Skip to content

Z-Image: Broken rendering with long prompt and higher NAG Scale #1412

Description

@GoldMath

See #1412 (comment) for updated issue.

Hello, I been noticing a strange behavior with Z-Image Turbo 6B when prompting and using the Enhance Prompt, the rendering seems to broke after these steps. Exemple use the same seed, negative, settings and profiles. Tested on Profile 2 & 3+, also with Compile Transformer On/Off.

Florence 2 + Llama Joy

and noticed this console message when using the Enhance Prompt:

A:\Wan2GP\Wan2GP\python_embeded\lib\site-packages\pygame\pkgdata.py:25: UserWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html. The pkg_resources package is slated for removal as early as 2025-11-30. Refrain from using this package or pin to Setuptools<81.
 from pkg_resources import resource_stream, resource_exists
Hooked to model 'prompt_enhancer_image_caption_model' (Florence2ForConditionalGeneration)
Hooked to model 'prompt_enhancer_llm_model' (LlavaForConditionalGeneration)
Setting `pad_token_id` to `eos_token_id`:128001 for open-end generation.
Loading Model 'ckpts\ZImageTurbo_quanto_bf16_int8.safetensors' ...
Loading Text Encoder 'ckpts\Qwen3\qwen3_quanto_bf16_int8.safetensors' ...
Hooked to model 'text_encoder' (Qwen3ForCausalLM)
Hooked to model 'vae' (AutoencoderKL)
 0%|                                                                                         | 0/8 [00:13<?, ?steps/s]
New image saved to Path: outputs\2026-01.jpg

Regards,

  1. Prompt and Generate

a cat with a ball toy
Image

  1. Click Enhance Prompt and Generate

A photograph of a cat with a ball toy, captured in mid-play. The cat is positioned slightly to the right of the frame, with its body angled to the left, creating a dynamic diagonal line. The cat's head is turned towards the ball, which is held in its mouth, with its eyes focused intently on it. The cat's fur is short and a light gray color, with a slight sheen under the soft, natural light. The background is a simple, blurred indoor setting, emphasizing the cat and the ball. The composition uses a shallow depth of field, with the cat in sharp focus and the background softly out of focus. The lighting is even, highlighting the cat's playful expression and the texture of its fur. The overall mood is playful and energetic.
Image

  1. If I get back to the initial Prompt and Generate

a cat with a ball toy
Image

  1. I then have restarted the whole app and tried the enhanced prompt directly (different seed/same settings):

A photograph of a cat with a ball toy, captured in mid-play. The cat is positioned slightly to the right of the frame, with its body angled to the left, creating a dynamic diagonal line. The cat's head is turned towards the ball, which is held in its mouth, with its eyes focused intently on it. The cat's fur is short and a light gray color, with a slight sheen under the soft, natural light. The background is a simple, blurred indoor setting, emphasizing the cat and the ball. The composition uses a shallow depth of field, with the cat in sharp focus and the background softly out of focus. The lighting is even, highlighting the cat's playful expression and the texture of its fur. The overall mood is playful and energetic.
Image

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions