CivArchive
    CielBleu ZIT - v1
    NSFW
    Preview 130087981
    Preview 130087986
    Preview 130088004
    Preview 130087982
    Preview 130087983
    Preview 130087988
    Preview 130087989
    Preview 130088009
    Preview 130088008
    Preview 130087984
    Preview 130087985
    Preview 130088033
    Preview 130088003
    Preview 130088011
    Preview 130088034
    Preview 130088035
    Preview 130088036
    Preview 130088037
    Preview 130088038
    Preview 130088039

    📝 Description

    CielBleu ZIT is an experimental fine-tuned merge, based on the original Z-Image Turbo and other ZIT models.

    It's designed for high-fidelity photorealism. It delivers photorealistic portraits with sophisticated lighting and organic texture rendering, maintaining full creative flexibility and prompt fidelity.

    Sampling Method : Euler
    Schedule Type : Simple
    Steps : 16
    CFG : 1
    Prompt : Long detailed natural language description, between 100-350 tokens.
    Resolution : Any width & height between 1024 and 2048, as long as they are divisible by 16.

    ⚠️ NSFW Note

    This model is capable of producing both SFW and NSFW content, but it has all the same flaws as the original Z-Image Turbo when it comes to NSFW content. I suggest either using a specialized NSFW Lora or performing SDXL inpainting.

    ℹ️ Required Additional Files

    Like the original Z-Image Turbo, CielBleu ZIT needs :
    Text Encoder : qwen3_4b.safetensors
    VAE : ae.safetensors

    Description

    Initial version.

    FAQ

    Comments (17)

    ralphjxMay 8, 2026
    CivitAI

    Can this checkpoint be used with a character LoRA without changing its identity?

    liftweightsMay 9, 2026· 2 reactions

    I can confirm it works with a character LoRA trained on "DARKMOP V61," which means it probably shares a lot of the same DNA. Normally, ZIT LoRAs have poor compatibility across checkpoints.

    cielbleu9991
    Author
    May 9, 2026

    @ralphjx It depends on which checkpoint your Character LoRA was trained on.
    I suggest you testing your LoRA with CielBleu ZIT, if the results differ too much, train your LoRA again on this model.

    ralphjxMay 9, 2026

    @cielbleu9991 i always train my loras on Z image Base in OstrisAI because is compatible with ZIT also... so...

    cielbleu9991
    Author
    May 14, 2026

    @ralphjx If your LoRA was trained on ZIB or ZIT, it should be compatible with CielBleu ZIT.

    ikrall001893May 9, 2026
    CivitAI

    release BF16 for making proper quants

    cielbleu9991
    Author
    May 9, 2026· 2 reactions

    Unfortunately, it's not possible to release a BF16 version, as most of the merged models are FP8 and FP16. Even though my script uses FP32 conversions, offering a BF16 version would only make the file larger without improving the result.

    TaiLongMay 9, 2026
    CivitAI

    in the end it all comes to inpainting, so can you please share you inpainting workflow used here...

    cielbleu9991
    Author
    May 9, 2026

    @TaiLong It usually produces good NSFW results even without inpainting.
    If you need to do SDXL inpainting, I recommend modifying an existing ComfyUI template.
    I use Forge Neo, and the inpainting is done directly in the "img2img" tab, then "inpaint" sub tab.

    TaiLongMay 9, 2026

    @cielbleu9991 ohh , which sdxl model? Lustify?

    cielbleu9991
    Author
    May 9, 2026

    @TaiLong It depends on the context, the SDXL models I use for inpainting are :

    - Analog Madness SDXL
    - GonzaLomo XL
    - IntoRealism Ultra / Ultra Amateur
    - HuslyoRealismXL

    TaiLongMay 9, 2026

    @cielbleu9991 ok thanks

    DelightofVisionMay 9, 2026
    CivitAI

    Please BF16 version

    cielbleu9991
    Author
    May 9, 2026

    Unfortunately, it's not possible to release a BF16 version, as most of the merged models are FP8 and FP16. Even though my script uses FP32 conversions, offering a BF16 version would only make the file larger without improving the result.

    RedAstronautMay 14, 2026
    CivitAI

    Does it work with character LoRAs? And what do you mean by SDXL inpainting for NSFW? Is there a guide on that?

    daceheg192491May 14, 2026· 4 reactions

    Worked with multiple character loras I tried it with.

    cielbleu9991
    Author
    May 14, 2026· 1 reaction

    @RedAstronaut CielBleu ZIT works with Character LoRAs trained on ZIT checkpoints.

    SDXL Inpainting is a technique that lets you fix or improve specific areas of an image without regenerating the whole thing. You paint a mask over the area you want to change, then a different checkpoint (typically an SDXL-based model) regenerates only that masked region. Since SDXL models have been trained on much more diverse data than Z-Image Turbo, they tend to produce significantly better NSFW anatomical details. The result is then seamlessly blended back into your original image.

    I suggest you have a look at an inpainting tutorial for the UI you’re using (Forge Neo, ComfyUI, etc.).

    Checkpoint
    ZImageTurbo

    Details

    Downloads
    653
    Platform
    CivitAI
    Platform Status
    Available
    Created
    5/8/2026
    Updated
    5/18/2026
    Deleted
    -

    Files

    ae.safetensors

    Mirrors

    HuggingFace (561 mirrors)
    TensorFiles (1 mirrors)
    ShakkerAI (1 mirrors)

    cielbleuZIT_v1.safetensors

    Mirrors

    CivitAI (1 mirrors)

    cielbleuZIT_v1_txt.safetensors

    Mirrors

    HuggingFace (194 mirrors)