Selfie Photography LoRA for SDXL — Realistic Selfie Portraits on Demand

SelfiePhotographyRedmond is a Stable Diffusion XL LoRA adapter that turns any subject prompt into a realistic selfie-style portrait. It handles a wide range of demographics, ages, and aesthetics — from casual Instagram shots to professional headshots.

• Trigger words: instagram model, discord profile picture
• Base model: Stable Diffusion XL 1.0
• Format: Safetensors, works in diffusers, ComfyUI, A1111, Forge

What This LoRA Does

The SelfiePhotographyRedmond LoRA was fine-tuned on SDXL 1.0 with GPU resources from Redmond.AI. It specializes in selfie and portrait photography framing — close-up compositions with natural lighting, expression, and depth of field typical of smartphone or webcam selfies.

The model's strength is versatility. Rather than locking into one demographic, it generates convincing selfie portraits across age groups, ethnicities, and styles. A young Asian woman with rainbow hair, a 95-year-old grandmother with pink hair, and a 35-year-old Indian tech executive all render with consistent photographic quality.

How to Use the LoRA

Prompt Format

The model expects this structure:

A photo of a {your subject description}, instagram model, discord profile picture

The trigger words instagram model and discord profile picture activate the selfie photography style. You can omit one of the triggers if you want less of the "profile picture" aesthetic.

Example Prompts

Prompt | Result

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

A photo of a blonde woman wearing rainbow sunglasses, 19 years, instagram model, discord profile picture | Casual young portrait A photo of an old granny with pink hair, 95 years old, instagram model, discord profile picture | Elderly selfie portrait A photo of an Indian Tech CEO, 35 years old, instagram model | Professional headshot A photo of a basketball player, instagram model, discord profile picture | Athletic portrait

Loading with Diffusers

from diffusers import AutoPipelineForText2Image
import torch

pipeline = AutoPipelineForText2Image.from_pretrained(
    'stabilityai/stable-diffusion-xl-base-1.0',
    torch_dtype=torch.float16
).to('cuda')

pipeline.load_lora_weights(
    'artificialguybr/selfiephotographyredmond-selfie-photography-lora-for-sdxl',
    weight_name='SelfiePhotographyRedmond.safetensors'
)

image = pipeline(
    'A photo of a young woman with glasses, instagram model, discord profile picture'
).images[0]
image.save('output.png')

Recommended Settings

• Base model: SDXL 1.0 (Stable Diffusion XL Base)
• LoRA weight: Start at 0.8–1.0; lower to 0.6–0.7 for subtler effect
• Resolution: 1024x1024 (native SDXL) or 832x1216 for vertical selfie framing
• Sampler: DPM++ 2M Karras or Euler a
• Steps: 20–30
• CFG scale: 5–8

The LoRA works with most SDXL checkpoints, but SDXL 1.0 base produces the most consistent results as shown in the author's examples.

Training Details

The model was fine-tuned on SDXL 1.0 as a LoRA adapter. The author describes it as a "Doodle LoRA" — a term used for their collection of focused-style adapters. GPU training was provided by Redmond.AI.

Training data focused on selfie-style photography: close-up portraits, face-forward compositions, and the kind of casual photographic aesthetic seen in social media profile pictures. The two trigger words (instagram model and discord profile picture) steer the generation toward this specific style.

Limitations

• The model is tuned for portrait/selfie framing — it will not produce landscape, still life, or abstract content.
• The trigger words are part of the trained style; omitting them weakens the selfie effect.
• Results are best with subjects described in natural language (age, appearance, style) rather than overly abstract prompts.
• This is a style adapter, not a face-specific model — it does not replicate a specific person.

Common Errors and Fixes

Error | Cause | Fix | Source

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

Flat/no selfie framing | Missing trigger words | Add instagram model, discord profile picture to prompt | HF README Wrong art style | Using with wrong base model | Use SDXL 1.0 base checkpoint | HF README Loading fails | Wrong filename | Use SelfiePhotographyRedmond.safetensors | HF README Weak effect | LoRA weight too low | Increase weight to 0.8–1.0 | Common SDXL LoRA practice

FAQ

Q: Can I use this with Stable Diffusion 1.5? A: No. This LoRA is trained on SDXL 1.0 and is not compatible with SD 1.5 or SD 2.x checkpoints.

Q: What are the exact trigger words? A: instagram model and discord profile picture. Both activate the selfie style; using both gives the strongest effect.

Q: Does this work in ComfyUI? A: Yes. Load the Safetensors file as a LoRA node, connect to any SDXL checkpoint, and use the same trigger words in your positive prompt.

Q: Can I use this commercially? A: The model uses a bespoke LoRA trained license that allows derivatives with attribution. Commercial use requires rental licensing per the license terms. Check the license link on the model page for details.

Q: How does this compare to other portrait LoRAs? A: This LoRA specifically targets the selfie/social media profile picture aesthetic. Other portrait LoRAs may focus on fine art, studio photography, or specific character types. The versatility across demographics is a distinguishing feature.

Sources

• HuggingFace Model Card — Full README with usage code and trigger word documentation
• Civitai Model Page — Community gallery, download stats (1,500+ downloads), and sample images

See Also

• More SDXL LoRA Models
• Guide to Using LoRAs with SDXL
• Stable Diffusion XL Getting Started