recode-lh

>  按现有文档，直接在环境变量中添加一个模型名称就可以，翻译的时候翻译器选择custom_openai。或者选openai也行，把`custom_`去掉，值也这么写 为什么设置了ollama翻译的时候一直报超时····

> 信息太少看不出原因x 您好，完整信息如下 我发现时是 get_hyphenation()的时候很慢，render设置no_hyphenation=true好像没有作用呢 参数： { "detector": { "detector": "default" }, "render": { "no_hyphenation":true, "direction": "auto", "font_size_offset": 0 }, "translator": { "translator": "custom_openai", "target_lang": "ENG" }, "inpainter": { "inpainter":...

> 我也是，用得README里面的图片看看效果，但是卡在rendering这一步；查看显卡内存确实有占用，应该使用了显卡。 直接使用得 python -m manga_translator local -v -i sourceTranslate.png > > [local] Namespace(verbose=True, attempts=0, ignore_errors=False, model_dir=None, use_gpu=True, use_gpu_limited=False, font_path='', pre_dict=None, post_dict=None, kernel_size=3, context_size=0, batch_size=1, batch_concurrent=False, disable_memory_optimization=False, mode='local', input=['sourceTranslate.png'], dest='',...

def select_hyphenator(lang: str): rerurn None

> RAGFlow 没有做任何语言翻译。除了调整提示外，还转到另一个 LLM。 Hello, how to solve the problem of language confusion

> I guess tuning the prompt might solve the issue. On the prompt words (on the web interface), I clearly told him that he needed to answer in Chinese, including...