[punchy]

Well... They released LTX-2.3 about 4 days ago.  It improves sound and prompt adherence.  Motion seems to be improved as well.  I am not going to deal with it as I am trying to get out of generating stuff, but for those of you interested...

DeepBeepMeep has something called Wan2GP that simplifies running these AI art models.  Qwen, LTX, WAN, Ace-Step, Flux, Z-Image, and more.  It requires some know-how to install as you have to use git and conda commands.

Here is a video on LTX-2.3 and half way through the video he walks you through installing Wan2GP.  Granted, this is only doing SFW material.  You really need to learn ComfyUI to do NSFW.

https://youtu.be/KaQbt6NYrzY?si=sj8J77pnuqwEtvVT

Punchy

[punchy]

Process.

I generate a picture from Chroma because I get the biggest boobs from this.

ŠPunchy

[punchy]

Here are my settings for Chroma in ComfyUI to generate that picture.

ŠPunchy

[punchy]

I crop the picture so I can do the "sexual action" without seeing her genitalia.  More POV that way.

ŠPunchy

[punchy]

Plug it into Phr00t's Qwen Image Edit and cleaned up the image to fix bad hands, remove extra stuff, etc.

ŠPunchy

[punchy]

Qwen prompt and settings for that.  I just love the plain language prompts and the adherence.  It didn't fix the zoom, but I will live with it. I didn't use the generation on this page, I selected the prior gen.

ŠPunchy

[punchy]

I went through all of that and now I am getting too many anomalies with slapping the image into Phr00t LTX-2 Rapid Merge.  Well, you get the idea.  It is a lot of generations to get to the final video.

Punchy

[punchy]

Ok.  I have deleted all of my AI generation tools except the SFW version of LTX-2 and my LLMs and Ace-Step (I like music). Good luck, folks.  I hope I taught you guys enough to create some new "artists".

Beware of 6/6/26.

Punchy

[wolpertinger]

Ok.  I have deleted all of my AI generation tools except the SFW version of LTX-2 and my LLMs and Ace-Step (I like music). Good luck, folks.  I hope I taught you guys enough to create some new "artists".

Beware of 6/6/26.

Punchy

what shall happen on 6, June 2026? Another y2k nonsense?

[punchy]

Hello, folks.  I have been honing my skills with AI, creating music and videos, using OpenShot, Audacity, various AI tools, etc, to create videos that are purely AI.  The music videos are so much fun, and I have produced a few bangers using AI, mostly Suno.  I tried Ace-Step, but the results are bupkis.

So I have honed and stabilized my AI installs at home.  I am using LTX-2.3 for video, I2V, and IS2V.  I still use Phr00t's version of Qwen Image Edit, just because of the simplicity of it, but I am likely to move on from that.  New tools I use include ImageID, ERNIE, and others.  It helps with face consistency across different scenes.

Anywho, I hope y'all are doing wonderful things with AI.  I am being forced to learn Claude at work.  They give us an allowance each month and I am going to take full advantage of it so I can learn AI skills to carry me through the end of my career. (I am an old man close to retirement age.)

Here is something I churned out in just a couple of minutes.  Now that I have this so honed, imagine what I could do if I ever returned here.   Again, I am hoping I sparked the creative bug in some of the younger BE fans.  BE is VERY alive on DeviantArt, but most of it is paid content.'

- You come home early from work and see your cleaning lady's car is still sitting in the driveway.  Having had a tough day at work, you walk in and head to your bedroom to change into your lounge clothes.  Your eyes bulge out of your head at what you see...

ŠPunchy

[punchy]

And I forgot how much this forum frustrates me.  No matter what format I use it doesn't work unless it is WEBP or an image format.  My LTX-2.3 videos have audio attached and to change the format to one of the others means stripping off the audio.  It is just a shame because I can actually make MP4 videos with audio that are smaller than those images I used to upload, but they just don't work with the forum.  I attached it as a WEBM, hoping it would work and it didn't.

Back into lurk mode.
Punchy

PS.  Here is the MEGA link to the same video as an MP4.
https://mega.nz/file/13wyTDQR#TXWjK8nnII5uUBGqQjnYCvepuL_K87bPVRfFjwSEsQc

[Bombayduk]

And I forgot how much this forum frustrates me.  No matter what format I use it doesn't work unless it is WEBP or an image format.  My LTX-2.3 videos have audio attached and to change the format to one of the others means stripping off the audio.  It is just a shame because I can actually make MP4 videos with audio that are smaller than those images I used to upload, but they just don't work with the forum.  I attached it as a WEBM, hoping it would work and it didn't.

Back into lurk mode.
Punchy

PS.  Here is the MEGA link to the same video as an MP4.
https://mega.nz/file/13wyTDQR#TXWjK8nnII5uUBGqQjnYCvepuL_K87bPVRfFjwSEsQc

Good work Punchy, at 4,3mb it is bigger than allowed here, which when it comes to all the new AI, I think 4.3 may be small.  Lots of stuff on DA that must exceed that.  I'm not going to pretend to know the difference today, but things have changed; back when the BEA started, I had a Gateway Pentium 75 with a whole 1gb hard drive that most people said was bigger than I would ever need.  The 3mb limit was huge here since most images were in the low kb range and most people were still using dial up.  The world has changed, and sorry to say, the BEA has not kept up with technology at all.  Wolpertinger and I noted on the site issues some of the current problems that are ignored; not sure who is responsible for correcting things, but very little has been done for the past 20+ years; people were complaining about the censored words back around 2003 and they are still censored.

[wolpertinger]

^
^
'Welcome' back, Punchy.
I can understand that you want the full video displayed including audio. But you can "embed" nearly any filesize via the {img}{/img} -function (replace curly brackets by square brackets) as long as the file-format is either *.gif (verified up to over 20 MB) or *.webp (verified up to over 16 MB).

But who I am, to teach you!?

[punchy]

^
^
'Welcome' back, Punchy.
I can understand that you want the full video displayed including audio. But you can "embed" nearly any filesize via the {img}{/img} -function (replace curly brackets by square brackets) as long as the file-format is either *.gif (verified up to over 20 MB) or *.webp (verified up to over 16 MB).

But who I am, to teach you!?

Thanks, Wolpertinger, but I am not really back.  Just checking in.  And my frustration is with the fact that GIF and WEBP formats are both without sound.  I embed my music videos into other forums, complete with sound and all, and I can shrink an MP4 to usually around 2 GB in size without making it unrecognizable. I never really had an issue with using the WEBP format, and I have tools now that can shrink them without the watermarks.  I am still monitoring the AI field and watching for new tools that could be useful here, but so far the only one I found that is new and powerful for making stills, but with mixed results, is Ideogram 4.0.  I have a workflow that allows me to create bounding boxes by clicking and dragging.  I can make a box for the woman, then create two boxes for her breasts and it will make her breasts fill the boxes I make up to a point.  I am not fond of the texture, though, as it gets seams on anything over the size of a basketball. It also has random results for what nipples look like.  However, Ideogram is new and LoRAs and checkpoints are just now being made for it.  Watch that one.  ERNIE NAVA is another one I was watching, but the censorship is too heavy with it so far.  I am unsure if anyone will create a less censored text encoder and model for it.  I can probably find an abliterated encoder, but the model hasn't been around for long.

The reason I am watching Ideogram 4.0 and ERNIE NAVA is because they are both small models that fit on consumer grade equipment and scored high against existing open source models and some commercial models.  They would be EXTREMELY useful for my music videos!

I will post an Ideogram example here.

Punchy

[punchy]

Here is the Ideogram example.  I will post a picture of the workflow.

ŠPunchy