CivArchive
    2025-08-30-15-37-22-another-midjourney-release-old-man - V2
    NSFW
    Preview undefined
    Preview undefined
    Preview undefined

    The Storyline

    A data scientist embarks on an ambitious project: to use artificial intelligence to preserve a piece of their personal history. The goal is to train a model to generate images of their "father," their "grandfather," and even an image combining the two.

    The project starts with a collection of 73 image files. These images are not just data; they are memories, moments captured in time. The AI begins its arduous task of learning, processing the images over 18 epochs. The training logs reveal the struggle of this process, with the system encountering technical warnings and a loss function that slowly decreases, indicating the model is learning but the process is challenging and takes many hours.

    The AI, named Illustrious, diligently works through the provided photos, learning the nuances of each person's face. The progress is a long, methodical journey, documented by the regular saving of intermediate checkpoints after each epoch.

    In the end, the AI succeeds, creating a new image that combines the features of both their father and grandfather. The final output isn't just a generated image; it's a testament to the power of AI to create something new from cherished memories, a digital family portrait that couldn't exist in the real world.

    {
      "training_parameters": {
        "engine": "kohya",
        "loraType": "lora",
        "resolution": 1024,
        "maxTrainEpochs": 18,
        "trainBatchSize": 4,
        "targetSteps": 9855,
        "learning_rate": 0.0005,
        "unet_lr": 0.0005,
        "text_encoder_lr": 5e-5,
        "network_dim": 32,
        "network_alpha": 16,
        "lr_scheduler": "cosine_with_restarts",
        "lr_scheduler_num_cycles": 3,
        "min_snr_gamma": 5.0,
        "noise_offset": 0.1,
        "clip_skip": 1,
        "optimizer_type": "Adafactor",
        "save_every_n_epochs": 1,
        "sample_every_n_epochs": 1,
        "shuffleCaption": true,
        "flipAugmentation": true,
        "keepTokens": 3,
        "save_precision": "bf16",
        "mixed_precision": "bf16"
      },
      "dataset_configuration": {
        "num_repeats": 30,
        "image_resolution": 1024,
        "caption_extension": ".txt",
        "enable_bucket": true,
        "bucket_reso_steps": 64,
        "min_bucket_reso": 256,
        "max_bucket_reso": 2048,
        "files": 73
      },
      "sample_prompts": [
        {
          "prompt": "my father"
        },
        {
          "prompt": "my grandfather"
        },
        {
          "prompt": "my father and my grandfather"
        }
      ]
    }

    Description

    The Story of the Pony Model

    The journey began on August 31, 2025, when a new training job was submitted at 9:10:27 AM. The goal was to create a new model based on the "pony" base model. The process was highly structured and technical, driven by a series of precise configurations.

    The dataset, a collection of 2482 images, was prepared for training. It was configured to a resolution of 1024 pixels, with a plan to shuffle captions and use a "flip" augmentation to improve the model's learning. The training would run for 20 epochs, with a total of 10,000 optimization steps. The team used the Prodigy optimizer with a specific learning rate to guide the model's development.

    Throughout the training, the system meticulously saved a new checkpoint at the end of each epoch, creating a detailed record of the model's progress. The logs documented the process, noting the loading of settings and the progress of each epoch.

    Sample prompts were used to test the model's ability to generate new images as it learned. These prompts were simple yet personal, including "my father," "my grandfather," and "my father and my grandfather".

    The training job progressed steadily, and by 5:39 PM on the same day, the job was marked as ready. The final model, a file with a .safetensors extension, was uploaded, followed by the sample images that were generated during the training process. The successful completion of the job was confirmed with metrics being pushed, and the final model and sample images being uploaded to their respective asset links.

    Physical Attributes
    
    Body Type & Muscles: muscular , muscular male , pectorals , abs , large pectorals , thick thighs.
    
    Hair: short hair , black hair , gray hair , beard , facial hair , hairy.
    
    Body Parts & Features: navel , nipples , chest hair , arm hair , leg hair.
    
    Other details: glasses , scar , tattoo , light skin , tan skin.
    Attire & Accessories
    
    Clothing: white shorts , white polo shirt , white shirt , shirt, tank top , black tank top , yellow shirt, collared shirt.
    
    Footwear: shoes , sneakers , white footwear.
    
    Accessories: belt , glasses , white gloves.
    
    Jewelry & Piercings: ring , jewelry , nipple piercing.
    Setting & Context
    
    Location: modern office-like room, car interior , church , modern kitchen , outdoors , golf course.
    
    Actions: standing , sitting , holding golf club , hands in pockets.
    
    Lighting & Style: natural lighting , artificial lighting , realistic style , high aesthetic quality.
    Technical Photo Details
    
    Camera Model: Canon EOS 5D Mark IV , Canon EOS 5D Mark III.
    
    Lens: 50mm lens , 85mm lens , Canon EF 24-70mm f/2.8L II USM , Canon EF 50mm f/1.8 STM.
    
    Settings: f/2.8 aperture , ISO 400 , ISO 100 , eye-level angle.
    Here are descriptive captions for the subject character, identified as "chweeee1" in the Pony version.
    
    ***
    
    ### Physical Attributes
    * **Build**: Muscular, muscular male, defined abs, large pectorals, thick thighs. One image caption also describes a "slightly chubby physique" and another a "bodybuilder physique."
    * **Age**: Middle-aged, older man, mature male.
    * **Hair**: Short black or gray hair, beard, facial hair, mustache, hairy. Specific hairiness is noted on the chest, arms, and legs.
    * **Skin**: Light skin, tan skin, dark brown skin. One image caption specifies "muscular Asian man."
    * **Other details**: Glasses, small tattoo on the left shoulder, and a scar.
    
    ***
    
    ### Attire
    * **Clothing**: White shorts, white polo shirt, white shirt, black tank top, gray sweatpants with white text, blue striped shirt, beige pants, and a yellow shirt with brown floral patterns (Hawaiian shirt).
    * **Accessories**: Belt, white gloves, and a captain's hat.
    * **Jewelry**: Nipple piercings (sometimes with a chain connecting them), rings, a gold wedding ring, and a gold necklace.
    
    ***
    
    ### Poses & Expressions
    * **Poses**: Standing with hands in pockets, sitting, sitting in a car, sitting on a wooden bench, holding a golf club, arms crossed over chest, hands clenched into fists, leaning against an oven, and holding his penis.
    * **Expressions**: Neutral, serious, confident, slight smile, slightly open mouth, and a slightly pouty expression.
    LORA
    Pony

    Details

    Downloads
    1
    Platform
    CivitAI
    Platform Status
    Available
    Created
    8/31/2025
    Updated
    8/25/2026
    Deleted
    -
    Trigger Words:
    bara

    Files

    2025-08-30-15-37-22-another-midjourney-release-old-man.safetensors