ArtificialGuyBR

Home / Blog / Training LoRAs

Snapchat Selfie LoRA for Qwen-Image

Generate high-quality Snapchat selfie style images with this Qwen-Image LoRA. Trigger words, ComfyUI usage, license, and example output.

2 sources cited Training LoRAs

On this page

  1. What This Model Does
  2. Training Details
  3. How to Use
  4. Recommended Settings
  5. Example Outputs
  6. Common Errors and Fixes
  7. FAQ
  8. Sources

Snapchat Selfie LoRA for Qwen-Image

TL;DR

What This Model Does

This LoRA adapts Qwen-Image-2512 to generate images in a Snapchat Selfie style. According to the model card, it was trained on Snapchat Selfie style images and delivers "high-quality snapchat selfie content with excellent consistency" and "high detail and consistency."

Example output from the Snapchat Selfie Redmond LoRA

The style is consistent across outputs, which makes the LoRA useful for generating series of selfie-style images that share the same look.

Training Details

How to Use

  1. Load the LoRA with the Load LoRA node
  2. Add the trigger phrase to the positive prompt: Snapchat Selfie. Snapchat pic.
  3. Start with a LoRA strength around 0.8 and adjust up or down
  4. Generate

Other Interfaces

Because the model ships as a diffusers LoRA in Safetensors format, it also loads in diffusers-compatible pipelines. Load the .safetensors weight as a LoRA and include the trigger words in the prompt.

Second example output from the Snapchat Selfie Redmond LoRA

Setting | Recommendation

|---|---

Base model | Qwen-Image-2512 Trigger phrase | Snapchat Selfie. Snapchat pic. LoRA strength | Start at 0.8, tune 0.6–1.0 Sampler | Euler a or DPM++ 2M Karras Steps | 20–30 CFG | 4–7

Example Outputs

Third example output from the Snapchat Selfie Redmond LoRA

The three examples above come from the model repository's sample gallery (images/001.png through images/003.png).

Common Errors and Fixes

Error | Cause | Fix

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

Style not applying | Trigger phrase missing or misspelled | Use the exact phrase: Snapchat Selfie. Snapchat pic. Style too strong | LoRA weight too high | Reduce to 0.6–0.7 Wrong look entirely | Base model mismatch | This LoRA is architecture-specific to Qwen-Image-2512

FAQ

Q: Does this LoRA work with SDXL or FLUX? A: No. It is trained for Qwen-Image-2512 and should be used with that base model.

Q: Can I use it commercially? A: Yes. It is licensed under Apache 2.0, which permits commercial use with attribution.

Q: Where can I download the weights? A: From the HuggingFace files tab: https://huggingface.co/artificialguybr/SNAPCHATSELFIE-REDMOND-QWENIMAGE

Sources