We're a Day-0 launch partner for MiniMax H3, a next-gen open-weights multimodal video model - and it's already in the generator. No waitlist. Open weights land in a few days.

MiniMax H3 is live: open-weights video, generate now

<span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">■This is an experimental ground for krea2 LoRAs.■If you notice pixel patterns in your generated images, try the VAE below. It removes the pixelation typical of the qwen_vae and makes the images much sharper.<a target="_blank" rel="ugc" href="https://civitai.red/models/2718533/hdr-vae-anima-qwen-image">https://civitai.red/models/2718533/hdr-vae-anima-qwen-image</a>■If the VAE looks a bit too saturated for your taste, try using this VAE merge node to blend it with the original qwen_vae to dial it in. A 0.5 merge should be just right.<a target="_blank" rel="ugc" href="https://civitai.red/models/2487550/vae-merger-for-comfyui-anima">https://civitai.red/models/2487550/vae-merger-for-comfyui-anima</a>

This is a style LoRA with a Kingdom Hearts vibe.The dataset consists of 24 images: 12 official illustrations plus 12 variations created through flipping and slight color adjustments. Because of this limited dataset, the overall variety is quite low.Since it was trained for quite a few epochs, it is heavily overfitted, which results in lower flexibility. It has a high tendency to generate two or more characters. Lowering the LoRA strength might help make it more flexible. Additionally, I saved checkpoints from several different steps in the backups, so you might want to look for the one that fits your preference best.The minimum is 500 steps, but since it was trained for around 80 epochs, it has a fairly strong effect. The 2,000-step version has very high style reproducibility, but at 300 epochs, it leans a bit towards being overfitted.

KingdomHearts_style

I created this using 1,600 carefully curated Digimon images. It mainly reproduces the style, atmosphere, and backgrounds of the TV series.I recommend adding the trigger word "digimon," as it was trained across the entire dataset.Please note that it was not trained on specific monster or character names. However, the auto-tagger may have included the tag "digimon (creature)" in some instances. So, while you can't generate specific monsters perfectly, you might still be able to capture their general characteristics.

Digimon_style

■This is a cute, cel anime-style LoRA. It was trained on a carefully curated dataset of 19,000 images from various classic cel anime titles.■Style Tendencies The overall style leans towards a "pop" aesthetic with large eyes. The tone is generally clear and bright, but it rarely becomes overly vivid like modern digital anime. Instead, you can easily achieve the calm colors and nostalgic atmosphere unique to traditional cel anime.■Trigger Tags &amp; Dataset Info Below are the anime titles used in the dataset. These names are also trained as style trigger tags, and the tag names are based on existing series tags on Danbooru. I've included brief notes on the style tendencies of each dataset. Please treat these as a general reference—they might not apply perfectly during every inference, but they serve as a useful guide. Some series naturally have a simpler overall style. If you want to generate images with more detail or dense information, I highly recommend using the tags for titles where I specifically praise the backgrounds. Or you might want to search the series names to get a sense of the overall vibe. I’ve also included the dataset resolutions for each title.Many of the "cute style" anime series were never released on Blu-ray, which limited what I could source, so some of the datasets rely on DVD-quality footage.You don't need to worry about this during inference, as I trained the model to ensure it doesn't negatively impact the output, but it can serve as a baseline indicator of the source quality. (BD sources were trained at 1024px, and DVD sources at 768px.)Please note that a lower resolution does not mean the anime itself is of low quality! Every single one of these shows features exceptionally high-quality animation that truly stood out in its era. ■taglist gunsmith cats [1024px] - An exquisite balance of a cute style and a hardboiled, calm tone. aozora shoujotai [768px] - Fantastic overall. Great backgrounds, lighting, fighter jets, and cute characters. taiho shichauzo (OVA + Movie) [1024px] - Convincing backgrounds, mechanical details, and stunning lighting. tenkuu no escaflowne (Movie) [1024px] - A bit quirky but highly artistic; the pinnacle of high-standard cel anime. cardcaptor sakura [1024px] - Achieves both a cute style and a calm tone with wonderful lighting. x (manga) (Movie) [1024px] - A fantastic work that perfectly blends the CLAMP style with a dark atmosphere. trigun (TV) [768px] - Amazing lighting and backgrounds that really make you feel the temperature and textures. di gi charat [768px] - The origin of otaku culture's "cute" style. Backgrounds are simple. slayers (OVA + Movie) [1024px] - The highly representative 90s style. detatoko princess [768px] - Similar lineage to Slayers. Great atmosphere and lighting. jungle de ikou [1024px] - Similar lineage to Slayers. Good backgrounds and a wide variety of character body types. hikyou tanken fam &amp; ihrie [768px] - Similar lineage to Slayers. A classic, orthodox fantasy style. A highly well-balanced style with very few quirks. dirty pair (Dirty Pair Flash OVA) [1024px] - A great mix of a cute style with mature body proportions. fatal fury (Movie) [1024px] - An exaggerated, quirky take on the 90s style. idol project [768px] - Also features an exaggerated, quirky 90s style. plastic little [768px] - The footage is older, but the core quality is wonderful, featuring great characters and backgrounds. sailor moon [1024px] - On a different wavelength from the others (more pop and colorful), but the composition and atmosphere are excellent. A certain sense of formal beauty and refined art. rurouni kenshin (Movie version, not the OVA) [1024px] - Beautiful depictions of Japanese scenery. Characters have cute, classic 90s large eyes. neon genesis evangelion (Death (True)² + EOE) [1024px] - The characters' eyes might be slightly smaller compared to the others.

retro anime_cute_style

■This is a Megaman X style LoRA, trained on 400 carefully selected images from "Megaman X DiVE" and "The Day of Σ" anime.Roughly 150 images from "Megaman X DiVE"About 250 images from "The Day of Σ" (official FHD video)Rather than a pure "style" LoRA, it might be more accurate to think of this as a LoRA trained specifically on the physical features of Megaman characters.■Features &amp; UsageRecommended Tags: Adding tags like android, robot, robot joints, joints ,etc... will really help bring out that classic Megaman character vibe.Visual Characteristics: It makes it easier to generate human-machine hybrid bodies. It also emphasizes a manufactured, metallic gloss on the body, and eyes tend to be drawn a bit larger. Since the DiVE sprites heavily use rim lighting, you'll probably see that effect a lot.■A Note on BackgroundsThis might be a bit divisive, but since the Megaman X DiVE dataset consisted entirely of character sprites on white backgrounds, I used "nanobanana" to add backgrounds to the images before training. I did this because training strictly on white backgrounds makes the model prone to generating unwanted white or overly simple backgrounds.I carefully considered what backgrounds would suit each character and used detailed AI prompts to generate them. Because of this, you might notice an "AI-generated" feel in the backgrounds, lighting, and overall style. If the style feels too strong or unnatural, lowering the LoRA weight/strength should help reduce the effect.

MEGA_MAN_X_Style

■This is a style LoRA focused on the aesthetics of cool, retro anime.It was trained on a dataset of 18,000 carefully selected images from what can be considered masterpieces of traditional cel animation. You can find the list of the included works below.■This LoRA was trained at 1024px.I trained this using booru-style tag captions.The number of epochs and steps might seem low compared to the size of the dataset. However, since I trained it with a Rank of 64 and an effective batch size of 16, I believe the learning process actually progressed quite well.Interestingly, when looking at the comparison images, it seems like the model picked up a few unknown concepts in addition to the overall style, which might provide some unexpected benefits.■depending on your seed or prompt, the LoRA's style might feel too weak. If that happens, please try increasing the weight. You will likely find a good sweet spot somewhere around 1.1 to 1.5.or, Since Krea2 is already capable of reproducing high-quality cel styles on its own, applying this LoRA at a lower weight for a hybrid look is also a great option.■Versatility:Because the dataset includes a wide range of genres—such as Sci-Fi, traditional Japanese style, and fantasy—this LoRA is highly flexible and can adapt to almost any genre of image you want to generate.■Character Genders:Although I personally only generate 1girl images, this is not a female-only LoRA. Anime naturally features male characters, and about half of the dataset consists of men, so generating male characters shouldn't be an issue.■About the Files &amp; BackupsTo stabilize the style, this main LoRA is a merge of several different training steps.I am also sharing the unmerged backup LoRAs that were used as ingredients for this merge.■List of Works: (The trigger tag does not exist.)Cowboy_Bebop_movie&amp;TVghost_in_the_shellRurouni_Kenshin_OVAMemoriespatlabor_movie_1.2.3Vampire_Hunter_DThe_Vision_of_Escaflowne_movieX (CLAMP 1996) movieMillennium_ActressGUNDAM_The 08th MS Team_OVAGUNDAM_0083: Stardust Memory_OVAGUNDAM_W Endless Waltz_movie■A Personal Note on the AnimeI truly believe the titles in this list represent the absolute pinnacle of traditional cel animation. The artwork in every single one is on par with fine art, and seeing it brought to life with such stunning animation is a truly rich visual experience. They are all fantastic masterpieces absolutely worth your time, so I highly recommend checking them out if you ever get the chance.

retoroanime_cool_style

■This is a style LoRA based on the anime masterpiece, Vampire Hunter D: Bloodlust.I trained this LoRA on a carefully selected dataset of 1,000 images.■This LoRA was trained at 1024px. However, since it is based on traditional cel animation, the output is not perfectly clean and naturally includes characteristic artifacts. If these artifacts bother you, lowering the LoRA weight might help.Here are the three main artifacts you might notice:<ul><li>Noise Patterns: You may see film or video noise patterns, especially in areas like the sky.</li><li>Edge Shadows: You might notice shadows along the edges of the characters. This is a unique feature of cel animation, caused by the character cel layer casting a shadow onto the background layer.</li><li>Soft Details: As is typical with many older film works, perfectly sharp scans or artificial sharpening are rare, which can sometimes result in softer or slightly blurry details.</li></ul>On the other hand, depending on your seed or prompt, the LoRA's style might feel too weak. If that happens, please try increasing the weight. You will likely find a good sweet spot somewhere around 1.1 to 1.5.■A Personal Note on the AnimeMadhouse's productions in the late 1990s represented the absolute pinnacle of cel animation, and among them, Vampire Hunter D: Bloodlust is a standout achievement. If you have never watched it, it is absolutely worth experiencing at least once.

krea2_lora_lab