Snapchat Selfie Z-Image LoRA Guide

TL;DR:

• This LoRA reproduces the casual, phone-camera Snapchat selfie look on top of Tongyi-MAI/Z-Image-Turbo.
• Trigger it with the phrases "Snapchat Selfie" and "Snapchat pic."
• Weights are in safetensors format; the author recommends ComfyUI. License: Apache 2.0.

What the model does

Snapchat Selfie Redmond Z-Image is a text-to-image LoRA that steers Z-Image Turbo toward selfie-style photography. The model card states it was trained on Snapchat Selfie style images and that it produces high-quality results with strong consistency across generations.

What that means in practice is visible in the gallery: subjects captured at arm's length in everyday indoor scenes, with the phone in frame and the loose, unposed composition of an actual selfie. The outputs read as social-media snapshots rather than studio portraiture or cinematic staging.

If you are looking for related takes on this aesthetic, the same author has a Snapchat Selfie variant for FLUX Klein and a broader selfie photography LoRA for SDXL.

Trigger words

The card lists two trigger phrases, both of which should be included in your prompt:

• Snapchat Selfie
• Snapchat pic.

Omitting the trigger is the most common reason the style does not appear — without it, the base model has no signal pointing at the selfie aesthetic.

How to use it

Load the safetensors weights into a ComfyUI workflow on top of the Tongyi-MAI/Z-Image-Turbo base model, then add the trigger phrases to your prompt. The model card explicitly recommends ComfyUI for best results and does not document a specific LoRA strength, CFG scale, or step count, so start from your usual Z-Image Turbo settings and adjust the LoRA weight up or down until the style sits where you want it.

The example below shows the style transferring to a different subject and setting — the model holds the casual selfie framing rather than collapsing into generic portraiture.

Diffusers users can also load it through the standard LoRA pipeline since the repository is tagged for the diffusers library.

Example use cases

• Social media content: consistent selfie-style imagery for posts, avatars, or mood boards.
• Character consistency testing: the same subject across multiple selfie-style renders, using the LoRA's reported consistency.
• Style transfer experiments: combine with other Z-Image Turbo LoRAs — e.g. a fluent emoji Z-Image LoRA or an isometric illustration Z-Image LoRA — to blend aesthetics.

Common errors and fixes

Error | Cause | Fix | Source

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

Output has no selfie style | Trigger phrases missing from prompt | Add "Snapchat Selfie" and "Snapchat pic." | HF model card Style too weak or too strong | LoRA weight not tuned | Adjust LoRA strength from your usual Z-Image Turbo baseline | Inference from card guidance

FAQ

Which base model does this LoRA need? Tongyi-MAI/Z-Image-Turbo, per the model metadata.

Does it work in AUTOMATIC1111 or ComfyUI? The card recommends ComfyUI. A1111 is not mentioned.

What license is it released under? Apache License 2.0.

What are the recommended sampling settings? The card does not document CFG scale or step counts; use standard Z-Image Turbo values and tune the LoRA weight.

Sources

• HuggingFace model card: https://huggingface.co/artificialguybr/SNAPCHATSELFIE-REDMOND-ZIMAGE