Automatically keep trying | ComfyUI workflow — v1.1

This ComfyUI workflow is for repeatedly generating images until BLIP says your conditions are met; it uses a selected model, prompts, natural-language questions, and answers, and produces generated images plus BLIP answers.

The workflow repeats image generation while BLIP checks whether your conditions are met, and it includes LoopManager.

The usage steps call for enabling Auto Queue, pressing Queue Prompt, and enabling Auto Queue again after success.

You can enter questions in natural language, including questions about facial expression or whether an image is classical art.

BLIP answers may appear as forms such as "1" or "one"; the console shows each answer.

Install WAS Node Suite and Cyclist custom nodes, then download and open the workflow.

Suggestion · not verified

The workflow includes the checkpoint file aZovyaRPGArtistTools_v4.safetensors.

The listed node names are BLIP Analyze Image, BLIP Model Loader, CheckpointLoaderSimple, CLIPTextEncode, CyclistCompare, EmptyImage, EmptyLatentImage, Interrupt, KSampler, Logic Comparison AND, LoopManager, OverrideImage, ReloadImage, Text String, and VAEDecode.

Select the model and prompts, then set the questions and answers BLIP should use for the conditions.

Suggestion · not verified

In ComfyUI, enable Extra Options and Auto Queue, press Queue Prompt, and enable Auto Queue again after success.

Suggestion · not verified

The base model is SD 1.5.

BLIP is not human, and its answers are not 100% correct.

Because answers can use forms such as "1" or "one," check the console when interpreting them.

Suggestion · not verified