joachimsallstrom

Made a little ghost LoRA since I find transparent, translucent ghosts hard to get right on SDXL. It's trained on ghost people, but you might squeeze out a few animals too. Try weights from 0.3-1.5 and beyond. Big thanks to Masslevel for all the awesome images shown here!

Click here to download the latest Double Exposure embedding for SD 2.x in higher resolution! This is version 2 of the Double Exposure Diffusion model, trained specifically on images of people and a few animals. The model file (DoubleExposurev2.ckpt) can be downloaded on the Files page. You trigger double exposure style images using token: dublex style or just dublex. Galaxy man (image 1): dublex man galaxy Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 3273014177, Size: 512x512 Emma Stone (image 2): dublex style Emma Stone, galaxy Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 250257155, Size: 512x512 Frodo (image 6): dublex style young Elijah Wood as (Frodo), portrait, dark nature Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 3717002975, Size: 512x512 Scarlett Johansson (image 1): dublex Scarlett Johansson, (haunted house), black background Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 3059560186, Size: 512x512 Frozen Elsa (image 3): dublex style Elsa, ice castle Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 2867934627, Size: 512x512 Wolf (image 4): dublex style wolf closeup, moon Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 312924946, Size: 512x512 This model was trained using Shivam's DreamBooth model on Google Colab @ 2000 steps. The previous version 1 of Double Exposure Diffusion is also available in the Files section. This model is open access and available to all, with a CreativeML OpenRAIL-M license further specifying rights and usage. The CreativeML OpenRAIL License specifies: 1. You can't use the model to deliberately produce nor share illegal or harmful outputs or content 2. The authors claims no rights on the outputs you generate, you are free to use them and are accountable for their use which must not go against the provisions set in the license 3. You may re-distribute the weights and use the model commercially and/or as a service. If you do, please be aware you have to include the same use restrictions as the ones in the license and share a copy of the CreativeML OpenRAIL-M to all your users (please read the license entirely and carefully) Please read the full license here

> vhs glitch a skeleton news reporter wearing a black wig and suit, they live 1988, movie still cinematic MAKE SURE TO CHECK THE IMAGES ON THIS PAGE FOR PROMPTS AND SETTINGS TO GET GOOD RESULTS. This is A e t h e r G l it c h , an SDXL LoRA that simulates vhs tape glitches over subjects. Make sure to trigger the LoRA with vhs glitch , and play with weights to get more or less balanced glitch or crt tv looks. Looks cool on both photoreal and animated stuff. Gaming, movies and horror works well, and of course synthwave and such. Overall a great mood setter! Thanks to RunDiffusion who sponsored the finetuning of this LoRA. It was made on Lastben’s SDXL LoRA trainer through RunDiffusion which will soon be made public for everyone. Aether Glitch is of course available on their platform to try it out. Also thanks to Masslevel, Warlock and Adam D for contributing with awesome images. > vhs glitch man on a mountain looking at the sunset, cinematic, black background > vhs glitch cinematic 80s movie still of Margot Robbie, Barbie movie, intricate, atom bomb in the background > vhs glitch award-winning photo of kung fury battling a monster, cinematic horror > vhs glitch movie still of cyberpunk 2077 female character, cinematic, hires 8k, intricate, white background > vhs glitch movie still of by 50s Jennifer Connelly jazz singer at a dark club, Dark City 1998, cinematic, heavy grain > vhs glitch movie still of cyberpunk 2077 black character, cinematic, hires 8k, intricate, white background > vhs glitch cinematic movie still of Cillian Murphy wearing a 40s fedora > vhs glitch found footage of horror girl in the woods, cinematic, hires 8k, intricate, black background > vhs glitch movie still of cyberpunk 2077 character with cigar in mouth, cinematic, hires 8k, intricate, white background

EMBEDDING DOWNLOAD LINK - This is the Double Exposure Embedding for SD 2.x. It was trained with the v2.1 768 ema pruned model on 768px images of people, layered with a variety of surroundings. It's been shown to handle objects and animals good as well. 1. Place the file in the embeddings folder of your Automatic1111 installation. 2. Trigger the effect in the prompt by writing dblx. Left image: a profile photo of a woman and a city, photo by dblx Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 1965858791, Size: 1088x768, Model hash: 4bdfc29c Right image: woman photo closeup soap bubbles, photo by dblx Negative prompt: deformed, ugly Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 3612934306, Size: 1152x768, Model hash: 4bdfc29c All prompts and settings for the other images can be found in this document. This embedding was trained locally with Automatic1111's webui. The previous Double Exposure models for Stable Diffusion 1.5 can be found here. Made by masslevel: woman rising from ocean cloud glass smiling geisha woman with ocean + flower hair woman with galaxy hair Made by spaablauw aka Stille Willem: diamond mountain fox This model is open access and available to all, with a CreativeML OpenRAIL-M license further specifying rights and usage. The CreativeML OpenRAIL License specifies: 1. You can't use the model to deliberately produce nor share illegal or harmful outputs or content 2. The authors claims no rights on the outputs you generate, you are free to use them and are accountable for their use which must not go against the provisions set in the license 3. You may re-distribute the weights and use the model commercially and/or as a service. If you do, please be aware you have to include the same use restrictions as the ones in the license and share a copy of the CreativeML OpenRAIL-M to all your users (please read the license entirely and carefully) Please read the full license here

🥊 Aether Punch – Face Impact LoRA for Wan 2.2 5B (i2v) Aether Punch is a custom LoRA trained to animate a cinematic face punch — a single boxing glove appearing from the left and striking the subject. - 📦 Base model: Wan 2.2 5B (image-to-video) - 🖼️ Input resolution: 768×768 (square) - 🕒 Clip length: 5 seconds (121 frames) - 🎥 FPS: 24 - ⚙️ Steps: 20 - 📏 CFG: 5 - 🎯 Optimized for: human subjects (portraits) > A single boxing glove appears from the left, punching the face Start with a square image (768x768) of a single person, preferably front-facing or 3/4 view, without gloves or action already present — the animation will add the punch effect. Here are a few example videos showing the effect in action: - punchexample01.mp4 - punchexample02.mp4 - punchexample03.mp4 - punchexample04.mp4 This LoRA was trained with clean, centered portrait clips showing the moment of impact from a boxing glove or fist. It works best with minimal prompts and clear subject definition. Created by Joachim Sallström – part of the Aether series of expressive video LoRAs.

💦 Aether Splash – Water Balloon Impact LoRA for Wan 2.2 5B (i2v) Aether Splash is a custom LoRA trained to animate a cinematic slow-motion water balloon impact — a single balloon bursting across the subject’s face with full liquid detail. - 📦 Base model: Wan 2.2 5B (image-to-video) - 🖼️ Input resolution: 768×768 (square) - 🕒 Clip length: 5 seconds (121 frames) - 🎥 FPS: 24 - ⚙️ Steps: 20 - 📏 CFG: 5 - 🎯 Optimized for: human subjects (portraits) Use a square (768×768) or landscape image of a single person, preferably front-facing or in a 3/4 view, without any water balloons or splashes — the animation will add the impact effect. A clean, well-lit portrait works best. Here are a few example videos showing the splash effect in action: - splashexample01.mp4 - splashexample02.mp4 - splashexample03.mp4 This LoRA was trained with clean, centered portraits showing the moment of liquid impact. Best used with minimal prompts and clear subject framing for maximum effect. Also available on Civitai: https://civitai.com/models/1852834/aether-splash-wan-22-5b-i2v-lora And TensorArt: https://tensor.art/models/895667821210524878/Aether-Splash-Wan-2.2-5b-i2v-LoRA-v1.0 Created by Joachim Sallström – part of the Aether series of expressive video LoRAs.