMa for inpainting.
+Translate manga and comics from Japanese/Korean/Chinese to any language using GLM-4.6V for OCR and LaMa for inpainting.
 
 ## Features
 
-- 🔍 **Smart Text Detection** - Qwen2-VL detects speech bubbles and text regions
+- 🔍 **Smart Text Detection** - GLM-4.6V detects speech bubbles and text regions
 - 🌐 **Multi-language Translation** - Japanese, Korean, Chinese → Any language
 - 🎨 **Clean Inpainting** - LaMa removes original text with GPU acceleration
 - ✍️ **Text Overlay** - Adds translated text back to the image
@@ -39,17 +39,17 @@ Translate manga and comics from Japanese/Korean/Chinese to any language using Qw
 ## How It Works
 
 1. Upload a manga/comic page
-2. Qwen2-VL detects and translates text
+2. GLM-4.6V detects and translates text
 3. LaMa inpainting removes original text (GPU-accelerated)
 4. Translated text is overlaid on the clean image
 
 ## Models Used
 
-- **Qwen2-VL-7B** (HuggingFace Inference API) - OCR + Translation
+- **GLM-4.6V** (Z.ai API) - OCR + Translation
 - **LaMa** - Text removal inpainting (GPU-accelerated)
 
 ## Credits
 
-- [Qwen2-VL](https://huggingface.co/Qwen/Qwen2-VL-7B-Instruct) by Alibaba
+- [GLM-4.6V](https://huggingface.co/zai-org/GLM-4.6V) by Zhipu AI
 - [LaMa](https://github.com/advimman/lama) for inpainting
 - Demo by [@lulavc](https://huggingface.co/lulavc)