I get different results with automatic1111 and cagliostro even if I set the same prompt, seed, size, step, and CFGscale. Do you know what is causing this?

[wainowaino]

Even with the same prompt, seed, size, step, and CFGscale, automatic1111 and cagliostro output different results.

I am seeing that perhaps some setting is turned on by default, but I can't figure out what it is... I found the "clip skip" item and tried setting it to 1, but there was no change and the output is still different.

The problem with this is that the picture style has changed.

I am trying to make the prompts consistent to some extent, but all my previous prompts are going to waste, and the same model looks like a different model. Please let me know if you know the cause of this problem.

Thank you.

[Linaqruf]

Hi, can you explain more? Because Cagliostro Colab UI is also using Automatic1111's, and even Anapnoe's uses it as the backend, so I believe there is no difference in terms of result.

If you set use_preset to True, I made changes to the ui-config.json file. Instead of empty values, I added default presets so that it launches using the provided values.

You can turn use_presets off if you believe this is the cause of the problem.

if use_presets:
    # ui-config.json
    default_prompt = "masterpiece, best quality,"
    default_neg_prompt = "(worst quality, low quality:1.4)"
    default_sampler = "DPM++ 2M Karras"
    if dpm_v2_patch:
        default_sampler = "DPM++ 2M Karras v2" 
    default_steps = 20
    default_width = 512
    default_height = 768
    default_denoising_strength = 0.55
    default_cfg_scale = 7

    # txt2img
    config = read_config(ui_config_file)
    config["txt2img/Prompt/value"] = default_prompt
    config["txt2img/Negative prompt/value"] = default_neg_prompt
    config["txt2img/Sampling method/value"] = default_sampler
    config["txt2img/Sampling steps/value"] = default_steps
    config["txt2img/Width/value"] = default_width
    config["txt2img/Height/value"] = default_height
    config["txt2img/Upscaler/value"] = "Latent (nearest-exact)"
    config["txt2img/Denoising strength/value"] = default_denoising_strength
    config["txt2img/CFG Scale/value"] = default_cfg_scale

    # img2img
    config["img2img/Prompt/value"] = default_prompt
    config["img2img/Negative prompt/value"] = default_neg_prompt
    config["img2img/Sampling method/value"] = default_sampler
    config["img2img/Sampling steps/value"] = default_steps
    config["img2img/Width/value"] = default_width
    config["img2img/Height/value"] = default_height
    config["img2img/Denoising strength/value"] = default_denoising_strength
    config["img2img/CFG Scale/value"] = default_cfg_scale
    write_config(ui_config_file, config)

[LukeOvermind]

Hi it could be the clip skip, it is set on 2, I tried to lower it to one but no luck. Only way around was using PNG info and send to to text2image and use the overide settings