CivArchive
    Technically Color Z - v1.0 - Z Image Turbo
    Preview 111930510
    Preview 111930202
    Preview 111932541
    Preview 111932963
    Preview 111928979
    Preview 111942828
    Preview 111929504
    Preview 111930236
    Preview 111931029
    Preview 111932260
    Preview 111931669
    Preview 111930027
    Preview 111930704
    Preview 112041741
    Preview 111936376
    Preview 111936690
    Preview 111936789

    Technically Color Z is meticulously crafted to capture the unmistakable essence of classic film.

    This LoRA was trained on approximately 100+ stills to excel at generating images imbued with the signature vibrant palettes, rich saturation, and dramatic lighting that defined an era of legendary classic film. This LoRA greatly enhances the depth and brilliance of hues, creating more realistic yet dreamlike textures, lush greens, brilliant blues, and sometimes even the distinctive glow seen in classic productions, making your outputs look truly like they've stepped right off a silver screen. Images were captioned using Joy Caption Batch, and the model was trained with ai-toolkit and tested in ComfyUI.

    The gallery contains examples with workflows attached. I'm running a very simple 2-pass workflow for most of these; drag and drop the first image into ComfyUI to see the workflow.

    Update: If you have updated ComfyUI to resolve the "LoRA not loaded error message" you may notice this LoRA makes skin too smooth, lower the strength of the LoRA to around 0.3-0.5 and it still retains the style and resolves a lot of the smoothing issues.

    If you've enjoyed using this LoRA and the other resources I share, please consider sponsoring my work on Ko-Fi. Any support covers essential costs like cloud compute, enabling me to keep creating and sharing these resources with you. Every contribution, no matter how small, makes a difference! 👍🏼

    renderartist.com

    Description

    Trained for 2000 steps

    FAQ

    Comments (18)

    paralaif_xoxNov 29, 2025· 1 reaction
    CivitAI

    Beatiful lora!!

    SeoulSeekerNov 29, 2025· 1 reaction
    CivitAI

    only a true gentleman would conceive of a LoRA such as this (assuming you're a guy) 😆

    Keroro_GunsoNov 29, 2025· 1 reaction
    CivitAI

    Nice lora!

    qekNov 29, 2025· 12 reactions
    CivitAI

    Bad lora, it makes my images blurry and unrealistic

    renderartist
    Author
    Nov 29, 2025

    I noticed the behavior of the LoRA changed when I updated ComfyUI and it enabled all of the LoRA layers, try lowering the strength to around 0.4 and see if that helps.

    qekNov 30, 2025· 1 reaction

    Stop dislikebombing, it isn't my bad

    eviltigz710Feb 7, 2026

    Yes it is.

    kirillallNov 29, 2025· 1 reaction
    CivitAI

    Cool one to use! Thank you!

    bot_ai_but_humanNov 29, 2025· 2 reactions
    CivitAI

    I really enjoy your Loras. What settings did you use to train? For example training rate? Can you provide the config file?

    renderartist
    Author
    Nov 29, 2025· 2 reactions

    https://pastebin.com/V4SJ8xGM This is the config.yaml - You can copy and paste text into the "Show Advanced" section of your training job. It's mostly standard settings, I used batch 2 and set the rank to 48 but I think for this model ai-toolkit defaults to 32 anyway, not sure.

    bot_ai_but_humanNov 29, 2025

    Thank you! Why is the trigger word 'null' but you use a trigger word in your sample prompts?

    qekNov 30, 2025

    @bot_ai_but_human No trigger word

    saryacatmander788Nov 29, 2025· 2 reactions
    CivitAI

    Produces good images as the examples and works well with my custom character lora, not changing the face. Thank you.

    renderartist
    Author
    Nov 29, 2025

    Thanks! 🙌🏼

    olivettyNov 30, 2025· 1 reaction
    CivitAI

    This looks so good! Thanks! <3

    NektarineDec 5, 2025· 2 reactions
    CivitAI

    Great! I'm having trouble training a style. Does your dataset have captions?

    My dataset with captions didn't turn out very well, so I was thinking of training without captions and without trigger words so that the model learns to generalize the style rather than learning what is present in each image. Thanks!

    renderartist
    Author
    Dec 5, 2025

    This did have simple captions about 170 or so tokens each, captioned with Joy Caption Batch, I've been experimenting without captions or triggers and it works really well for some harder to train styles. Just experiment with what your gut tells you based on your results, if it looks garbled lower the learning rate. I think with the Turbo model less is more for captions, maybe even skip captioning on things you think it might tend to overly focus on.

    scriptr78Jan 6, 2026
    CivitAI

    To cold (blue) and burning at str0.9. Better str0.15-0.25, but cold (blue) filter there is too.