topogoogles4 – Model description

topogoogles4 is a custom Flux fine‑tune trained on a curated set of 30 high‑quality 1024×1024 images, optimized for generating detailed, consistent visuals in this specific style. It is designed for users who want repeatable character/scene styling with Flux‑level detail without having to engineer complex prompts from scratch.

Key characteristics:

• Trained on 30 carefully selected 1024×1024 images in a single coherent style.
• Optimized for detailed, high‑frequency textures and consistent subject rendering across multiple generations.
• Runs on H100 hardware on Replicate, making it suitable for higher‑resolution or batch workflows.

Intended use

Use this model for:

• Consistent character or brand visuals in the same aesthetic across a series (thumbnails, social posts, storyboards, etc.).
• Style‑locked concept art where you want Flux‑quality detail but tightly controlled look and feel.

Not ideal for:

• Photorealistic outputs outside the training style.
• Very abstract or radically different aesthetics that diverge from the training set.

How to prompt

General tips:

• Mention the subject and action first, then let the style be implied by the fine‑tune (avoid over‑styling with extra “in the style of …” modifiers).
• Keep prompts concrete: subject, setting, mood, camera/framing, and level of detail.

Example prompts:

• “Close‑up portrait of a young woman looking at the camera, soft studio lighting, highly detailed, 4k.”
• “Full‑body shot of a character walking through a neon‑lit city street at night, cinematic composition, detailed environment.”

If your fine‑tune expects a trigger word (e.g. a special token), document it here and show 1–2 minimal example prompts using that token.

Usage

Web UI

1. Open the model page on Replicate.
2. Paste one of the example prompts, adjust guidance/steps as needed, and run a test generation.

API (JavaScript)

import Replicate from "replicate";

const replicate = new Replicate({
  auth: process.env.REPLICATE_API_TOKEN,
});

const output = await replicate.run(
  "topogoogles/topogoogles4", // or specific version hash
  {
    input: {
      prompt: "Close-up portrait of a young woman, looking at the camera, soft studio lighting, highly detailed",
      seed: 42,
      // add other supported inputs here: width, height, guidance, steps, etc.
    },
  },
);

console.log(output);

API (Python)

import replicate
import os

client = replicate.Client(api_token=os.environ["REPLICATE_API_TOKEN"])

output = client.run(
    "topogoogles/topogoogles4",
    input={
        "prompt": "Full-body character in the model style, standing in a city street at night, cinematic lighting",
        "seed": 42,
        # "width": 1024,
        # "height": 1024,
        # add other supported parameters here
    },
)

print(output)

Adapt parameter names and version string to match the actual input schema of your deployed version.

Inputs and outputs

Document the key inputs your version exposes (adjust to the actual schema):

• prompt (string, required): Text description of the image to generate.
• seed (integer, optional): Fixed seed for reproducible results.
• width, height (integers, optional): Output resolution, typically 1024×1024 for best quality.
• Any other important knobs (guidance, steps, scheduler, etc.).

Outputs:

• One or more image URLs in PNG/JPEG format, suitable for direct display or download.

Limitations and ethics

• The model is specialized to its training style; outputs outside that domain may look off‑style or low quality.
• Do not use the model to generate harmful, illegal, or infringing content, and respect Replicate’s and Hugging Face’s content and license guidelines.