👉 Join my Discord for help, prompt design and news on upcoming LoRAs
💫 Aether Exposure – Double Exposure LoRA for Subjects and Environments (t2v)
Aether Exposure is a first version of a paired LoRA set (low and high noise) that creates layered double exposure compositions — blending a human subject with a landscape, city or other background.
Trained specifically for silhouette separation and coherent overlay, it’s ideal for stylized, surreal storytelling sequences focused on people. Also works great to generate single one frames double exposure images. Best compositional output there is.
💬 Default Trigger Phrase
Use this exact format to trigger the effect — then insert your subject and background context:
double exposure of a [subject] and a [context], [white/black] background
Prompt example: double exposure of a woman and a forest, black background
You can also experiment with different types of motion applied to either the subject or the background context:
Example: double exposure of a turning woman and a panning over a forest, black background
🧠 How to Use (t2v)
• Works best with either black or white backgrounds
• Combine with camera movement (better controlled) motion
⚙️ Inference Settings
• Mode: text-to-video (t2v)
• Model: Wan 2.2 14B
• Resolution: 480–1024px (approx. 720p)
• FPS: 16-24
• Steps: 20 (10 steps each low and high noise)
• CFG: 3.5
• Clip length: 3–5 seconds (73–121 frames)
Thanks to @masslevel for contributing videos!
Description
FAQ
Comments (9)
Would you mind uploading the safetensors files directly, please? All LoRA management apps depend on civitAI's API and require a safetensors URL endpoint. If you upload a zip then none of the scripts, apps, or software can fetch metadata for the LoRA.
ok. I didnt see that I could upload two files at once. will try
@Gongoloid so I tried this and just get error. will see if civitai can help me. for now it's the zip.
@joachim_s you have to add a new version, like as if you would add a V2 of your lora ;)
@WhatTheGuy Yeah. I won’t do that.
@joachim_s Well you'll never get credit for people using your LoRA's. Civitai rely on the hash codes and AIR tags of the file being uploaded (In this case a ZIP file) compared to the meta data used in the generation (the hash of the actual loras) Since the zip file has a different HASH, you'll never get credit for the use of the lora, and you're literally missing out on free Buzz.
So you either upload two version a "low" and a "high" or you don't and get nothing for your effort.
@joachim_s - literally everyone else does, though.
@Gongoloid @WhatTheGuy @SencneS - fixed
This lora would be awesome if ppl had the time and the GPU to run it