In the interest of honesty I will disclose that many of these pictures here have been cherry picked, hand-edited and re-generated. Now enjoy those fine gens and get this sick mix! Peace! ATTENTION: This model DOES NOT contain all my clothing baked in. Updated - SECO: SECO = Second-stage Engine Cutoff (I watch too many SpaceX launches!!) - am cutting this model off now, and there may be an ICBINP XL release, but will see what happens. Civitai is the ultimate hub for AI art generation. WD 1. Download the User Guide v4. 0 Status (Updated: Nov 14, 2023): - Training Images: +2300 - Training Steps: +460k - Approximate percentage of completion: ~58%. This should be used with AnyLoRA (that's neutral enough) at around 1 weight for the offset version, 0. Version 3: it is a complete update, I think it has better colors, more crisp, and anime. That name has been exclusively licensed to one of those shitty SaaS generation services. Use it at around 0. . For next models, those values could change. Trigger is arcane style but I noticed this often works even without it. I wanted to share a free resource compiling everything I've learned, in hopes that it will help others. Dreamlike Diffusion 1. Stable Diffusion Webui Extension for Civitai, to download civitai shortcut and models. Choose from a variety of subjects, including animals and. Select the custom model from the Stable Diffusion checkpoint input field Use the trained keyword in a prompt (listed on the custom model's page) Make awesome images! Textual Inversions Download the textual inversion Place the textual inversion inside the embeddings directory of your AUTOMATIC1111 Web UI instance 有问题/错误请及时联系 千秋九yuno779 修改，谢谢。 备用同步链接： Stable Diffusion 从入门到卸载 ② Stable Diffusion 从入门到卸载 ③ Civitai | Stable Diffusion 从入门到卸载 【中文教程】 前言 介绍说明 Stable D. To mitigate this, weight reduction to 0. How to use Civit AI Models. The official SD extension for civitai takes months for developing and still has no good output. It supports a new expression that combines anime-like expressions with Japanese appearance. And full tutorial on my Patreon, updated frequently. articles. The pursuit of perfect balance between realism and anime, a semi-realistic model aimed to ach. To reference the art style, use the token: whatif style. BeenYou - R13 | Stable Diffusion Checkpoint | Civitai. I used CLIP skip and AbyssOrangeMix2_nsfw for all the examples. If you like it - I will appreciate your support. I don't remember all the merges I made to create this model. Please let me know if there is a model where both "Share merges of this model" and "Use different permissions on merges" are not allowed. Motion Modules should be placed in the WebUIstable-diffusion-webuiextensionssd-webui-animatediffmodel directory. PEYEER - P1075963156. Once you have Stable Diffusion, you can download my model from this page and load it on your device. 1, FFUSION AI converts your prompts into captivating artworks. This checkpoint includes a config file, download and place it along side the checkpoint. Civitai is a website where you can browse and download lots of Stable Diffusion models and embeddings. lora weight : 0. Stable Diffusion:. To reproduce my results you MIGHT have to change these settings: Set "Do not make DPM++ SDE deterministic across different batch sizes. Architecture is ok, especially fantasy cottages and such. 增强图像的质量，削弱了风格。. The Civitai Link Key is a short 6 character token that you'll receive when setting up your Civitai Link instance (you can see it referenced here in this Civitai Link installation video). Posted first on HuggingFace. Facbook Twitter linkedin Copy link. still requires a. animatrix - v2. Even animals and fantasy creatures. If you gen higher resolutions than this, it will tile. Refined-inpainting. 5 for generating vampire portraits! Using a variety of sources such as movies, novels, video games, and cosplay photos, I've trained the model to produce images with all the classic vampire features like fangs and glowing eyes. You just drop the pose image you want into controlnet extensions dropzone (the one saying "start drawing") and select open pose as model. 1 recipe, also it has been inspired a little bit by RPG v4. Civitai is a platform where you can browse and download thousands of stable diffusion models and embeddings created by hundreds of. Browse snake Stable Diffusion models, checkpoints, hypernetworks, textual inversions, embeddings, Aesthetic Gradients, and LORAsThis model was trained on images from the animated Marvel Disney+ show What If. Action body poses. . As a bonus, the cover image of the models will be downloaded. 0 update 2023-09-12] Another update, probably the last SD upda. Hello my friends, are you ready for one last ride with Stable Diffusion 1. Highest Rated. Civitai Helper 2 also has status news, check github for more. 5. Expect a 30 second video at 720p to take multiple hours to complete with a powerful GPU. This might take some time. Huggingface is another good source though the interface is not designed for Stable Diffusion models. 5 and 10 CFG Scale and between 25 and 30 Steps with DPM++ SDE Karras. 5D/3D images) Steps : 30+ (I strongly suggest 50 for complex prompt) AnimeIllustDiffusion is a pre-trained, non-commercial and multi-styled anime illustration model. These are the concepts for the embeddings. The split was around 50/50 people landscapes. VAE: Mostly it is recommended to use the “vae-ft-mse-840000-ema-pruned” Stable Diffusion standard. Making models can be expensive. Noosphere - v3 | Stable Diffusion Checkpoint | Civitai. 1 version is marginally more effective, as it was developed to address my specific needs. . The Civitai Discord server is described as a lively community of AI art enthusiasts and creators. Worse samplers might need more steps. 45 | Upscale x 2. If you use Stable Diffusion, you probably have downloaded a model from Civitai. AI一下子聪明起来，目前好看又实用。 merged a real2. This is a fine-tuned Stable Diffusion model (based on v1. It's a more forgiving and easier to prompt SD1. This model works best with the Euler sampler (NOT Euler_a). This is a general purpose model able to do pretty much anything decently well from realistic to anime to backgrounds All the images are raw outputs. . 3. To utilize it, you must include the keyword " syberart " at the beginning of your prompt. Installation: As it is model based on 2. Cmdr2's Stable Diffusion UI v2. Some Stable Diffusion models have difficulty generating younger people. LoRAデータ使用時もTrigger Wordsをコピペする手間がかからないため、画像生成も簡単です。. Conceptually elderly adult 70s +, may vary by model, lora, or prompts. It allows users to browse, share, and review custom AI art models, providing a space for creators to showcase their work and for users to find inspiration. Merge everything. Originally posted by nousr on HuggingFaceOriginal Model Dpepteahand3. Sensitive Content. This model was finetuned with the trigger word qxj. The first version I'm uploading is a fp16-pruned with no baked vae, which is less than 2 GB, meaning you can get up to 6 epochs in the same batch on a colab. Although this solution is not perfect. Installation: As it is model based on 2. pt file and put in embeddings/. It is advisable to use additional prompts and negative prompts. If you see a NansException error, Try add --no-half-vae (causes slowdown) or --disable-nan-check (may generate black images) to the commandline arguments. Except for one. What kind of. The GhostMix-V2. Which equals to around 53K steps/iterations. 2. Use it with the Stable Diffusion Webui. The training resolution was 640, however it works well at higher resolutions. X. The Ally's Mix II: Churned. Thank you thank you thank you. Create stunning and unique coloring pages with the Coloring Page Diffusion model! Designed for artists and enthusiasts alike, this easy-to-use model generates high-quality coloring pages from any text prompt. 65 weight for the original one (with highres fix R-ESRGAN 0. More up to date and experimental versions available at: Results oversaturated, smooth, lacking detail? No. pth. 2. See HuggingFace for a list of the models. Sci Fi is probably where it struggles most but it can do apocalyptic stuff. Counterfeit-V3 (which has 2. Created by ogkalu, originally uploaded to huggingface. This model is capable of generating high-quality anime images. Research Model - How to Build Protogen ProtoGen_X3. This is a simple extension to add a Photopea tab to AUTOMATIC1111 Stable Diffusion WebUI. fixed the model. Afterburn seemed to forget to turn the lights up in a lot of renders, so have. You can swing it both ways pretty far out from -5 to +5 without much distortion. 5D RunDiffusion FX brings ease, versatility, and beautiful image generation to your doorstep. Requires gacha. Civitai Helper. Sampler: DPM++ 2M SDE Karras. The third example used my other lora 20D. This checkpoint includes a config file, download and place it along side the checkpoint. Colorfulxl is out! Thank you so much for the feedback and examples of your work! It's very motivating. art. Install the Civitai Extension: Begin by installing the Civitai extension for the Automatic 1111 Stable Diffusion Web UI. This extension allows you to manage and interact with your Automatic 1111 SD instance from Civitai, a web-based image editor. See the examples. 55, Clip skip: 2, ENSD: 31337, Hires upscale: 4. Look no further than our new stable diffusion model, which has been trained on over 10,000 images to help you generate stunning fruit art surrealism, fruit wallpapers, banners, and more! You can create custom fruit images and combinations that are both beautiful and unique, giving you the flexibility to create the perfect image for any occasion. The purpose of DreamShaper has always been to make "a. Silhouette/Cricut style. We couldn't solve all the problem (hence the beta), but we're close! We tested hundreds of SDXL prompts straight from Civitai. Sensitive Content. Provide more and clearer detail than most of the VAE on the market. 4 + 0. 5d, which retains the overall anime style while being better than the previous versions on the limbs, but the light and shadow and lines are more like 2. No results found. 5) trained on screenshots from the film Loving Vincent. Welcome to Stable Diffusion. This is good around 1 weight for the offset version and 0. Highres-fix (upscaler) is strongly recommended (using the SwinIR_4x,R-ESRGAN 4x+anime6B by myself) in order to not make blurry images. Android 18 from the dragon ball series. This model imitates the style of Pixar cartoons. 5D ↓↓↓ An example is using dyna. To exploit any of the vulnerabilities of a specific group of persons based on their age, social, physical or mental characteristics, in order to materially distort the behavior of a person pertaining to that group in a manner that causes or is likely to cause that person or another person physical or psychological harm; For any use intended to. . stable Diffusion models, embeddings, LoRAs and more. Things move fast on this site, it's easy to miss. This option requires more maintenance. It does portraits and landscapes extremely well, animals should work too. For example, “a tropical beach with palm trees”. NeverEnding Dream (a. To reproduce my results you MIGHT have to change these settings: Set "Do not make DPM++ SDE deterministic across different batch sizes. 1_realistic: Hello everyone! These two are merge models of a number of other furry/non furry models, they also have mixed in a lot. Please consider joining my. It can be used with other models, but. For v12_anime/v4. 5 with Automatic1111's checkpoint merger tool (Couldn't remember exactly the merging ratio and the interpolation method)About This LoRA is intended to generate an undressed version of the subject (on the right) alongside a clothed version (on the left). 5 version please pick version 1,2,3 I don't know a good prompt for this model, feel free to experiment i also have. Prepend "TungstenDispo" at start of prompt. Civitai is a great place to hunt for all sorts of stable diffusion models trained by the community. pth <. The lora is not particularly horny, surprisingly, but. This LoRA model was finetuned on an extremely diverse dataset of 360° equirectangular projections with 2104 captioned training images, using the Stable Diffusion v1-5 model. The effect isn't quite the tungsten photo effect I was going for, but creates. You can ignore this if you either have a specific QR system in place on your app and/or know that the following won't be a concern. Usually this is the models/Stable-diffusion one. Click Generate, give it a few seconds, and congratulations, you have generated your first image using Stable Diffusion! (you can track the progress of the image generation under the Run Stable Diffusion cell at the bottom of the collab notebook as well!) Click on the image, and you can right-click save it. . They are committed to the exploration and appreciation of art driven by. . A startup called Civitai — a play on the word Civitas, meaning community — has created a platform where members can post their own Stable Diffusion-based AI. This resource is intended to reproduce the likeness of a real person. 75, Hires upscale: 2, Hires steps: 40, Hires upscaler: Latent (bicubic antialiased) Most of the sample images are generated with hires. Some tips Discussion: I warmly welcome you to share your creations made using this model in the discussion section. Browse weapons Stable Diffusion models, checkpoints, hypernetworks, textual inversions, embeddings, Aesthetic Gradients, and LORAsA dreambooth-method finetune of stable diffusion that will output cool looking robots when prompted. Performance and Limitations. stable Diffusion models, embeddings, LoRAs and more. • 9 mo. Its main purposes are stickers and t-shirt design. You can view the final results with sound on my. v5. Instead, the shortcut information registered during Stable Diffusion startup will be updated. You can check out the diffuser model here on huggingface. Results are much better using hires fix, especially on faces. Mix of Cartoonish, DosMix, and ReV Animated. 20230529更新线1. veryBadImageNegative is a negative embedding trained from the special atlas generated by viewer-mix_v1. SafeTensor. The only thing V5 doesn't do well most of the time are eyes, if you don't get decent eyes try adding perfect eyes or round eyes to the prompt and increase the weight till you are happy. Activation words are princess zelda and game titles (no underscores), which I'm not gonna list, as you can see them from the example prompts. It is a Latent Diffusion Model that uses two fixed, pretrained text encoders ( OpenCLIP-ViT/G and CLIP-ViT/L ). Ohjelmiston on. 6. It tends to lean a bit towards BoTW, but it's very flexible and allows for most Zelda versions. art. 🎨. 5 weight. 结合 civitai. 103. All the images in the set are in png format with the background removed, making it possible to use multiple images in a single scene. This is a model trained with text encoder on about 30/70 SFW/NSFW art, primarily of realistic nature. 直接Civitaiを使わなくても、Web UI上でサムネイル自動取得やバージョン管理ができるようになります。. All models, including Realistic Vision. Included 2 versions, 1 for 4500 steps which is generally good, and 1 with some added input images for ~8850 steps, which is a bit cooked but can sometimes provide results closer to what I was after. mutsuki_mix. It is more user-friendly. Refined v11 Dark. . It can make anyone, in any Lora, on any model, younger. V7 is here. . Whether you are a beginner or an experienced user looking to study the classics, you are in the right place. Then go to your WebUI, Settings -> Stable Diffusion on the left list -> SD VAE, choose your downloaded VAE. Realistic Vision V6. Now I am sharing it publicly. Hey! My mix is a blend of models which has become quite popular with users of Cmdr2's UI. Explore thousands of high-quality Stable Diffusion models, share your AI-generated art, and engage with a vibrant community of creators. . The overall styling is more toward manga style rather than simple lineart. 5 ( or less for 2D images) <-> 6+ ( or more for 2. So, it is better to make comparison by yourself. The Stable Diffusion 2. This was trained with James Daly 3's work. 適用すると、キャラを縁取りしたような絵になります。. Very versatile, can do all sorts of different generations, not just cute girls. I'm just collecting these. 🎓 Learn to train Openjourney. Restart you Stable. Raising from the ashes of ArtDiffusionXL-alpha, this is the first anime oriented model I make for the XL architecture. It tends to lean a bit towards BoTW, but it's very flexible and allows for most Zelda versions. I tried to alleviate this by fine tuning the text-encoder using the class nsfw and sfw. For next models, those values could change. Originally shared on GitHub by guoyww Learn about how to run this model to create animated images on GitHub. 5 ( or less for 2D images) <-> 6+ ( or more for 2. To mitigate this, weight reduction to 0. Stable Diffusion Webui Extension for Civitai, to help you handle models much more easily. If you get too many yellow faces or you dont like. Non-square aspect ratios work better for some prompts. 5, but I prefer the bright 2d anime aesthetic. 4-0. If you want to get mostly the same results, you definitely will need negative embedding: EasyNegative, it's better to use it at 0. v1 update: 1. This model is very capable of generating anime girls with thick linearts. yaml). a. iCoMix - Comic style Mix! Thank you for all Reviews, Great Model/Lora Creator, and Prompt Crafter!!! Step 1: Make the QR Code. Anime Style Mergemodel All sample images using highrexfix + ddetailer Put the upscaler in the your "ESRGAN" folder ddetailer 4x-UltraSharp. Denoising Strength = 0. Overview. If you gen higher resolutions than this, it will tile the latent space. 有问题/错误请及时联系 千秋九yuno779 修改，谢谢。 备用同步链接： Stable Diffusion 从入门到卸载 ② Stable Diffusion 从入门到卸载 ③ Civitai | Stable Diffusion 从入门到卸载 【中文教程】 前言 介绍说明 Stable D. 增强图像的质量，削弱了风格。. Final Video Render. This checkpoint includes a config file, download and place it along side the checkpoint. 6 version Yesmix (original). In my tests at 512,768 resolution, the good image rate of the Prompts I used before was above 50%. Kenshi is my merge which were created by combining different models. Activation words are princess zelda and game titles (no underscores), which I'm not gonna list, as you can see them from the example prompts. Join us on our Discord: collection of OpenPose skeletons for use with ControlNet and Stable Diffusion. 5 weight. It DOES NOT generate "AI face". 0 (B1) Status (Updated: Nov 18, 2023): - Training Images: +2620 - Training Steps: +524k - Approximate percentage of completion: ~65%. HuggingFace link - This is a dreambooth model trained on a diverse set of analog photographs. When comparing civitai and stable-diffusion-ui you can also consider the following projects: ComfyUI - The most powerful and modular stable diffusion GUI with a. Shinkai Diffusion is a LORA trained on stills from Makoto Shinkai's beautiful anime films made at CoMix Wave Films. Browse nsfw Stable Diffusion models, checkpoints, hypernetworks, textual inversions, embeddings, Aesthetic Gradients, and LORAsRecommend: DPM++2M Karras, Clip skip 2 Sampler, Steps: 25-35+. It's a mix of Waifu Diffusion 1. Different models available, check the blue tabs above the images up top: Stable Diffusion 1. Browse controlnet Stable Diffusion models, checkpoints, hypernetworks, textual inversions, embeddings, Aesthetic Gradients, and LORAsSeeing my name rise on the leaderboard at CivitAI is pretty motivating, well, it was motivating, right up until I made the mistake of running my mouth at the wrong mod, didn't realize that was a ToS breach, or that bans were even a thing,. Add dreamlikeart if the artstyle is too weak. Robo-Diffusion 2. Realistic Vision V6. Follow me to make sure you see new styles, poses and Nobodys when I post them. Since I use A111. Works only with people. Therefore: different name, different hash, different model. You may further add "jackets"/ "bare shoulders" if the issue persists. Avoid anythingv3 vae as it makes everything grey. com/models/38511?modelVersionId=44457 的DDicon模型使用，生成玻璃质感web风格B端元素。 v1和v2版本建议对应使用，v1. 合并了一个real2. SynthwavePunk - V2 | Stable Diffusion Checkpoint | Civitai. 日本人を始めとするアジア系の再現ができるように調整しています。. Andromeda-Mix | Stable Diffusion Checkpoint | Civitai. Thank you for your support!CitrineDreamMix is a highly versatile model capable of generating many different types of subjects in a variety of styles. While we can improve fitting by adjusting weights, this can have additional undesirable effects. 5) trained on images taken by the James Webb Space Telescope, as well as Judy Schmidt. Reuploaded from Huggingface to civitai for enjoyment. In second edition, A unique VAE was baked so you don't need to use your own. 6/0. Stable Diffusion is a powerful AI image generator. Example: knollingcase, isometic render, a single cherry blossom tree, isometric display case, knolling teardown, transparent data visualization infographic, high-resolution OLED GUI interface display, micro-details, octane render, photorealism, photorealistic. . Latent upscaler is the best setting for me since it retains or enhances the pastel style. Space (main sponsor) and Smugo. The resolution should stay at 512 this time, which is normal for Stable Diffusion. How to use: A preview of each frame is generated and outputted to \stable-diffusion-webui\outputs\mov2mov-images\<date> if you interrupt the generation, a video is created with the current progress. That is because the weights and configs are identical. Civitai hosts thousands of models from a growing number of creators, making it a hub for AI art enthusiasts. 5 (512) versions: V3+VAE Same as V3 but with the added convenience of having a preset VAE baked in so you don't need to select that each time. Submit your Part 2 Fusion images here, for a chance to win $5,000 in prizes!Browse lora Stable Diffusion models, checkpoints, hypernetworks, textual inversions, embeddings, Aesthetic Gradients, and LORAsUPDATE DETAIL (中文更新说明在下面) Hello everyone, this is Ghost_Shell, the creator. This upscaler is not mine, all the credit go to: Kim2091 Official WiKi Upscaler page: Here License of use it: Here HOW TO INSTALL: Rename the file from: 4x-UltraSharp. 0+RPG+526, accounting for 28% of DARKTANG. Upscaler: 4x-Ultrasharp or 4X NMKD Superscale. SCMix_grc_tam | Stable Diffusion LORA | Civitai. 2版本时，可以. 8346 models. For v12_anime/v4. 本文档的目的正在于此，用于弥补并联. , "lvngvncnt, beautiful woman at sunset"). fix is needed for prompts where the character is far away in order to make decent images, it drastically improve the quality of face and eyes! Sampler: DPM++ SDE Karras: 20 to 30 steps. vae. 構図への影響を抑えたい場合は、拡張機能の「LoRA Block Weight」を使用して調整してください。. C:stable-diffusion-uimodelsstable-diffusion)Redshift Diffusion. Copy the file 4x-UltraSharp. Expect a 30 second video at 720p to take multiple hours to complete with a powerful GPU. 9). 2 and Stable Diffusion 1. posts. Saves on vram usage and possible NaN errors. Copy this project's url into it, click install. The Civitai model information, which used to fetch real-time information from the Civitai site, has been removed. Mix from chinese tiktok influencers, not any specific real person. 起名废玩烂梗系列，事后想想起的不错。. Supported parameters. Originally posted to HuggingFace by leftyfeep and shared on Reddit. 5 fine tuned on high quality art, made by dreamlike. Trained on AOM2 . In the tab, you will have an embedded Photopea editor and a few buttons to send the image to different WebUI sections, and also buttons to send generated content to the embeded Photopea. Download (1. Originally Posted to Hugging Face and shared here with permission from Stability AI. Refined_v10. Stable Diffusion Models, or checkpoint models, are pre-trained Stable Diffusion weights for generating a particular style of images. Plans Paid; Platforms Social Links Visit Website Add To Favourites. All the examples have been created using this version of. hopfully you like it ♥. Pony Diffusion is a Stable Diffusion model that has been fine-tuned on high-quality pony, furry and other non photorealistic SFW and NSFW images. Pixar Style Model. Please Read Description Important : Having multiple models uploaded here on civitai has made it difficult for me to respond to each and every comme. 3. bounties. 111 upvotes · 20 comments. . 41: MothMix 1. Civitai stands as the singular model-sharing hub within the AI art generation community. To use this embedding you have to download the file aswell as drop it into the "stable-diffusion-webuiembeddings" folder. bounties. The GhostMix-V2. このよう. Face restoration is still recommended. 0 is SD 1. 推荐设置：权重=0. Out of respect for this individual and in accordance with our Content Rules, only work-safe images and non-commercial use is permitted. IF YOU ARE THE CREATOR OF THIS MODEL PLEASE CONTACT US TO GET IT TRANSFERRED TO YOU! This is the fine-tuned Stable Diffusion model trained on screenshots from a popular animation studio. The purpose of DreamShaper has always been to make "a better Stable Diffusion", a model capable of doing everything on its own, to weave dreams. Even animals and fantasy creatures. Instead, use the "Tiled Diffusion" mode to enlarge the generated image and achieve a more realistic skin texture. Example images have very minimal editing/cleanup. Trang web cũng cung cấp một cộng đồng cho người dùng chia sẻ các hình ảnh của họ và học hỏi về AI Stable Diffusion. Version 4 is for SDXL, for SD 1. IF YOU ARE THE CREATOR OF THIS MODEL PLEASE CONTACT US TO GET IT TRANSFERRED TO YOU! This is the fine-tuned Stable Diffusion model trained on screenshots from a popular animation studio. Enter our Style Capture & Fusion Contest! Part 2 of our Style Capture & Fusion contest is running until November 10th at 23:59 PST. 45 GB) Verified: 14 days ago. Prohibited Use: Engaging in illegal or harmful activities with the model. Mistoon_Ruby is ideal for anyone who loves western cartoons and animes, and wants to blend the best of both worlds. ”. Use the LORA natively or via the ex. For more example images, just take a look at More attention on shades and backgrounds compared with former models ( Andromeda-Mix | Stable Diffusion Checkpoint | Civitai) Hands-fix is still waiting to be improved. SafeTensor. You can customize your coloring pages with intricate details and crisp lines. Step 3. AS-Elderly: Place at the beginning of your positive prompt at strength of 1. models. Just another good looking model with a sad feeling . Use the token lvngvncnt at the BEGINNING of your prompts to use the style (e. iCoMix - Comic style Mix! Thank you for all Reviews, Great Model/Lora Creator, and Prompt Crafter!!!Step 1: Make the QR Code. 🎨. This Stable diffusion checkpoint allows you to generate pixel art sprite sheets from four different angles. There are tens of thousands of models to choose from, across. character western art my little pony furry western animation. Civitai. Other upscalers like Lanczos or Anime6B tends to smoothen them out, removing the pastel-like brushwork. . KayWaii. Inspired by Fictiverse's PaperCut model and txt2vector script. Stable Diffusion (稳定扩散) 是一个扩散模型，2022年8月由德国CompVis协同Stability AI和Runway发表论文，并且推出相关程序。. .