Search results

Okay, so here is the prompt used on every run of individual clips that were later combined: (bl0wj0b:1.2), POV cock in a woman's hand, thrusting in and out repeatedly, <lora: DR34ML4Y_I2V_14B_LOW.safetensors:1.0>, This is using the dual High/Low Noise DR34ML4Y_I2V_14B LoRAs with one of its...

This is FANTASTIC!

Wow forgot to mention some important details so I will do that now as to not come across as disingenuous. :D My comparative test I'm doing will be using just a straight Wan 2.2 Image-to-Video workflow to generate VideoA. The workflow will also save out every uncompressed frame of the output...

I found a 1 year old YouTube video of a workflow with a custom node pack the guy created to help build longer videos. I tried it, after working out how to install the node pack, and it basically works, but it takes the last frame of a generated video and uses it as the first frame of the next...

I only have a 12GB 3060 so I'd have to use H3 with certain versions and optimizations to even give that ago. I do plan to look into it. If you've been playing around with it, anything you'd be willing to share?

Hi, has anyone been playing around with LTX or Wan lately and have videos or pointers to share?

Has anyone tried the new MOSS Image model?

Saw the 404 on their pages then tried a search, this came up

I think he's talking about the stream captures? That's about the size for the StreamTape versions at least.

Here is an example of effective vs ineffective prompting. For this I used the same Wan 2.2 workflow with the same seed. The only difference is the length of detail in the prompt. The video at the left had this prompt: Japanese woman smiles and walks in place, as if she were on a treadmill...

These are AI, no code. You can tell if you really look at the images. For the first on here is what I spotted: The chair designs do no t match In the background there is a color difference in the same window and the square pattern of the window is inconsistent Her hand does not have five...

Another great thing about image generation are the tools available and whatever is to come! Over on the AI Video Generation thread I posted something showing how I injected a character I created into the real world. Basically took an existing screenshot from a video clip, used flux 2 klein to...

I just cursed myself to waste all my free time, at least for a bit. I had an idea and I know I'm not the first out there, but I thought about using some of the things I've learned on the way here to do something interesting. Something fun yet playfully sinister. I thought I'd try inviting...

Yeah, I think because her breasts are visibly (physically) larger and gravity-stricken, maybe the model gives more motion to them haha!

Tried a little physics experiment featuring some lovely ladies. :cool: https://mega.nz/embed/mJE0XL6A#7Oh15MACH1yHOt0L7RdP5k3v6p5sNjx9h4STuPl8eCw

I'm going to try to train a LoRA using Kohya_ss. For my use case, though, it seems because I want to train a LoRA for the model I like using, gonzalomoXLFluxPony_v40UnityXLDMD, the training should be very successful since the subject (woman) I'm using was created using that model. It would...

Has anyone else tried Image to Image generation? I'm really liking this method of using a Pony XL model for the initial image, which will be more on the anime (hentai) side, producing some pretty bombastic figures. Then when the second model comes in, keeping the denoise around 0.56 or so (for...

Wow you really did a purge off this stuff. :eek: The default workflow uses the full 14B high and low models, I've been using the light GGUF UNET models, that cuts the generation time down a lot, especially when using the lightning loras. Slight quality loss, but not all that noticeable.

Would you mind sharing your workflow JSON file here? You'll have to rename it as a TXT file, though, to attach it. I'm curious what it looks like and what it's doing.

I'm not familiar with that performer in your screenshot so I don't know the consistency of her look. Keep in mind if you throw a "bad" or fuzzy/blurry or grainy image at Wan as a starting point I'd imagine there will be some significant drift since it doesn't have details to keep things as...

Hey, pal! I was out of town until today and just missed the 3 day mark on your first ones. I was able to grab your second one. The first thing is your image is a non-standard vertical size, not even true 16:9 (or here not even 9:16 since it is vertical) and your video size is 640x480 which is...

I finally found an upscale workflow that didn't have OOM crashes. And it was a very unassuming video from an equally unassuming YT channel. Although, not sure this qualifies as a true upscale because the node that does the size increase is a simple resize node. But it works. This was also a...

SORRY I meant my FireFox extension Video DownloadHelper was able to grab them. BUT I also just fed the m3u8 link to JDownloader2 and it also brought it down in less than a minute.

IDM Video DownloadHelper seemed to grab them just fine, at least the few I tried.

I've only ever watched them on DMM. JAVLibrary used to hotlnk to them but that probably stopped after the IP geo-restrictions

If the LoRA you're after is of a specific looking (or just specific) person, just my opinion, but I would opt for working on the LoRA with an image model. That way I could use it to place that person in any environment I wanted via prompts, and then send the images through a video model like...

Here are some examples of what I'm talking about. First model that generates the initial image at 480p: Fantasm | Pony XL Second model used for the image-to-image at 1080p (after an upscale to 576p): GonzaLomo XL/Flux/Pony v4.0 Unity XL DMD The Pony model creates some fantastic hips, waists...

Hopefully not, friend! I know you don't trust Google that much, but using the Google AI search I asked: What is the minimum VRAM on a graphics card required to train a LoRA for a video model? The answer: The Local Minimum: 24 GB of VRAM (e.g., RTX 3090 / 4090, or the RTX 5090 with 32GB). To...

LOL Yes he would! Outstanding!

I think I'm going back to school...as in old-school...as in back when I first started getting into this stuff and I was paying for a service, the models were creating almost near outlandish versions of women with nice hips, thighs, breasts that were anime-exaggerated bodies. That's what I...

The last thing I did with H3 was add the action prompt, then an empty line, then started the new line with "Audio: " and described the audio. That didn't produce that beginning pop, but that could have just been the seed, perhaps But I didn't try any further. :oops:

I've given up ( for now! ) on MiniMax H3. I'll wait for a few more distilled versions. It's just to chunky for me, the video looses consistent color almost immediately...it took me a while to fix a pop or cracking sound at the beginning of any generation with sound/voice applied, and the...

I was surprised how nice it looked from such a low res source! SeedVR2. Likely newer versions out, but here is the ComfyUI workflow I currently use. Of course, rename to SeedVR2_4K_image_upscale.json

These are from two different DVDs, I was scouring the Duga site for screenshots to try out. The top title is MMBS-009 from Marrion https://duga.jp/ppv/marrion2-0212/ The bottom title is ASW-337 from Asfur https://duga.jp/ppv/spc-0348/

That is Yuri Honma :D

That's great!!! Thanks for sharing! Per your comment on the other thread, it is pretty cool how they are able to get audio in there as well. Certainly increases the model size and generation time, but, the goal is the outcome. I'm still tinkering with it. Great stuff!

Try DMM directly. I think they started with new titles and are hopefully making their way backwards to update covers. For example, here is the recent FLAV-426 releasing 8/22/2026: https://awsimgsrc.dmm.co.jp/pics_dig/digital/video/flav00426/flav00426pl.jpg And here is from what I could...

I'm just trying it today! I've seen lots of YouTube videos on how to use and what to use and everything else. Found a simple workflow I'm trying. If you have any generations you'd like to share, I'd be interested to see what you've been able to do so far. :D I'll see what I can come up...

Hi, friend, sorry for the late reply. Yes, used ComfyUI for. The models (like the Z-Image Turbo flavor I used) are what allow for natural language prompting. I tend to use that mostly to see which models respond better to it. Glad you've incorporated that into your methods! Always great to...

I used a NSFW variant of Z-Image Turbo and the prompt below to try to recreate the image @_2old_ shared back in July. I generated maybe 12 or so but this one had the cutest face and the smallest band aid across the nose. The model kept putting thicker or longer ones with this prompt. I didn't...

I'm so glad you're working on this. I wouldn't have the patience to create a checkpoint, but being able to create something that will help you create what you want has to be exciting!

OUTSTANDING! I've been trying to find time to try this out...now I'm going to have to do whatever it takes to try it. Great examples!

Please give visual examples of what this is...for those who don't know.

Did you check your Task Manager to make sure there isn't a ghost instance of it hanging or stuck coming up?

Just to try, can you give an example? I can take a shot at a deep dive :belajar:

Make sure you're using NoScript or something similar to avoid bad ads or unexpected popups! This link seems to show 571 albums. Likely not complete or full resolution I would think? https://www.xasiat.com/albums/tags/lovepop/

There are sites that have a bunch of these, not sure how far back or if they are original or full resolution. In gallery form.

I voted Download, but the kicker is I used to download from stream sites. Sometimes I still do if the Streamtape video hasn't finished encoding.

I'm experimenting with a Flux 2 Klein variant instead of Qwen. Interesting and promising results so far! So it's likely possible to create a virtual set, so to speak, by using a background and having the model produce different viewing angles from it...

Now trying my hand at a Flux 2 Klein variant with pretty good results. What these models can do is take an image and change things based on your prompt, like changing a color of an object or even removing or inserting objects. And it does pretty well at even slight change in the angle of the...