Customization and Extensibility

Contents

Editing and Adding Modes

All of the modes can be edited in the config.yaml file.

Adding a new mode is as simple as adding it in the config. To use it when generating add m=<name of your mode> to the end of your prompt. To change the defaults create a default mode.

Mode configuration options Default
cfg Guidance scale 3.5
steps Number of sampling steps 20
lora Name of LORA to use in your ComfyUI models/loras folder. No LORA
lora_strength Strength of LORA effect 1.0
sampler Sampling method to use "euler"
scheduler Scheduler to use "simple"
prompt_template_pre_pe Template for prompt before enhancement. {} is replaced with the prompt before prompt enhancement. "{}"
prompt_template_post_pe Template for prompt after enhancement. {} is replaced with the prompt after prompt enhancement. "{}"
description Description of the mode No description

Changing the Model and editing the Workflow

This project works with only one workflow. This limits it to UNET models.

If you want to use a different workflow or model you will have to change some code.

To make the project work with a custom workflow edit the _prepare_workflow method in the comfyui_telegram_bot/image_gen.py file. Edit the method to match the node IDs and node attributes in your workflow file.

To export a workflow from ComfyUI in the correct format:

  1. Open your workflow in the ComfyUI web interface.
  2. Go to Settings (gear icon in the right vertical menu) -> Dev Mode -> Enable dev mode options -> ON
  3. Close settings
  4. Click the Save (API Format) button in the right vertical menu.

Prompt Enhancement

Prompt Enhancement is fully customizeable. This means that you can change the system prompts for the LLMs and you can also change what LLM you want to use.

System Prompts

All of the system prompts can be edited in the config.yaml file.

Adding a new prompt enhancement 'types' is as simple as adding it in the config. To use it when generating add pe=<name of your prompt enhancement type> to the end of your prompt.

Prompt enhancement type configuration options
description Description of the prompt enhancement type
system_prompt System Prompt of the LLM. Important: The LLM has to be instructed to return the enhanced prompt in <prompt></prompt> tags!

Services

There is currently only one prompt enhancement service implemented and that is for the Anthropic API.

Adding more is as simple as creating a new Python script in the comfyui_telegram_bot/services folder with the name of your service. Then you write a class extending the PromptEnhanceService class and implementing the enhance_prompt method.

For reference on how this can be done, you can read the comfyui_telegram_bot/services/anthropic.py file.

Contributing

I welcome all contributions. Simply create a pull request on the GitHub repository and I will review it.

Ideas for Improvements: