msalsas/alexa-lora

Alexa Pavlova LoRA

Prompt

-

Prompt

-

Prompt

-

Model description

---

license: mit tags: - lora - stable-diffusion-xl - sdxl - text-to-image pipeline_tag: text-to-image

AlexaPavlova LoRA

A LoRA adapter that renders the visual identity of AlexaPavlova, the AI persona from the amanuensis project. It is trained to reproduce a single consistent synthetic character across generations.

TL;DR

• Type: LoRA adapter (not a full checkpoint — contains only trained low-rank deltas, none of the base model's weights)
• Base model: Juggernaut XL "Ragnarok" by RunDiffusion (KandooAI, SDXL) — required at inference time, not included here. Civitai
• Resolution: 768×1024 (portrait)
• Trigger word: 41ex4_p4v10v4
• Recommended weight: 0.3
• License: MIT

What it is

This adapter defines the face and look of a fully synthetic character. AlexaPavlova is not a real person, and the LoRA is not trained on any real individual's likeness.

Training data — provenance

The training dataset was generated entirely by me using Juggernaut XL "Ragnarok". Every training image is synthetic SDXL output; no photographs, no real people, no third-party or scraped images were used. The character is an original synthetic identity with no real-world referent.

Because the LoRA file stores only the trained deltas, it does not redistribute Juggernaut XL or any SDXL weights.

Usage

1. Get Juggernaut XL "Ragnarok" by RunDiffusion from Civitai (it is not distributed here).

2. Load this LoRA on top of it at weight 0.3, generating at 768×1024.

3. Build the prompt with the trigger word and a fixed set of appearance tags, then append whatever you want to vary per image:

   41ex4_p4v10v4, short hair, blue eyes, blonde hair, selfie, tough, <scene words here>

   The trigger + appearance tags keep the character consistent; the trailing words (e.g. at a desk, coffee shop, beach) change the scene.

This adapter is intended for use with that SDXL-based checkpoint. Inference with the base model is subject to the base model's own license terms (see its Civitai page).

Sample images

All samples are synthetic outputs of Juggernaut XL + this LoRA. The character has no real-world referent.

License

The adapter weights in this repository are released under the MIT License and are my own work. The base model and its use restrictions are governed separately by their respective licenses.

Trigger words

You should use 41ex4_p4v10v4 to trigger the image generation.

You should use short hair to trigger the image generation.

You should use blue eyes to trigger the image generation.

You should use blonde hair to trigger the image generation.

You should use selfie to trigger the image generation.

You should use tough to trigger the image generation.

Download model

Download them in the Files & versions tab.