i cant get the sampler and image resize to download on your workflow.. i downloaded the whole RES4LYF folder but nothing is working... but otherwise this is amazing!

what is PP image in workflow? Picture profile image of the character for the final scene?

To test this in portrait view I just took a random portrait image and 512x768 to try it out. works well. I used 269 frames though so a little repeat due to the extra frames. Nice job on the lora. Should be posted in gallery unless Deadpool gets it banned.

Good job! Also great wf thanks you

How to generate these images? I only have the start image

Amazing LoRAs and workflow; thank you so so much!

Thanks for this. Your username checks out. Great work.

I've played around with it but I'm having some issues.

1. The workflow seems to drop random interpolated videos and doesn't save properly. Sometime the first one, sometimes the last one. Any idea what could be causing this?

2. When the workflow is finished my ComfyUI (v0.5.1) crashes. But I didn't try to update, maybe that's going to fix that.

Thanks again for sharing. It's great fun to use.

edit: Actually it saves the files correctly. Nevermind. It was my dumb ass not handling the file management correctly. 😁 And lowering the resolution also remidies the Comfy crash. 👍🏻

When trying to use portrait photos and changing the aspect ratio to be portrait using the workflow you provided, the auto-crop nodes crop the portrait photos too much. And it doesn't seem like anything prevents this. This didn't happen in your other Multiscene lora.

Good lora, great workflow. I learn so much from these, thank you for sharing.

I seem to be missing a step. How do I combine the videos? I have four videos in the interpolation step, looks like the same four were saved in the outputs folder, is this not supposed to put them together?

I like the concept, but boy does this require a shit ton of work. The workflow demands for multiple input images with different angles, this takes a lot of extra effort when you have just one input image to work with.

I would like to see some example images that you used for your successful generations, the crop zones are giving me oddly scaled outputs and I'm not sure if I need a different angle or zoom level for better results.

I'd say by far the weakest link of the work flow is the cumshot. If you try to add any description to the cumshot in the prompt, all of the clips will end up with cum, and not in a good way. We need a different positive for the fourth video, but I'm not sure if we can just add extra conditioning after the WanImageToVideo node.

Overall great job on the concept, it does work, but it's leaving a bit to be desired with the difficulty of setup.

that workflow is insane! Thank you so much for both the lora and the WF!

best creator here. we need a pov missionary lora with that quality

Only thing missing is one final shot of the subject looking directly at the camera after the cumshot. Would love to to see the facial aftermath

Wow, thanks, man....you're my idol!!!

If you don't mind, I have a request for a lora featuring doggy-style sex from different angles and/or cowgirl sex looking back and looking front

Awesome work!

this is one of the best lora and workflow i have come across , hope u can create same stuff for different , sex positions

I have no idea how to start this. I am able to get a combined picture of the 5 starting ones, but then I don't know what to do exactly

Love the lora and WF!

I wanted to play with the combine of all 4 videos at the end. I enabled the combiner you have disabled by default. It doesn't seem to do anything at all.

Any idea how to get it to work so that I can test out idea to keep comfy from crashing?

This is amazing! Really first time I got such high quality out of Wan! Are you a magician? Really great work and thank you so much! If I may ask something, is there anyway to share a short workflow that upgrades the quality of a lower resolution video, like you do in your current workflow? I would really appreciate it. Anyway, well done once more and thank you very much for sharing it with us!

Try a freememory node before Fl_imagebatch for stitching.

pretty please make a intercourse version!!!!!!

Can't wait to get this working. Unfortunately none of the crops do any cropping. They all say "Out of sync, run graph to get preview" but running doesn't make them crop either - or show the image in the crop box

Amazing workflow and Lora! Does anyone knows how to get final highest resolutions? it states that 1920 x 1080 is also possible but I get probs with anything higher than 1344 x 896. I have a 16 Vram card and 64 GB Ram. Thanks in advance!

the idea of your workflow sounds pretty amazing. Do you reckon it would work if I don't use the Lora and change the prompts to generate a different sequence ?

Using your WF with unchanged settings , there is consistently a kind of dark ghosting on the results.

It's from motion, so either the girl moves and it ghosts or the camera does and so the environment gets the ghosting.

I have tried everything I can think of to fix it: changed samplers, schedulers, steps, step split, shift, ETA, denoise, lightning loras, resolutions, tried different images, seeds, changed re-size methods. I have no idea what is causing it. I am typically using the max resolution for the low and high res as per your list.

It seems present on the initial low res pass but very slight, sometimes I can't tell if it's there, and then becomes worse on the high res pass. Tried HQ photoshoot images and also insta pics, same result. Any idea what could be the cause?

I had to modify the lora loader a bit to get the dick not to be a potato. Essentially, I added a multi lora loader for low and run Dr3amley v2 I2V @ .6 strength across the entire gen. Made a WORLD of difference and fixed my issues.

Great WF!!

how to combine all generated videos in one? I turn on last nodes, but they dont started.

Simply amazing work again. Thank you!

About the recombining of the image batches, I just use the KJNodes Image Batch Multi node and it's no issue to recombine the videos. I only do 1152 x 768 right now though because the generation alone takes too long if I go higher

Try merging the latents instead of the images. That will reduce your memory consumption by around 100x. If that still isn't enough, you can add tiled vae decode.

Is it your lora that somehow crops the images? If i use another lora like someone said, like the blink instead of your multiscene, i just get the preview image with all 5 pics moving as the video. This workflow confuses me how it uses the images. Any explanation would be helpful.

Love this, any tips on how to go about training something like this with such a long length? Would love to throw a few together for other actions but unsure on how to pass the normal 81 frames in training.

Why does it output a simple image file when I click run?

The more I play around with this lora, the more I am convinced that it is one of the best loras ever made. The overall intended motion is excellent and the character consistency is practically 100% without any remarkable changes to the face and clothing. And if you don't want any clothing, you just make the character nude. Even using it for a single scene instead of the multiscene concept is still unmatched compared with other similar loras. I would really love to see the training data and study the training parameters to apply it to my own first lora I am planning to train.

This workflow works well, but I have a problem: after running it successfully the first time, all image output nodes disconnect on the second run, rendering the workflow unusable. I don't know why. Could you please explain?

when using your workflow i get a ksampleradvanced error: returned non-zero exit status 1.

any ideas?

Does anyone know how to add cumshots? He says in the comments in the fourth section but I don't know what he means or what he's talking about. I don't see a place where it goes

anyonee know how to make it longer? every time i try i get some kind of error regarding expected frames....

Can you provide the picture collage that you use in your examples? The combined picture.

This is GREAT! Works like a charm, though it looks like the combining part doesn't work as it should. But no worry, the seperate parts can easily be combined with an external video editor. I'm planning to try your workflow as a multiscene penetration version, but don't really know what changes will be necessary. Thanks man!!!!

Is there a mirror to download the high noise file somewhere else? I'm getting the "error getting file" crap when trying to download the link. I got the low noise file fine. Ugh.

Just wow ! Did u want to make Cowgirl Missi or doggy Lora? Incredable ! THX