ArtificialGuyBR

Home / Blog / Training LoRAs

SDXL Cute Fruits LoRA: Cartoon Fruit Faces Made Easy

CuteFruitsRedmond is an SDXL LoRA that turns fruit into funny cartoon characters with faces. Trigger word, settings, and usage inside.

2 sources cited Training LoRAs

On this page

  1. What the model does
  2. How to use it
  3. Recommended settings
  4. Where it fits
  5. Common errors and fixes
  6. FAQ
  7. Sources

SDXL Cute Fruits LoRA: Cartoon Fruit Faces Made Easy

CuteFruitsRedmond, by ArtificialGuyBR, is a Stable Diffusion XL LoRA purpose-built for one job: generating funny, cute images of fruits with faces. Instead of fighting prompts to get a strawberry to smile, you load the LoRA, type the trigger word, and get consistent cartoon fruit characters.

Example generated by the CuteFruitsRedmond LoRA showing a cute cartoon fruit character

What the model does

The LoRA was trained on a large dataset of fruits with faces and fine-tuned on SDXL 1.0, per the HuggingFace model card. Gallery output shows the signature result: bright, toy-like 3D renders of fruit characters with big expressive faces — the kind of image you'd expect on a sticker sheet or a kids' app. Because it's a LoRA rather than a full checkpoint, the base model's general capabilities stay intact and you can blend the style with whatever prompts you already know.

The model card credits GPU time from Redmond.AI for the training run, which is where the "Redmond" in the name comes from.

Another CuteFruitsRedmond example showing a different cute fruit character with a face

How to use it

ComfyUI

  1. Download the LoRA from the HuggingFace model page or Civitai.
  2. Drop the file into your models/loras folder.
  3. Add a LoRA loader node, pick the base SDXL checkpoint, and connect it to your positive conditioning.
  4. Put CtFruitsRedmAF in the positive prompt and describe the fruit: "CtFruitsRedmAF, a cheerful strawberry with big eyes, soft studio lighting".

If you're new to the node-based pipeline, our ComfyUI LoRA workflow guide walks through the loader, weight, and sampler basics.

Automatic1111 / ComfyUI WebUI

Diffusers

from diffusers import DiffusionPipeline
import torch

pipe = DiffusionPipeline.from_pretrained(
    "stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16
).to("cuda")
pipe.load_lora_weights("artificialguybr/CuteFruitsRedmond")

image = pipe("CtFruitsRedmAF, a happy watermelon character").images[0]

The card doesn't specify exact sampler or step recommendations, so start from standard SDXL LoRA defaults and tune from there:

The fruit character style is forgiving, so moderate weights usually keep the fruit looking like fruit while the face stays consistent. For a full walkthrough of loader setups, see our Stable Diffusion LoRA guide.

Third CuteFruitsRedmond example showing the range of cute fruit characters the LoRA produces

Where it fits

If you generate a lot of cute character art, this LoRA is a fast shortcut: no prompt-wrestling to anthropomorphize fruit. It also pairs well with other style models — try stacking it with a sticker-style or clay-render LoRA to push the look further. Browse the rest of the cute LoRA collection for matching styles.

Common errors and fixes

Error | Cause | Fix | Source

|---|---|---|---

Style not activating | Trigger word missing | Add CtFruitsRedmAF to the positive prompt | Model card Distorted fruit | Wrong base model | Use SDXL 1.0 checkpoint; other bases (Flux, Qwen, ZImageTurbo) have their own loaders | Civitai base models Faces look wrong at low weight | LoRA weight too low | Raise weight toward 1.0 | Inference

FAQ

What is the trigger word? CtFruitsRedmAF. It appears in the model card as instance_prompt and in the Civitai trained words list.

What base model do I need? SDXL 1.0 is the training base. Civitai also lists Flux.2 Klein 9B, Qwen, and ZImageTurbo as compatible, but those use their own model formats.

Is it free to use? The HuggingFace card is licensed under creativeml-openrail-m; the Civitai page lists Apache 2.0. Check the base model's license too for commercial use.

Sources