CivArchive

    Relive the golden age of the Wild West myth.

    minimax h3: I noticed I didn't use the minimax captions when training, so it's still using the z-image format below. However, it seems to work well anyway.

    z-image: I've changed captioning process for this model, and this one is a bit more elaborate in the descriptions. I also left in some comments about the quality, not sure if it was a good idea or not. So, compared to my other models, it responds better to long prompts. Example:

    "50s_western. a group of cowboys in various period clothing, riding slowly through a ravine. The background features steep, rocky hillsides with sparse vegetation. The camera shot is taken from a ahead of the group, focusing on them and the imposing rocks. The lighting is natural, with morning light. The photograph has a vintage, cinematic quality."

    These movies were made for cinema, and a a very wide format. I had to cut them down to save my poor gpu, but I tried to maintain composition and a wide aspect ratio. Try going beyond 16:9.

    Description

    FAQ

    Comments (4)

    MikeflowerOct 1, 2026· 1 reaction
    CivitAI

    💕👍🏻

    boobkake22Oct 1, 2026· 1 reaction
    CivitAI

    Wonderful.

    aixartsOct 1, 2026· 1 reaction
    CivitAI

    Fantastic!!

    boobkake22Oct 2, 2026
    CivitAI

    I did some testing with this last night. Works quite well. Really like how it adjusts the tone of dialog to better match the period performances.

    I had a question with regards to film grain. Seems like H3 finds film grain to be an anathma, and it tends to render it as a static texture that mottles and them moves with everything. Have you found a way to prompt or otherwise configure settings in such a way to improve that?

    LORA
    MiniMax H3

    Details

    Downloads
    371
    Platform
    CivitAI
    Platform Status
    Available
    Created
    10/1/2026
    Updated
    10/4/2026
    Deleted
    -

    Files

    1950s_western_h3_150.safetensors

    Mirrors