[Bug]: [SEP] no longer works, the first face gets put on everybody about adetailer HOT 3 OPEN

Hello @Zimtschock! I'm here to help you with any bugs, questions, or contributions you might have. Let's squash those bugs together! 🐞

The issue with the [SEP] separator causing all faces to be rendered at the same time and only the first face to be chosen for all characters is likely due to the way prompts are being processed. The _get_prompt method in the Adetailer script splits the prompt using the [SEP] separator and replaces blank prompts with a default value or a replacement prompt. If the replacement logic is not correctly handling multiple faces, it could result in the behavior you are experiencing.

To address this issue, you can review and adjust the _get_prompt method to ensure it correctly processes each face individually. Here is the relevant part of the code:

def _get_prompt(
    self,
    ad_prompt: str,
    all_prompts: list[str],
    i: int,
    default: str,
    replacements: list[PromptSR],
) -> list[str]:
    prompts = re.split(r"\s*\[SEP\]\s*", ad_prompt)
    blank_replacement = self.prompt_blank_replacement(all_prompts, i, default)
    for n in range(len(prompts)):
        if not prompts[n]:
            prompts[n] = blank_replacement
        elif "[PROMPT]" in prompts[n]:
            prompts[n] = prompts[n].replace("[PROMPT]", blank_replacement)

        for pair in replacements:
            prompts[n] = prompts[n].replace(pair.s, pair.r)
    return prompts

Ensure that the ad_prompt string is correctly formatted and that the replacements are applied as intended. Additionally, verify that the prompt_blank_replacement method is providing the correct default values for each face ^[1].

If the issue persists, you may need to further debug the prompt processing logic to identify any other potential causes.

To continue the conversation, mention @dosu.

from adetailer.

I also have encountered this issue @dosu . It's very recent and not sure what is causing it. I've gone through the !adetailer script and found the code you mentioned to compare it with mine and the two are identical.

Just to add some extra detail. It seems to mostly not work when I try to use the person yolov8n-seg/person yolov8s-seg models.

Some extra, extra detail. Checking on huggingface for Bing - https://huggingface.co/Bingsu/adetailer/tree/main - the 'Person_' files have been marked as unsafe so could this be having an effect?

from adetailer.

When making a report like this, could you please also upload screenshots and prompts?

from adetailer.