ComfyDeploy: How ComfyUI_Anytext works in ComfyUI?

What is ComfyUI_Anytext?

Unofficial Simple And Rough Implementation Of [a/AnyText](https://github.com/tyxsspa/AnyText) and [a/Glyph-ByT5] (https://github.com/AIGText/Glyph-ByT5) and [a/JoyType](https://github.com/jdh-algo/JoyType)

How to install it in ComfyDeploy?

Head over to the machine page

  1. Click on the "Create a new machine" button
  2. Select the Edit build steps
  3. Add a new step -> Custom Node
  4. Search for ComfyUI_Anytext and select it
  5. Close the build step dialig and then click on the "Save" button to rebuild the machine

Look at the node tooltip for instruction.

1、AnyText: Unofficial custom_node for AnyText: Multilingual Visual Text Generation And Editing

Option 1: full AnyText checkpoint.

Huggingface:

Modelscope:

Workflow in workflow dir:

Option 2: any sd1.5 checkpoint with AnyText control model.

Huggingface:

Workflow in workflow dir:

If want chinese prompt without translator, use MiaoBi clip.

Huggingface:

Workflow in workflow dir:

Font: select Auto_Download from huggingface or manual download from huggingface or modelscope or use any other fonts in ComfyUI\models\fonts.

Huggingface:

Modelscope:

Zh to En Translators:

T5: Auto_Download from huggingface or manual download or git clone from huggingface into ComfyUI\models\prompt_generator\models--utrobinmv--t5_translate_en_ru_zh_base_200.

nlp_csanmt_translation_zh2en: Auto_Download from modelscope or manual download or git clone from modelscope into ComfyUI\models\prompt_generator\modelscope--damo--nlp_csanmt_translation_zh2en.

2、Glyph-ByT5: Unofficial custom_node for Glyph-ByT5: A Customized Text Encoder for Accurate Visual Text Rendering

Limitations: background result worse than original SDXL checkpoint, image resolution locked at 1024x1024, input mask and image aspect ratio must be 1:1.

Place Glyph-ByT5 checkpoints in ComfyUI\custom_nodes\ComfyUI_Anytext\Image_Generation\Glyph_SDXL\checkpoints

Place google/byt5-small text_encoder into ComfyUI\models\text_encoders or select Auto_Download from huggingface.

For sdxl checkpoints, download original_config_yaml and inpaint_original_config_yaml into ComfyUI\models\configs.

Workflow in workflow dir:

3、JoyType: Unofficial custom_node for JoyType: A Robust Design for Multilingual Visual Text Creation

It's a typical canny controlnet implementation by diffusers library.

Comfy sample workflow recommended, sd1.5、sd2.1、sdxl canny works.

For JoyType controlnet, download from https://huggingface.co/jdh-algo/JoyType-v1-1M into ComfyUI\models\controlnet. It's a diffusers format canny controlnet, compitable with comfy sampler.

Workflow in workflow dir: