This LoRA is designed to bring more motion and detail to Hyvid hetero sex scenes. It should be used with other motion LoRAs, using straightforward natural language prompts. Consult the prompting for LoRAs you combine with this one.
This LoRA improves other sex LoRAs in the following ways:
- More motion when combining with character LoRAs.
- Larger range of motions and facial expressions
- Better lighting and environments
- Higher detail of bodies and anatomy
- Crisper outputs at higher resolutions - lower the strength of other LoRAs if you get blurriness
Tips
If you get jitter or sudden movements, try turning up flow shift. Going from the standard 9 to 17 can improve fluidity, especially when combining character LoRAs.
Adding the Boreal LoRA at 0.2 - 0.3 strength can improve hands, feet, and other fine edges.
The simplest way to use this is to start with an existing generation that you want to add more motion or detail to, then add this LoRA at 0.2, 0.3, etc using the same seed until you find the sweet spot. The amount of strength needed is dependent on the other LoRAs, so you will need to experiment a bit.
When using as a standalone, you'll get very different results in widescreen vs. vertical aspect ratios. The training data is all widescreen and includes things like camera pans and zooms, and also a large diversity of angles. Widescreen seems capable of better results but also takes more work to get the prompts right.
Training Notes
v0.1 and v0.2 were trained on a large diversity of short PiV clips, but the resolution of the originals was not great so there were lots of artifacts.
In v0.3 I started over from scratch with 5 videos, each 4k and about 30 minutes long. I used diffusion-pipe to make frame buckets of 121 frames @ 244px and 33 frames @ 512px, using the overlapping_middle method so that the model saw the entire video twice, but with different samples. The idea was to give it higher res with short term motions and then help it understand long term motions with the 244px samples. I made a mistake having one video feature an actress's face more than the others, and it's overfit to her face a bit. The LR was 5e-5 and took about 7k steps.
v0.4 trained on top of 0.3 using 8 new videos of about 1hr each, also 4k sources, this time focusing on 1st person POV. Since most other sex LoRAs are 1st person I wanted it to improve when combining with them. The sources are all 16:9 aspect ratio so they perform a bit better in widescreen. I reduced the learning rate to 1e-5 to prevent any major forgetting, and the total steps was about 12k. So it has seen about 19k individual samples from 13 videos so far.
An important thing to note if is that there were no captions of any sort, so it does not come with any new text biases. The idea is more to fill Hunyuan's pretraining gap, which seems to have not included any video porn (hence why it can do porn stills but not motion).
I have done one fine-tune on top of this (the anal version) just to see if it can pick up on new concepts quickly. This was also trained without captions using one video, and it definitely picked up on the concept quite well just through the video examples. My hunch is that fine tunes should perform even better with source diversity and captions.
Version Notes
V0.4
This version is trained on 1st person POV sex scenes with a goal of improving the motions and understanding of sex from a male's POV. The model does have some understanding of 3rd person POV as well but it's currently less developed. This version will work best in combination with 1st person POV LoRAs such as Missionary, Cowgirl, Reverse Cowgirl, etc.
Widescreen aspect ratios will yield more fluid and interesting results. The training data is all 16:9 ratio 4k videos, and so using that aspect ratio is going to get you closer to the source material. Using vertical aspect ratio will push you more toward other LoRAs in your stack that use more of those ARs.
This version does still suffer from some sameface from getting overbiased by a long previous video at too high of a learning rate. This version reduces that but it will take more training to completely neutralize that. The next version will focus on face details, facial expressions, and diversity.
There are example workflows in several of the gallery posts. The main strategy is to use an existing sex LoRA at about half its normal strength, then combine with this LoRA at 0.5 to 1.5 strength. You will have to experiment a bit, especially when combining image-trained character LoRAs.
V0.3
This version is trained differently than previous versions, using higher res samples and greater diversity of scenes. It currently has a bias toward skinny blondes but can be prompted away from that.
It works quite well on its own as well, just use descriptive prompts to get the pose right and play with the LoRA strength and guidance levels. I have had success using character LoRAs trained on images only up to 1.3 strength with almost no loss of motion.
It works great with vid2vid, so for those looking for more interesting poses, that is a good option.
V0.3 - Anal Edition
This is an even more experimental version trained on one anal video. This is mostly to serve as a proof of concept that the base model can be guided by specific trainings, which opens the door to more specific variants. There is a workflow embedded in the gallery video to reproduce it in Comfy. Sorry for the lack of prompt guidance, I've been forgetting to set save meta to true when re-using past workflows.
Description
This version trains on a variety of POV sex videos. The goal is to increase position diversity, body movements, and facial expressions. This version still suffers from residual sameface, so use at lower strengths to find the sweet spot.
FAQ
Comments (30)
I don't get it, is v0.4 an upgrade from 01, 02, 03 or are they individual "helpers"? since you labeled them with different positions.
It's an upgrade, and also has a specific focus for that version. 0.1 - 0.3 were focused more on general sex motions, this is focused more on 1st person POV. I will continue to train and refine it over time, adding different capabilities. The idea is that it gradually gets better the more training data it sees, but rather than wait until it's perfect to release it to the world, I'm releasing it in stages so people can try it out and give feedback. The next version will focus on better faces.
@blonkus each version is trained on different dataset or you are just adding new data into next version (thinking about merge)
@bartoszaugul Each version sees entirely new data. No sample is shown more than once in any version (to prevent overfitting). Merging should be possible, I haven't tried yet! Curious to know how that goes for folks.
@blonkus can you point how big was dataset for each of the versions? will play with merging it and this info might be helpful. also, which tool have u used to train it?
@bartoszaugul v0.1 and v0.2 were trained on a large diversity of short PiV clips, but the resolution of the originals was not great so there were lots of artifacts. In v0.3 I started over with 5 videos, each 4k and about 30 minutes long. I used diffusion-pipe to make frame buckets of 121 frames @ 244px and 33 frames @ 512px, using the overlapping_middle method so that the model saw the entire video twice, but with different samples. The idea was to give it higher res with short term motions and then help it understand long term motions with the 244px samples. I made a mistake having one video feature an actress's face more than the others, and it's overfit to her face a bit. The LR was 5e-5 and took about 7k steps.
v0.4 trained on top of 0.3 using 8 new videos of about 1hr each, also 4k sources, this time focusing on 1st person POV. Since most other sex LoRAs are 1st person I wanted it to improve when combining with them. The sources are all 16:9 aspect ratio so they perform a bit better in widescreen. I reduced the learning rate to 1e-5 to prevent any major forgetting, and the total steps was about 12k. So it has seen about 19k individual samples from 13 videos so far.
An important thing to note if it wasn't clear is that there are no captions, so it does not come with any existing text biases. The idea is more to fill Hunyuan's pretraining gap, which seems to have not included any video porn (hence why it can do porn stills but not motion).
I have done one fine-tune on top of this (the anal version) just to see if it can pick up on new concepts quickly. This was also trained without captions using one video, and it definitely picked up on the concept quite well. My hunch is that fine tunes should perform even better with source diversity and captions.
@blonkus Thanks for response, what about alpha and dim? Also you have used kohya (musubi) or other scripts? it is important while picking up the script for merging, right now playing with svd merger but output is having 1kb so need to play a bit more ;)
@bartoszaugul Ah yeah good questions. Double blocks only. Rank/Dim 32, Alpha 32 (diffusion-pipe forces them to be equal). I haven't tried musubi, mostly because I am just now figuring out how to make DP do what I want!
@blonkus merged your loras (4, didnt use the anal one) with alphas of .25 for each (used different script thank kohya as the svd did not work - it is caused because it was trained on different script). running some tests but with addition with my model LoRA trained on pictures so there is not much motion. If you want the file to play with/publish I'll put it in my gdrive, send me private message.
This is great, you seem to have nailed some excellent training techniques.
@Suppressor Appreciate that from you! Love your work!
@blonkus Thank you!
@blonkus Quick question, can this do standalone yet or is it still focused on enhancement? The low rez training materials on some other NSFW loras really messes up my image quality, looking for a replacement.
@Suppressor It can do a small amount of standalone, but it's best at enhancement still since other LoRAs are trained with text prompts. Once I think it's reached the limit of its abilities based on the small rank (32) I want to explore fine tuning it for specific positions, body types, etc. My goal is to make a base model that is very easy to fine tune for specific needs once it's seen a jillion NSFW scenes.
If you were to pick three things you'd want a general NSFW model be capable of, what would they be?
@blonkus I lean toward traditional tastes, so I’d go with camera-aware missionary, doggy, and cowgirl—just my preference. Training for different angles could make it stand out and feel more polished. Most NSFW stuff out there is pretty limited in scope and not especially well-executed, so that’s where I see an opportunity. If other creators’ usage rights allow, mixing in some additional loras with your high-quality training might lead to solid results. I’ve messed around with merging Hunyuan models, and it’s shown some potential.
@Suppressor We have similar tastes! Something I'd like to incorporate is camera and scene control - zooms, pans, transitions, lighting, composition. Something to give it a more professional and artistic feel rather than just smut (I also love smut, but that gets old after a while). I'm also interested in collaborating with creators who want to use my model as a base for further training or as a merge.
@blonkus Might want to reach out to the guy that did the Matrix bullet time lora, that is masterful camera angle training.
Why doesnt 0.4 have any meta? This is important to always add this so we can get close to generations. it is very difficult when ther is no data to start from. Great work, just no data! i am sure you will correct this!
Some of the gens have meta - sorry about that! v0.5 will include meta for all gens.
@blonkus you are the man! These are the best out there ! keep doing you!
Getting this error in Comfy
HyVideoVAELoader
Error(s) in loading state_dict for AutoencoderKLCausal3D: Missing key(s) in state_dict: "encoder.down_blocks.0.resnets.0.norm1.weight", "encoder.down_blocks.0.resnets.0.norm1.bias", "encoder.down_blocks.0.resnets.0.conv1.conv.weight", "encoder.down_blocks.0.resnets.0.conv1.conv.bias", "encoder.down_blocks.0.resnets.0.norm2.weight", "encoder.down_blocks.0.resnets.0.norm2.bias", "encoder.down_blocks.0.resnets.0.conv2.conv.weight", "encoder.down_blocks.0.resnets.0.conv2.conv.bias", "encoder.down_blocks.0.resnets.1.norm1.weight", "encoder.down_blocks.0.resnets.1.norm1.bias", "encoder.down_blocks.0.resnets.1.conv1.conv.weight", "encoder.down_blocks.0.resnets.1.conv1.conv.bias", "encoder.down_blocks.0.resnets.1.norm2.weight", "encoder.down_blocks.0.resnets.1.norm2.bias", "encoder.down_blocks.0.resnets.1.conv2.conv.weight", "encoder.down_blocks.0.resnets.1.conv2.conv.bias", "encoder.down_blocks.0.downsamplers.0.conv.conv.weight", "encoder.down_blocks.0.downsamplers.0.conv.conv.bias", "encoder.down_blocks.1.resnets.0.norm1.weight", "encoder.down_blocks.1.resnets.0.norm1.bias", "encoder.down_blocks.1.resnets.0.conv1.conv.weight", "encoder.down_blocks.1.resnets.0.conv1.conv.bias", "encoder.down_blocks.1.resnets.0.norm2.weight", "encoder.down_blocks.1.resnets.0.norm2.bias", "encoder.down_blocks.1.resnets.0.conv2.conv.weight", "encoder.down_blocks.1.resnets.0.conv2.conv.bias", "encoder.down_blocks.1.resnets.0.conv_shortcut.conv.weight", "encoder.down_blocks.1.resnets.0.conv_shortcut.conv.bias", "encoder.down_blocks.1.resnets.1.norm1.weight", "encoder.down_blocks.1.resnets.1.norm1.bias", "encoder.down_blocks.1.resnets.1.conv1.conv.weight", "encoder.down_blocks.1.resnets.1.conv1.conv.bias", "encoder.down_blocks.1.resnets.1.norm2.weight", "encoder.down_blocks.1.resnets.1.norm2.bias", "encoder.down_blocks.1.resnets.1.conv2.conv.weight", "encoder.down_blocks.1.resnets.1.conv2.conv.bias", "encoder.down_blocks.1.downsamplers.0.conv.conv.weight", "encoder.down_blocks.1.downsamplers.0.conv.conv.bias", "encoder.down_blocks.2.resnets.0.norm1.weight", "encoder.down_blocks.2.resnets.0.norm1.bias", "encoder.down_blocks.2.resnets.0.conv1.conv.weight", "encoder.down_blocks.2.resnets.0.conv1.conv.bias", "encoder.down_blocks.2.resnets.0.norm2.weight", "encoder.down_blocks.2.resnets.0.norm2.bias", "encoder.down_blocks.2.resnets.0.conv2.conv.weight", "encoder.down_blocks.2.resnets.0.conv2.conv.bias", "encoder.down_blocks.2.resnets.0.conv_shortcut.conv.weight", "encoder.down_blocks.2.resnets.0.conv_shortcut.conv.bias", "encoder.down_blocks.2.resnets.1.norm1.weight", "encoder.down_blocks.2.resnets.1.norm1.bias", "encoder.down_blocks.2.resnets.1.conv1.conv.weight", "encoder.down_blocks.2.resnets.1.conv1.conv.bias", "encoder.down_blocks.2.resnets.1.norm2.weight", "encoder.down_blocks.2.resnets.1.norm2.bias", "encoder.down_blocks.2.resnets.1.conv2.conv.weight", "encoder.down_blocks.2.resnets.1.conv2.conv.bias", "encoder.down_blocks.2.downsamplers.0.conv.conv.weight", "encoder.down_blocks.2.downsamplers.0.conv.conv.bias", "encoder.down_blocks.3.resnets.0.norm1.weight", "encoder.down_blocks.3.resnets.0.norm1.bias", "encoder.down_blocks.3.resnets.0.conv1.conv.weight", "encoder.down_blocks.3.resnets.0.conv1.conv.bias", "encoder.down_blocks.3.resnets.0.norm2.weight", "encoder.down_blocks.3.resnets.0.norm2.bias", "encoder.down_blocks.3.resnets.0.conv2.conv.weight", "encoder.down_blocks.3.resnets.0.conv2.conv.bias", "encoder.down_blocks.3.resnets.1.norm1.weight", "encoder.down_blocks.3.resnets.1.norm1.bias", "encoder.down_blocks.3.resnets.1.conv1.conv.weight", "encoder.down_blocks.3.resnets.1.conv1.conv.bias", "encoder.down_blocks.3.resnets.1.norm2.weight", "encoder.down_blocks.3.resnets.1.norm2.bias", "encoder.down_blocks.3.resnets.1.conv2.conv.weight", "encoder.down_blocks.3.resnets.1.conv2.conv.bias", "encoder.mid_block.attentions.0.group_norm.weight", "encoder.mid_block.attentions.0.group_norm.bias", "encoder.mid_block.attentions.0.to_q.weight", "encoder.mid_block.attentions.0.to_q.bias", "encoder.mid_block.attentions.0.to_k.weight", "encoder.mid_block.attentions.0.to_k.bias", "encoder.mid_block.attentions.0.to_v.weight", "encoder.mid_block.attentions.0.to_v.bias", "encoder.mid_block.attentions.0.to_out.0.weight", "encoder.mid_block.attentions.0.to_out.0.bias", "encoder.mid_block.resnets.0.norm1.weight", "encoder.mid_block.resnets.0.norm1.bias", "encoder.mid_block.resnets.0.conv1.conv.weight", "encoder.mid_block.resnets.0.conv1.conv.bias", "encoder.mid_block.resnets.0.norm2.weight", "encoder.mid_block.resnets.0.norm2.bias", "encoder.mid_block.resnets.0.conv2.conv.weight", "encoder.mid_block.resnets.0.conv2.conv.bias", "encoder.mid_block.resnets.1.norm1.weight", "encoder.mid_block.resnets.1.norm1.bias", "encoder.mid_block.resnets.1.conv1.conv.weight", "encoder.mid_block.resnets.1.conv1.conv.bias", "encoder.mid_block.resnets.1.norm2.weight", "encoder.mid_block.resnets.1.norm2.bias", "encoder.mid_block.resnets.1.conv2.conv.weight", "encoder.mid_block.resnets.1.conv2.conv.bias", "encoder.conv_norm_out.weight", "encoder.conv_norm_out.bias", "decoder.up_blocks.0.resnets.0.norm1.weight", "decoder.up_blocks.0.resnets.0.norm1.bias", "decoder.up_blocks.0.resnets.0.conv1.conv.weight", "decoder.up_blocks.0.resnets.0.conv1.conv.bias", "decoder.up_blocks.0.resnets.0.norm2.weight", "decoder.up_blocks.0.resnets.0.norm2.bias", "decoder.up_blocks.0.resnets.0.conv2.conv.weight", "decoder.up_blocks.0.resnets.0.conv2.conv.bias", "decoder.up_blocks.0.resnets.1.norm1.weight", "decoder.up_blocks.0.resnets.1.norm1.bias", "decoder.up_blocks.0.resnets.1.conv1.conv.weight", "decoder.up_blocks.0.resnets.1.conv1.conv.bias", "decoder.up_blocks.0.resnets.1.norm2.weight", "decoder.up_blocks.0.resnets.1.norm2.bias", "decoder.up_blocks.0.resnets.1.conv2.conv.weight", "decoder.up_blocks.0.resnets.1.conv2.conv.bias", "decoder.up_blocks.0.resnets.2.norm1.weight", "decoder.up_blocks.0.resnets.2.norm1.bias", "decoder.up_blocks.0.resnets.2.conv1.conv.weight", "decoder.up_blocks.0.resnets.2.conv1.conv.bias", "decoder.up_blocks.0.resnets.2.norm2.weight", "decoder.up_blocks.0.resnets.2.norm2.bias", "decoder.up_blocks.0.resnets.2.conv2.conv.weight", "decoder.up_blocks.0.resnets.2.conv2.conv.bias", "decoder.up_blocks.0.upsamplers.0.conv.conv.weight", "decoder.up_blocks.0.upsamplers.0.conv.conv.bias", "decoder.up_blocks.1.resnets.0.norm1.weight", "decoder.up_blocks.1.resnets.0.norm1.bias", "decoder.up_blocks.1.resnets.0.conv1.conv.weight", "decoder.up_blocks.1.resnets.0.conv1.conv.bias", "decoder.up_blocks.1.resnets.0.norm2.weight", "decoder.up_blocks.1.resnets.0.norm2.bias", "decoder.up_blocks.1.resnets.0.conv2.conv.weight", "decoder.up_blocks.1.resnets.0.conv2.conv.bias", "decoder.up_blocks.1.resnets.1.norm1.weight", "decoder.up_blocks.1.resnets.1.norm1.bias", "decoder.up_blocks.1.resnets.1.conv1.conv.weight", "decoder.up_blocks.1.resnets.1.conv1.conv.bias", "decoder.up_blocks.1.resnets.1.norm2.weight", "decoder.up_blocks.1.resnets.1.norm2.bias", "decoder.up_blocks.1.resnets.1.conv2.conv.weight", "decoder.up_blocks.1.resnets.1.conv2.conv.bias", "decoder.up_blocks.1.resnets.2.norm1.weight", "decoder.up_blocks.1.resnets.2.norm1.bias", "decoder.up_blocks.1.resnets.2.conv1.conv.weight", "decoder.up_blocks.1.resnets.2.conv1.conv.bias", "decoder.up_blocks.1.resnets.2.norm2.weight", "decoder.up_blocks.1.resnets.2.norm2.bias", "decoder.up_blocks.1.resnets.2.conv2.conv.weight", "decoder.up_blocks.1.resnets.2.conv2.conv.bias", "decoder.up_blocks.1.upsamplers.0.conv.conv.weight", "decoder.up_blocks.1.upsamplers.0.conv.conv.bias", "decoder.up_blocks.2.resnets.0.norm1.weight", "decoder.up_blocks.2.resnets.0.norm1.bias", "decoder.up_blocks.2.resnets.0.conv1.conv.weight", "decoder.up_blocks.2.resnets.0.conv1.conv.bias", "decoder.up_blocks.2.resnets.0.norm2.weight", "decoder.up_blocks.2.resnets.0.norm2.bias", "decoder.up_blocks.2.resnets.0.conv2.conv.weight", "decoder.up_blocks.2.resnets.0.conv2.conv.bias", "decoder.up_blocks.2.resnets.0.conv_shortcut.conv.weight", "decoder.up_blocks.2.resnets.0.conv_shortcut.conv.bias", "decoder.up_blocks.2.resnets.1.norm1.weight", "decoder.up_blocks.2.resnets.1.norm1.bias", "decoder.up_blocks.2.resnets.1.conv1.conv.weight", "decoder.up_blocks.2.resnets.1.conv1.conv.bias", "decoder.up_blocks.2.resnets.1.norm2.weight", "decoder.up_blocks.2.resnets.1.norm2.bias", "decoder.up_blocks.2.resnets.1.conv2.conv.weight", "decoder.up_blocks.2.resnets.1.conv2.conv.bias", "decoder.up_blocks.2.resnets.2.norm1.weight", "decoder.up_blocks.2.resnets.2.norm1.bias", "decoder.up_blocks.2.resnets.2.conv1.conv.weight", "decoder.up_blocks.2.resnets.2.conv1.conv.bias", "decoder.up_blocks.2.resnets.2.norm2.weight", "decoder.up_blocks.2.resnets.2.norm2.bias", "decoder.up_blocks.2.resnets.2.conv2.conv.weight", "decoder.up_blocks.2.resnets.2.conv2.conv.bias", "decoder.up_blocks.2.upsamplers.0.conv.conv.weight", "decoder.up_blocks.2.upsamplers.0.conv.conv.bias", "decoder.up_blocks.3.resnets.0.norm1.weight", "decoder.up_blocks.3.resnets.0.norm1.bias", "decoder.up_blocks.3.resnets.0.conv1.conv.weight", "decoder.up_blocks.3.resnets.0.conv1.conv.bias", "decoder.up_blocks.3.resnets.0.norm2.weight", "decoder.up_blocks.3.resnets.0.norm2.bias", "decoder.up_blocks.3.resnets.0.conv2.conv.weight", "decoder.up_blocks.3.resnets.0.conv2.conv.bias", "decoder.up_blocks.3.resnets.0.conv_shortcut.conv.weight", "decoder.up_blocks.3.resnets.0.conv_shortcut.conv.bias", "decoder.up_blocks.3.resnets.1.norm1.weight", "decoder.up_blocks.3.resnets.1.norm1.bias", "decoder.up_blocks.3.resnets.1.conv1.conv.weight", "decoder.up_blocks.3.resnets.1.conv1.conv.bias", "decoder.up_blocks.3.resnets.1.norm2.weight", "decoder.up_blocks.3.resnets.1.norm2.bias", "decoder.up_blocks.3.resnets.1.conv2.conv.weight", "decoder.up_blocks.3.resnets.1.conv2.conv.bias", "decoder.up_blocks.3.resnets.2.norm1.weight", "decoder.up_blocks.3.resnets.2.norm1.bias", "decoder.up_blocks.3.resnets.2.conv1.conv.weight", "decoder.up_blocks.3.resnets.2.conv1.conv.bias", "decoder.up_blocks.3.resnets.2.norm2.weight", "decoder.up_blocks.3.resnets.2.norm2.bias", "decoder.up_blocks.3.resnets.2.conv2.conv.weight", "decoder.up_blocks.3.resnets.2.conv2.conv.bias", "decoder.mid_block.attentions.0.group_norm.weight", "decoder.mid_block.attentions.0.group_norm.bias", "decoder.mid_block.attentions.0.to_q.weight", "decoder.mid_block.attentions.0.to_q.bias", "decoder.mid_block.attentions.0.to_k.weight", "decoder.mid_block.attentions.0.to_k.bias", "decoder.mid_block.attentions.0.to_v.weight", "decoder.mid_block.attentions.0.to_v.bias", "decoder.mid_block.attentions.0.to_out.0.weight", "decoder.mid_block.attentions.0.to_out.0.bias", "decoder.mid_block.resnets.0.norm1.weight", "decoder.mid_block.resnets.0.norm1.bias", "decoder.mid_block.resnets.0.conv1.conv.weight", "decoder.mid_block.resnets.0.conv1.conv.bias", "decoder.mid_block.resnets.0.norm2.weight", "decoder.mid_block.resnets.0.norm2.bias", "decoder.mid_block.resnets.0.conv2.conv.weight", "decoder.mid_block.resnets.0.conv2.conv.bias", "decoder.mid_block.resnets.1.norm1.weight", "decoder.mid_block.resnets.1.norm1.bias", "decoder.mid_block.resnets.1.conv1.conv.weight", "decoder.mid_block.resnets.1.conv1.conv.bias", "decoder.mid_block.resnets.1.norm2.weight", "decoder.mid_block.resnets.1.norm2.bias", "decoder.mid_block.resnets.1.conv2.conv.weight", "decoder.mid_block.resnets.1.conv2.conv.bias", "decoder.conv_norm_out.weight", "decoder.conv_norm_out.bias". Unexpected key(s) in state_dict: "encoder.down.0.block.0.conv1.conv.bias", "encoder.down.0.block.0.conv1.conv.weight", "encoder.down.0.block.0.conv2.conv.bias", "encoder.down.0.block.0.conv2.conv.weight", "encoder.down.0.block.0.norm1.bias", "encoder.down.0.block.0.norm1.weight", "encoder.down.0.block.0.norm2.bias", "encoder.down.0.block.0.norm2.weight", "encoder.down.0.block.1.conv1.conv.bias", "encoder.down.0.block.1.conv1.conv.weight", "encoder.down.0.block.1.conv2.conv.bias", "encoder.down.0.block.1.conv2.conv.weight", "encoder.down.0.block.1.norm1.bias", "encoder.down.0.block.1.norm1.weight", "encoder.down.0.block.1.norm2.bias", "encoder.down.0.block.1.norm2.weight", "encoder.down.0.downsample.conv.conv.bias", "encoder.down.0.downsample.conv.conv.weight", "encoder.down.1.block.0.conv1.conv.bias", "encoder.down.1.block.0.conv1.conv.weight", "encoder.down.1.block.0.conv2.conv.bias", "encoder.down.1.block.0.conv2.conv.weight", "encoder.down.1.block.0.nin_shortcut.conv.bias", "encoder.down.1.block.0.nin_shortcut.conv.weight", "encoder.down.1.block.0.norm1.bias", "encoder.down.1.block.0.norm1.weight", "encoder.down.1.block.0.norm2.bias", "encoder.down.1.block.0.norm2.weight", "encoder.down.1.block.1.conv1.conv.bias", "encoder.down.1.block.1.conv1.conv.weight", "encoder.down.1.block.1.conv2.conv.bias", "encoder.down.1.block.1.conv2.conv.weight", "encoder.down.1.block.1.norm1.bias", "encoder.down.1.block.1.norm1.weight", "encoder.down.1.block.1.norm2.bias", "encoder.down.1.block.1.norm2.weight", "encoder.down.1.downsample.conv.conv.bias", "encoder.down.1.downsample.conv.conv.weight", "encoder.down.2.block.0.conv1.conv.bias", "encoder.down.2.block.0.conv1.conv.weight", "encoder.down.2.block.0.conv2.conv.bias", "encoder.down.2.block.0.conv2.conv.weight", "encoder.down.2.block.0.nin_shortcut.conv.bias", "encoder.down.2.block.0.nin_shortcut.conv.weight", "encoder.down.2.block.0.norm1.bias", "encoder.down.2.block.0.norm1.weight", "encoder.down.2.block.0.norm2.bias", "encoder.down.2.block.0.norm2.weight", "encoder.down.2.block.1.conv1.conv.bias", "encoder.down.2.block.1.conv1.conv.weight", "encoder.down.2.block.1.conv2.conv.bias", "encoder.down.2.block.1.conv2.conv.weight", "encoder.down.2.block.1.norm1.bias", "encoder.down.2.block.1.norm1.weight", "encoder.down.2.block.1.norm2.bias", "encoder.down.2.block.1.norm2.weight", "encoder.down.2.downsample.conv.conv.bias", "encoder.down.2.downsample.conv.conv.weight", "encoder.down.3.block.0.conv1.conv.bias", "encoder.down.3.block.0.conv1.conv.weight", "encoder.down.3.block.0.conv2.conv.bias", "encoder.down.3.block.0.conv2.conv.weight", "encoder.down.3.block.0.norm1.bias", "encoder.down.3.block.0.norm1.weight", "encoder.down.3.block.0.norm2.bias", "encoder.down.3.block.0.norm2.weight", "encoder.down.3.block.1.conv1.conv.bias", "encoder.down.3.block.1.conv1.conv.weight", "encoder.down.3.block.1.conv2.conv.bias", "encoder.down.3.block.1.conv2.conv.weight", "encoder.down.3.block.1.norm1.bias", "encoder.down.3.block.1.norm1.weight", "encoder.down.3.block.1.norm2.bias", "encoder.down.3.block.1.norm2.weight", "encoder.mid.attn_1.k.bias", "encoder.mid.attn_1.k.weight", "encoder.mid.attn_1.norm.bias", "encoder.mid.attn_1.norm.weight", "encoder.mid.attn_1.proj_out.bias", "encoder.mid.attn_1.proj_out.weight", "encoder.mid.attn_1.q.bias", "encoder.mid.attn_1.q.weight", "encoder.mid.attn_1.v.bias", "encoder.mid.attn_1.v.weight", "encoder.mid.block_1.conv1.conv.bias", "encoder.mid.block_1.conv1.conv.weight", "encoder.mid.block_1.conv2.conv.bias", "encoder.mid.block_1.conv2.conv.weight", "encoder.mid.block_1.norm1.bias", "encoder.mid.block_1.norm1.weight", "encoder.mid.block_1.norm2.bias", "encoder.mid.block_1.norm2.weight", "encoder.mid.block_2.conv1.conv.bias", "encoder.mid.block_2.conv1.conv.weight", "encoder.mid.block_2.conv2.conv.bias", "encoder.mid.block_2.conv2.conv.weight", "encoder.mid.block_2.norm1.bias", "encoder.mid.block_2.norm1.weight", "encoder.mid.block_2.norm2.bias", "encoder.mid.block_2.norm2.weight", "encoder.norm_out.bias", "encoder.norm_out.weight", "decoder.mid.attn_1.k.bias", "decoder.mid.attn_1.k.weight", "decoder.mid.attn_1.norm.bias", "decoder.mid.attn_1.norm.weight", "decoder.mid.attn_1.proj_out.bias", "decoder.mid.attn_1.proj_out.weight", "decoder.mid.attn_1.q.bias", "decoder.mid.attn_1.q.weight", "decoder.mid.attn_1.v.bias", "decoder.mid.attn_1.v.weight", "decoder.mid.block_1.conv1.conv.bias", "decoder.mid.block_1.conv1.conv.weight", "decoder.mid.block_1.conv2.conv.bias", "decoder.mid.block_1.conv2.conv.weight", "decoder.mid.block_1.norm1.bias", "decoder.mid.block_1.norm1.weight", "decoder.mid.block_1.norm2.bias", "decoder.mid.block_1.norm2.weight", "decoder.mid.block_2.conv1.conv.bias", "decoder.mid.block_2.conv1.conv.weight", "decoder.mid.block_2.conv2.conv.bias", "decoder.mid.block_2.conv2.conv.weight", "decoder.mid.block_2.norm1.bias", "decoder.mid.block_2.norm1.weight", "decoder.mid.block_2.norm2.bias", "decoder.mid.block_2.norm2.weight", "decoder.norm_out.bias", "decoder.norm_out.weight", "decoder.up.0.block.0.conv1.conv.bias", "decoder.up.0.block.0.conv1.conv.weight", "decoder.up.0.block.0.conv2.conv.bias", "decoder.up.0.block.0.conv2.conv.weight", "decoder.up.0.block.0.nin_shortcut.conv.bias", "decoder.up.0.block.0.nin_shortcut.conv.weight", "decoder.up.0.block.0.norm1.bias", "decoder.up.0.block.0.norm1.weight", "decoder.up.0.block.0.norm2.bias", "decoder.up.0.block.0.norm2.weight", "decoder.up.0.block.1.conv1.conv.bias", "decoder.up.0.block.1.conv1.conv.weight", "decoder.up.0.block.1.conv2.conv.bias", "decoder.up.0.block.1.conv2.conv.weight", "decoder.up.0.block.1.norm1.bias", "decoder.up.0.block.1.norm1.weight", "decoder.up.0.block.1.norm2.bias", "decoder.up.0.block.1.norm2.weight", "decoder.up.0.block.2.conv1.conv.bias", "decoder.up.0.block.2.conv1.conv.weight", "decoder.up.0.block.2.conv2.conv.bias", "decoder.up.0.block.2.conv2.conv.weight", "decoder.up.0.block.2.norm1.bias", "decoder.up.0.block.2.norm1.weight", "decoder.up.0.block.2.norm2.bias", "decoder.up.0.block.2.norm2.weight", "decoder.up.1.block.0.conv1.conv.bias", "decoder.up.1.block.0.conv1.conv.weight", "decoder.up.1.block.0.conv2.conv.bias", "decoder.up.1.block.0.conv2.conv.weight", "decoder.up.1.block.0.nin_shortcut.conv.bias", "decoder.up.1.block.0.nin_shortcut.conv.weight", "decoder.up.1.block.0.norm1.bias", "decoder.up.1.block.0.norm1.weight", "decoder.up.1.block.0.norm2.bias", "decoder.up.1.block.0.norm2.weight", "decoder.up.1.block.1.conv1.conv.bias", "decoder.up.1.block.1.conv1.conv.weight", "decoder.up.1.block.1.conv2.conv.bias", "decoder.up.1.block.1.conv2.conv.weight", "decoder.up.1.block.1.norm1.bias", "decoder.up.1.block.1.norm1.weight", "decoder.up.1.block.1.norm2.bias", "decoder.up.1.block.1.norm2.weight", "decoder.up.1.block.2.conv1.conv.bias", "decoder.up.1.block.2.conv1.conv.weight", "decoder.up.1.block.2.conv2.conv.bias", "decoder.up.1.block.2.conv2.conv.weight", "decoder.up.1.block.2.norm1.bias", "decoder.up.1.block.2.norm1.weight", "decoder.up.1.block.2.norm2.bias", "decoder.up.1.block.2.norm2.weight", "decoder.up.1.upsample.conv.conv.bias", "decoder.up.1.upsample.conv.conv.weight", "decoder.up.2.block.0.conv1.conv.bias", "decoder.up.2.block.0.conv1.conv.weight", "decoder.up.2.block.0.conv2.conv.bias", "decoder.up.2.block.0.conv2.conv.weight", "decoder.up.2.block.0.norm1.bias", "decoder.up.2.block.0.norm1.weight", "decoder.up.2.block.0.norm2.bias", "decoder.up.2.block.0.norm2.weight", "decoder.up.2.block.1.conv1.conv.bias", "decoder.up.2.block.1.conv1.conv.weight", "decoder.up.2.block.1.conv2.conv.bias", "decoder.up.2.block.1.conv2.conv.weight", "decoder.up.2.block.1.norm1.bias", "decoder.up.2.block.1.norm1.weight", "decoder.up.2.block.1.norm2.bias", "decoder.up.2.block.1.norm2.weight", "decoder.up.2.block.2.conv1.conv.bias", "decoder.up.2.block.2.conv1.conv.weight", "decoder.up.2.block.2.conv2.conv.bias", "decoder.up.2.block.2.conv2.conv.weight", "decoder.up.2.block.2.norm1.bias", "decoder.up.2.block.2.norm1.weight", "decoder.up.2.block.2.norm2.bias", "decoder.up.2.block.2.norm2.weight", "decoder.up.2.upsample.conv.conv.bias", "decoder.up.2.upsample.conv.conv.weight", "decoder.up.3.block.0.conv1.conv.bias", "decoder.up.3.block.0.conv1.conv.weight", "decoder.up.3.block.0.conv2.conv.bias", "decoder.up.3.block.0.conv2.conv.weight", "decoder.up.3.block.0.norm1.bias", "decoder.up.3.block.0.norm1.weight", "decoder.up.3.block.0.norm2.bias", "decoder.up.3.block.0.norm2.weight", "decoder.up.3.block.1.conv1.conv.bias", "decoder.up.3.block.1.conv1.conv.weight", "decoder.up.3.block.1.conv2.conv.bias", "decoder.up.3.block.1.conv2.conv.weight", "decoder.up.3.block.1.norm1.bias", "decoder.up.3.block.1.norm1.weight", "decoder.up.3.block.1.norm2.bias", "decoder.up.3.block.1.norm2.weight", "decoder.up.3.block.2.conv1.conv.bias", "decoder.up.3.block.2.conv1.conv.weight", "decoder.up.3.block.2.conv2.conv.bias", "decoder.up.3.block.2.conv2.conv.weight", "decoder.up.3.block.2.norm1.bias", "decoder.up.3.block.2.norm1.weight", "decoder.up.3.block.2.norm2.bias", "decoder.up.3.block.2.norm2.weight", "decoder.up.3.upsample.conv.conv.bias", "decoder.up.3.upsample.conv.conv.weight".
Great job man. I've done a lot of testing now with it on and off and have found this lora to be essential for getting the best results.
Edit: After a lot of testing and isolation, I realized this lowers breast size on models with large breasts and has to be counter-balanced with the m4crom4sti4 lora to restore them. It would be nice if future loras included large breast size models.
Hi. What kind workflow is used for this lora?
I am just starting with comfy and HunyuanVideo
Look up models on this site and filter Hunyuan Video and Workflow type. I like JS61 one, it's got a lot of options and great workflow control where you can turn ON/OFF certain options easily without having to disconnect nodes.
https://civitai.com/models/1134115/hunyuan-yaw-64-yet-another-workflow-easy-t2v-i2vskyreels-v2v-audio-random-lora-preview-pause-upscale-multi-res-interpolateprompt-saveloadteacachenew-interface-fast
Stupid question probably..., but are there any sites I can generate these on that anyone has used successfully, without installing Stable Dif? My computer can't handle Stable. I know some sites host models to do online, but don't know any that do video, can anyone advise? I've only found sites that host image civit models :( Really want to test this out
Mage space can do it
god bless you. please keep killing hunyuan
What prompt can improve?? I add this lora, start with a typical cowgirl position scene, then the penis stay at outside without penetrating but just keep trembling.
very good model video lora I almost put it in allzq my generation video thanks to the creater good job ;) by the way if you could create a lora special video femboys ke time I train myself in this practice ;)