The goal of this lora is to reproduce the video style similar to live wallpaper, for those who play league of legends remember the launcher opening videos, that's the goal, but you can also use it to create your lofi videos :D enjoy. [Wan2.2 TI2V 5B - Motion Optimized Edition] Trained on 51 curated videos (24fps, 96 frames) for 5,000 steps across 100 epochs with rank 48. Optimized specifically for Wan2.2's unified TI2V 5B dense model and high-compression VAE. My Workflow (It's not organized, the important thing is that it works hahaha): <a target="_blank" rel="ugc" href="https://www.patreon.com/posts/135689184">🎮 Live Wallpaper LoRA - Wan2.2 5B (Workflow) | Patreon</a> Loop Workflow: <a target="_blank" rel="ugc" href="https://civarchive.com/models/1931348?modelVersionId=2185956">WAN 2.2 5b WhiteRabbit InterpLoop - v1.0 - Hardline | Wan Video Workflows | Civitai</a>Trigger word: l1v3w4llp4p3r [Wan2.2 I2V A14B - Full Timestep Edition]Trained on 301 curated videos (256px, 16fps, 49 frames) for 24 hours using Diffusion Pipe with Automagic optimizer, rank 64. Uses extended timestep range (0-1) instead of standard (0-0.875), enabling compatibility with both Low and High models despite training only on Low model.Trigger word: l1v3w4llp4p3rWorks excellently with LightX2V v2 (256 rank) for faster inference [Wan I2V 720P Fast Fusion - 4 (or more) steps]Wan I2V 720P Fast Fusion combines 2 Live Wallpaper LoRA (1 Exclusive) with Lightx2v, AccVid, MoviiGen and Pusa LoRAs for ultra-fast 4+ steps generation while maintaining cinematic quality.🚀 <a target="_blank" rel="ugc" href="https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors">Lightx2v LoRA</a> – accelerates generation by 20x through 4-step distillation, enabling sub 2-minute videos on RTX 4090 with only 8GB VRAM requirements. 🎬 <a target="_blank" rel="ugc" href="https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan21_AccVid_T2V_14B_lora_rank32_fp16.safetensors">AccVid LoRA</a> – improves motion accuracy and dynamics for expressive sequences. 🌌 <a target="_blank" rel="ugc" href="https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan21_T2V_14B_MoviiGen_lora_rank32_fp16.safetensors">MoviiGen LoRA </a>– adds cinematic depth and flow to animation, enhancing visual storytelling. 🧠 <a target="_blank" rel="ugc" href="https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Pusa/Wan21_PusaV1_LoRA_14B_rank512_bf16.safetensors">Pusa LoRA </a>– provides fine-grained temporal control with zero-shot multi-task capabilities (start-end frames, video extension) while achieving 87.32% VBench score. 🧠 <a target="_blank" rel="ugc" href="https://huggingface.co/Wan-AI/Wan2.1-I2V-14B-720P">Wan I2V 720p (14B)</a> base model – providing strong temporal consistency and high-resolution outputs for expressive video scenes.[Wan I2V 720P]The dataset used consists of 149 videos (each one hand-selected) in 1280x720x96 resolution but was trained in 244p and 480p and 64 frames with 64 dim (L40s).Trigger word was used so it needs to be included in the prompt: l1v3w4llp4p3r[Hunyuan T2V]The dataset used consists of 529 videos (each one hand-selected) in 1280x720x96 resolution but was trained in 244p and 72 frames with 64 dim (multiple RTX 4090).No captions or activation words were used, the only control you will need to adjust is the lora strength.Another important note is that it was trained in full blocks, I don't know how it will behave when mixing 2 or more loras, if you want to mix and are not getting a good result, try disabling single blocks.I recommend using lora strength between 0.2 and 1.2 maximum, resolution 1280x720 or generate at 512 and upscale later, minimum 3 seconds (72 frames + 1). [LTXV I2V 13b 0.9.7 – Experimental v1]The model was trained on 140 curated videos (512px, 24fps, 49 frames), using 250 epochs, 32 dim, and AdamW8bit. It was trained using Diffusion Pipe with support for LTXV I2V v0.9.7 (13B). Captions were used and generated with Qwen2.5-VL-7B via a structured prompt format.This is an experimental first version, so expect some variability depending on seed and prompt detail.Recommended:Scheduler: sgm_uniformSampler: eulerSteps: 30<h2 id="long-prompts-are-highly-recommended-to-avoid-motion-artifacts.">⚠️ Long prompts are highly recommended to avoid motion artifacts.</h2>You can generate captions using the <a target="_new" rel="ugc" href="https://github.com/alisson-anjos/ComfyUI-Ollama-Describer">Ollama Describer</a> or optionally use the official LTXV Prompt Enhancer.For more details, see the About this version tab. ------------------------------------------------------------------------------------------------------For more details see the version description Share your results.

Live Wallpaper Style

Animate static artwork into subtle, seamless live wallpapers with MiniMax H3 Ref2VA. The LoRA supports one source image plus optional midpoint and ending-frame references for better motion continuity. Two versions are included:- R32: often preserves facial features more faithfully, but camera behavior can be less predictable.- R64: continued with camera-aware captions, offering better control over locked cameras, zooms, pans, tilts and subtle camera drift. It can animate hair, clothing, foliage, water, lighting, energy, petals, particles and other environmental effects while preserving the original subject and composition. Recommended strength: 1.0 When using multiple references, connect the same images in the same order to the MiniMax H3 Reference to Video node in both workflow passes. Always describe the desired camera behavior explicitly; use a locked-off camera when no movement is wanted. More Details Here: <a target="_blank" rel="ugc" href="https://huggingface.co/Alissonerdx/Minimax-H3-ComfyUI">Alissonerdx/Minimax-H3-ComfyUI · Hugging Face</a>

minimax_h3_live_wallpaper_v1.0_ref2va_r64.safetensors

minimax_h3_live_wallpaper_v1.0_ref2va_r32.safetensors

minimax_h3_live_wallpaper_workflow.json

Minimax H3 - Ref2VA

LTX 2.3 I2V v1.0

Wan2.2 TI2V 5B 720P

Wan2.2 I2V A14B

Wan2.1 I2V 720P 14B F. F.

LTXV 13b 0.9.7 I2V

Wan2.1 I2V 14B 480P

Wan2.1 I2V 14B 720P

Hunyuan T2V

Live Wallpaper Plus — Wan I2V 720P (Exclusive Tensor.Art)🧾 Model DescriptionThis LoRA is designed to recreate the dynamic and immersive style of live wallpapers, reminiscent of the League of Legends launcher intros. It is also perfect for lofi loops, cinematic ambient scenes, and AI-powered visual storytelling. 🎮🎧This is the second generation of the original LoRA Live Wallpaper, trained from scratch using a larger, hand-curated dataset sourced from 4K video. While the first version focused on general motion and parallax, this version places a special emphasis on particles, atmospheric effects, visual detail enhancements, camera moviments, making it ideal for dynamic loops with magical or cinematic ambiance. It is still in the development stage, and needs more time to train for at least 50 epochs, meaning this version is experimental.<a target="_blank" rel="ugc noopener noreferrer nofollow" href="http://Tensor.Art">Tensor.Art</a> limits the generation of the 720P model to 480P, which makes no sense to me, in other words, you are not generating with the best possible quality, it makes no sense to generate 480 in a model that was mostly trained with 720P data.🧩 Training &amp; Specs- Trigger word: `lvwprx` (must be included in the prompt)- Dataset: 575 carefully selected 4K videos, downscaled to 256p- Captions: focused on dynamic motion (e.g. hair, particles, parallax, ambient effects, camera moviments)- Frames per sample: 49- LoRA Dim: 64- Training resolution: 256p- Hardware: Trained using H100 (80GB) + RTX 5090 (32GB)Note: It took 3 days to tune the dataset, train and test hyperparameters, and finally evaluate the model. The considerations I have regarding the model are that I need a few more days until the version of this model is really 100% improved, this model needs to be trained for more epochs because I used 3 times more data than the previous model and in this case a smaller number of epochs, so based on this I will use the next few days to improve it.⚙️ Usage Guidelines- LoRA strength: 0.2 to 1.2 (default = 1.0)- Generation resolution: must be 720P (1280x720, 720x1280...) (native to Wan I2V 720P)- Input image (if used):- Should be exactly or as close as possible to 720P- Must be sharp and detailed — avoid blurry or compressed images- Pay special attention to fine details like hair, fabric, and background layers- Prompting tip: clearly describe intended motion, parallax, particles, hair flow, or ambient effects — this strongly guides the animation- Duration: between 49 and 81 frames (~3 to 5 seconds at 16 FPS)- Compatible with other LoRAs — tune strength and merging settings as needed✍️ Prompt Generation Template (Optional)----------------------------------------If you want to generate high-quality prompts for this LoRA, you can use the following template with any LLM, along with an input image. Your prompt should always be in tags format for this template.You must always wait for an input image before generating output.Do not generate captions or tags unless an image has been provided.You are a captioning assistant for video generation using anime-style models like Wan 2.1.Given an image, you must generate two outputs optimized for animated video generation with smooth loops and subtle movement.Your task is to output one line:🔹 1. Tag-style Caption (lvwprx+ compact tags)On the second line, list 8–14 tags separated by commas. These define motion, style, and environment.✅ Example:lvwrpx, anime style, glowing eyes, flowing scarf, rising embers, twilight sky, parallax effect, static camera, smooth animation, seamless loop⚠️ Important: Use only one of the captions above — either the sentence or the tags — never both together.🙏 Special Thanks---------------------Special thanks to users bellow, for providing high-quality images used in the showcase examples. Your support helped shape the visual direction of this model.<a target="_blank" rel="noopener noreferrer nofollow" href="https://tensor.art/u/844693610673855217">Litch</a><a target="_blank" rel="noopener noreferrer nofollow" href="https://tensor.art/u/66282218543757****">Semi</a><a target="_blank" rel="noopener noreferrer nofollow" href="https://tensor.art/u/864713822257042838">Edson Rodrigues</a>L9🧾 LTXV I2V 0.9.7 (13B) – Experimental v1This LoRA is trained for the LTXV I2V model version 0.9.7 (13B parameters), with the goal of generating smooth, seamless live wallpaper-style video loops with detailed, localized motion — such as hair fluttering, blinking, ambient particles, and parallax — while keeping rigid structures like armor, weapons, chairs, and body parts perfectly static.This is the first experimental version of the LoRA, and was trained on a modest but curated dataset using conservative hyperparameters and high-precision captioning. If better techniques are discovered (e.g., improved prompting, training without captions, or more stable convergence), I plan to release future, higher-quality versions.🔧 Training Details:<ul><li>Base Model: LTXV I2V v0.9.7 (13B)</li><li>Training Tool: <a target="_blank" rel="ugc noopener noreferrer nofollow" href="https://github.com/tdrussell/diffusion-pipe">Diffusion Pipe (unofficial LTXV I2V support)</a></li><li>Dataset: 140 handpicked videos, 512px resolution, 24fps, 49 frames per clip</li><li>Epochs: 250 (~35,000 steps)</li><li>LoRA Dim: 32</li><li>Optimizer: AdamW8bit</li><li>Batch Size: 1</li><li>Prompts: Used during training, generated via Qwen2.5-VL-7B with a strict motion-focused template</li><li>Initial Loss: High — LTXV I2V is known to require many steps before reaching stability</li></ul>🧠 Prompting Recommendation:This LoRA is sensitive to prompt quality and seed variation. Short prompts often lead to:<ul><li>Deformed or “rubber-like” rigid objects</li><li>Motion in areas that should remain static</li><li>Unstable background or scene jitter</li></ul>✅ Use long, structured prompts that clearly describe motion and stillness. For best results, use the custom ComfyUI node: <a target="_blank" rel="ugc noopener noreferrer nofollow" href="https://github.com/alisson-anjos/ComfyUI-Ollama-Describer">Ollama Describer</a> (supports structured motion captions via LLMs)Alternatively, the official LTXV Prompt Enhancer node can also help improve prompt effectiveness.✍️ Prompt Generation TemplateUse the following prompt with any vision-enabled LLM (like Qwen-VL, GPT-4o, or Gemini). It is designed to describe only what moves, and explicitly what must stay still, to avoid motion artifacts in rigid structures.Prompt Template:You are an expert in motion design for seamless animated loops. Given a single image as input, generate a richly detailed description of how it could be turned into a smooth, seamless animation. Your response must include: ✅ What elements should move: – Hair (e.g., swaying, fluttering) – Eyes (e.g., blinking, subtle gaze shifts) – Clothing or fabric elements (e.g., ribbons, loose parts reacting to wind or motion) – Ambient particles (e.g., dust, sparks, petals) – Light effects (e.g., holograms, glows, energy fields) – Floating objects (e.g., drones, magical orbs) if they are clearly not rigid or fixed – Background ambient motion (e.g., fog, drifting light, slow parallax) 🚫 And explicitly specify what should remain static: – Rigid structures (e.g., chairs, weapons, metallic armor) – Body parts not involved in subtle motion (e.g., torso, limbs unless there’s idle shifting) – Background elements that do not visually suggest movement ⚠️ Guidelines: – The animation must be fluid, consistent, and seamless, suitable for a loop – Do NOT include sudden movements, teleportation, scene transitions, or pose changes – Do NOT invent objects or effects not present in the image – Do NOT describe static features like colors, names, or environment themes – The output must begin with the trigger word: lvwpr – Return only the description (no lists, no markdown, no instructions)✅ Example output:lvwpr The character’s hair flows gently with ambient wind, while glowing particles drift across the background. Her eyes blink slowly, and her outfit’s ribbons sway softly. The armor plates, weapons, and mechanical chair remain perfectly still.⚙️ Usage Guidelines:<ul><li>Trigger word: none (prompt-based control only)</li><li>Suggested LoRA Strength: 0.8 – 1.2</li><li>Recommended Duration: 49–81 frames (~3–5 seconds)</li><li>Resolution: Generate at 512px or upscale later</li><li>Style: Anime, stylized, or cinematic images with clean structure work best</li></ul>

Live Wallpaper Plus

Animate static artwork into subtle, seamless live wallpapers with MiniMax H3 Ref2VA. The LoRA supports one source image plus optional midpoint and ending-frame references for better motion continuity.


Two versions are included:

- R32: often preserves facial features more faithfully, but camera behavior can be less predictable.

- R64: continued with camera-aware captions, offering better control over locked cameras, zooms, pans, tilts and subtle camera drift.

It can animate hair, clothing, foliage, water, lighting, energy, petals, particles and other environmental effects while preserving the original subject and composition.
Recommended strength: 1.0

When using multiple references, connect the same images in the same order to the MiniMax H3 Reference to Video node in both workflow passes. Always describe the desired camera behavior explicitly; use a locked-off camera when no movement is wanted.

More Details Here: https://huggingface.co/Alissonerdx/Minimax-H3-ComfyUI

Minimax H3 REF2VA

Wan I2V 720P

Files with SHA256: 7ea5e75ae1514583a714e10caa7b215c04061c0dc27594284a9d8274b3a53348