WARNING: Consider this a beta, it still needs a lot of work.Please follow me on X/Twitter: <a target="_blank" rel="ugc" href="https://x.com/tazmannner">https://x.com/tazmannner</a>I am now posting on there more often, you can get early previews of what I'm working on there. It means a lot to grow outside just here. Thank you.<h3 id="triggers-tips">Triggers / Tips</h3>For style : 2d anime style (always use this). (you can mix styles from other anime with this though with additional style keywords or style loras).For specific words, it was consistently trained on these terms:penis, vagina, vaginal, testicles, penetration, insert, breasts, bouncing, jiggle, anus, anal, buttocks, plopping sound (for both sucking and penetrating), cowgirl, reverse cowgirl, doggy style, missionary, deep throat, thrust (for thrusting hips), blow job, suck, cum, cumming, deep breathing, heavy breathing, moan, grunt, insert, pull out, tentacle, ass ripple (for jiggling ass cheeks)It is trained on sound: moan, grunt, use "plopping" sounds fucking / sucking sounds.Adding other NSFW or motion loras (which are trained on 3d/IRL) will make the generation look less 2d. Just a warning.Recommended use with <a target="_blank" rel="ugc" href="https://huggingface.co/Kijai/MiniMax-H3-experimental/blob/main/loras/minimax_h3_DMAD_4step_full_lora_avg_rank_39_bf16.safetensors">DMAD turbo lora</a> if you must use a speed up, 20 steps default is best though. 5 steps 0.8 strength on DMAD and 1 strength on this. Euler Beta57.<h2 id="v0.5">V0.5</h2>This went through an additional 9 training attempts! I wanted to apply the knowledge I've learned since I started this lora with a fresh method of training.I bumped up the # of screen captures from the video dataset to around 480 images and videos are the same 422 clips, though I changed the target FPS to 25 to match the source. And overall its trained on higher resolution.This was trained in three stages: Stage 1: Images only training for detail. Using circle stone adapter 1.8k steps (1376x768; 1024x1024) with circle stone labs image adapterStage 2: Images + video for motion. Same as before but videos at 832x480 124frames added in.Stage 3: Image + video + sound for audio. Switch to Ostris v3 adapter with Audio Layers setting (Attention; mlp;audio), video 0.20 strength and audio 1.0 strength on weights.Each time load from weights, with Automagicv3.I think though the video has drifted a bit, stage 2 had much better penetration. So probably expect a better fixed version soon if additional training will help. If not I may start over from Stage 1 again and do everything with just the Ostris v3 adapter since it can do Stage2/3 together most likely.Overall its more stable in the sense that it has less distortion on the coloring and lines. And I think concepts like bj and hj are stable, penetration suffers the most. Sound is good. Motion in terms of jiggle/bounce I think look quite nice. This is almost 100% upgrade from last version.<h2 id="v0.4">V0.4</h2>Brushed up the dataset a bit, added in around 10 new videos and like 150 or so new images. I trained the videos and images together, but I think the best way would've been to do images first then add the videos. I restarted training and ran from 0 zero this time instead of loading weights.I wasn't able to get around to fixing paizuri ("tit fuck") and anuses still don't really show up despite making up a large part of the dataset. I think overall its a bump up in improvement from the last version, whatever that version did this can do better now and genitals look better.I'm gonna train it a bit more and see if more steps will help too, as the 23k steps version wasn't all that bad. 19.5k (most stable) is what's uploaded here, and a few other checkpoints in optional files if you wanna test those out too. Samples provided are not cherry picked.<h2 id="v0.3">V0.3</h2>edit: added 8100 steps checkpoint, use that one, seems better. There is 9450 steps one which may work for i2v better but its not stable, 8.1k I like the best.Added back in general NSFW data, so vagina, cumming, anal etc. are back in the dataset. I also added in an extra 50 or so images, making the total now: 130 images and 412 videos. It can now do pretty much all the stuff the v0.1 could do but actually stable and working now.I didn't have much time to check the checkpoints but the 5k steps one seems stable enough. Some of the samples seem weird with the vagina if you are prompting insertion. That I think can be fixed with better prompting. Sometimes bj the lips dont attach but rerolling seed or adjusting prompt fixes. I'll see if more training steps will help or not.I'll work on another version to improve the vagina shape and try to get the anus to appear, but I think might take a little longer, I'm getting a little burnt out working on this so fast. And I finally feel the lora is in a good place now. Please post your examples and constructive feedback.Did limited tests with lightx 8 step lora, and it does work with this lora (1str light 0.75-0.8 my lora) but you need to play around with the strength of both loras otherwise it goes off the rails. I recommend to not use them as they effect style quite a bit, let me know good settings and I can test. All samples are generated with 20 steps fully no speed ups. I also recommend the euler beta57 scheduler but simple seems more stable now.<h2 id="v0.2">V0.2</h2>Note: Do not use euler simple otherwise it will be distorted, try euler beta57 (my favorite so far) or some other scheduler/sampler. Strength 1 recommended, maybe 0.75 if you feel distortion too strong but I keep it at 1. Not tested with lightx. I had to go back to the drawing board and start over. I pruned the dataset down to 105 high quality clips involving only hand jobs or blow jobs and truncated them down to 75max frames. I also removed all furry content to remove any confusion for how dicks should be shaped. Then I added in a dataset of uncensored 2d anime style penis images. I also converted my captions from LTX format to H3 format with the help of gemma4 12b. I trained the videos on 672x384 resolution and the images on 1024x1024 resolution. Batch 3 for images, 1 for videos. I adjusted h3_guidance_distillation_scale down from 4 to 3.5.By reducing frames, resolution, adding in images, I went from 20s/it to around 4s/it! And I reached a good enough check point at around 6k steps to conclude this approach is a success! So I will release this now, but understand that the model has drifted away from the 1300 clips base knowledge. Penetration is not working anymore including doggy style. But blow job, hand jobs, and penis shape has improved ten-fold.So the next step is to add back in more data and train again so all the concepts are back.There is a 5.7 and 6.1k steps version, 5.7k has less defined penis shape but seems more like normal size, 6.1k seems to make the penis bigger but better shape. I don't wanna train further because I want to add more data and concepts and start over.<h2 id="v0.1">V0.1</h2>Trained on 1300 clips, so there is a lot of variety beyond just these. Some furry stuff in there but not a ton.Write the prompts in h3 syntax, start with "2d anime style"Works for t2v and i2v, i2v is better you can play with the strength, but I like 1 strength on both. Not tested well with lightx loras. In I2V if the genitals are already in frame it will keep their shape better. I think it does blow jobs especially well for 2d. The base model cannot do blowjobs right for 2d. Not tested with lightx, these are generated with 20 steps on euler simple with comfykitchen optimization only. Everything else is native/kijai's nodes.<h3 id="training">Training</h3>This was trained on musubi fork on 1300 uncensored high quality hentai clips at 832x480 resolution 124 max frames. And it only reached around 2 epochs on this version. I will train more tonight and see if there is improvement. I think its under trained, but at the same time I had to revert to a earlier checkpoint to get the best result. In this state it's usable for certain things, especially with i2v so I wanted to put it out. While I maybe just rework everything.The dataset and captions were originally meant for an even bigger LTX lora, but H3 was released so I decided to pivot this to H3. This is more of a learning experiment for me. I used a few LLM's to do a first pass on the captions, and then by hand brushed up around 1600 captions, I took out 300 dataset pieces due to odd aspect ratios that I didn't wanna deal with.The size of the dataset is my biggest yet, and this is still not finished, consider it a beta/experimental. I'll do another round of training and if nothing improves, I'm going to start over from square one with a smaller dataset and fit everything around H3 instead of LTX.I trained 4 versions before landing on this. H3 is very sensitive. I did 5k steps on V1, then loaded from safetensors and did 2400 steps in V4. Genitals in t2v are still not good enough, but it's picked up the sex positions and motions that go along with them.<h3 id="plans-for-v2.0:">Plans for v2.0: </h3>Reduce dataset down to like 300-500 clips max. Rewrite captions in H3 format. Brush up on terminology ("plopping" is a bit awkward word, and I need to caption sucking and fucking sounds separately).<h3 id="big-thanks">Big Thanks</h3>Thanks to <a target="_blank" rel="ugc" href="https://civarchive.com/user/JonXL">JonXL</a> who helped me test and tinker with pretty much every version, not possible without his help. Also to everyone in discord Sulphur and Banodoco discord with their feedback and tips.

2D Anime Style NSFW LORA - H3

This went through an additional 9 training attempts! I wanted to apply the knowledge I've learned since I started this lora with a fresh method of training.I bumped up the # of screen captures from the video dataset to around 480 images and videos are the same 422 clips, though I changed the target FPS to 25 to match the source. And overall its trained on higher resolution.This was trained in three stages: Stage 1: Images only training for detail. Using circle stone adapter 1.8k steps (1376x768; 1024x1024) with circle stone labs image adapterStage 2: Images + video for motion. Same as before but videos at 832x480 124frames added in.Stage 3: Image + video + sound for audio. Switch to Ostris v3 adapter with Audio Layers setting (Attention; mlp;audio), video 0.20 strength and audio 1.0 strength on weights.Each time load from weights, with Automagicv3.I think though the video has drifted a bit, stage 2 had much better penetration. So probably expect a better fixed version soon if additional training will help. If not I may start over from Stage 1 again and do everything with just the Ostris v3 adapter since it can do Stage2/3 together most likely.Overall its more stable in the sense that it has less distortion on the coloring and lines. And I think concepts like bj and hj are stable, penetration suffers the most. Sound is good. Motion in terms of jiggle/bounce I think look quite nice. This is almost 100% upgrade from last version.

Files with SHA256: 2353de7443b9693a73d23124923b18eab3f0850a68fc32fd36876537c641a721