TenStrip

/

10Eros-Max

Like 858

There's a weird thing going on on my side. Beta5 doesn't follow my prompt at all. I'm getting a video that's completely different from the one I wanted (and censored). I'm using it with WanGP as a finetune. Any idea?

My previous test was flawed , the anatomical structures aren't distorted, the bodily fluid effects are excellent, and there's no background noise—it turned out I had mistakenly enabled other LoRAs. This version is great, and I look forward to the next one being even better, Cheers!

My previous test was flawed , the anatomical structures aren't distorted, the bodily fluid effects are excellent, and there's no background noise—it turned out I had mistakenly enabled other LoRAs. This version is great, and I look forward to the next one being even better, Cheers!

Would you say best by itself or is there a "must" have lora with the turbo version? you think? or??? I can't wait to test.

My previous test was flawed , the anatomical structures aren't distorted, the bodily fluid effects are excellent, and there's no background noise—it turned out I had mistakenly enabled other LoRAs. This version is great, and I look forward to the next one being even better, Cheers!

Would you say best by itself or is there a "must" have lora with the turbo version? you think? or??? I can't wait to test.

The turbo version must be used on its own to avoid unexpected interference and noise , an 8-step setting yields the best results. The output is fantastic more realistic and true to life than beta4.

My previous test was flawed , the anatomical structures aren't distorted, the bodily fluid effects are excellent, and there's no background noise—it turned out I had mistakenly enabled other LoRAs. This version is great, and I look forward to the next one being even better, Cheers!

Would you say best by itself or is there a "must" have lora with the turbo version? you think? or??? I can't wait to test.

The turbo version must be used on its own to avoid unexpected interference and noise , an 8-step setting yields the best results. The output is fantastic more realistic and true to life than beta4.

Thanks for the reply brother. That eases my worry. Thanks again

idk what happend, but beta 4 much better than beta 5? maybe something wrong? it not listen my prompt, and always produce weird movement. i do same prompt and setting like on beta 4 , or beta 5 need change something?

anyone having problem with avoiding nudity in beta5? Characters are mostly naked even if not generating NSFW

anyone having problem with avoiding nudity in beta5? Characters are mostly naked even if not generating NSFW

I've been describing the outfit and saying they wear that for the entirety of the video clip.

I'm torn, but from my limited testing I'm leaning towards beta4 having better outputs. The difference is not that big, but in my testing the micro-movements and relations between objects/body parts is just more natural in beta4... I'll test more, but for now, it's like 65% towards beta4.

I'm torn, but from my limited testing I'm leaning towards beta4 having better outputs. The difference is not that big, but in my testing the micro-movements and relations between objects/body parts is just more natural in beta4... I'll test more, but for now, it's like 65% towards beta4.

I think I may be in the same boat... beta5 seems to be oookaayyy in most scenarios. but damn do you have to be specific! makes me think that the prompt adherence lora was like solely imbedded into it. Tenstrip did say, don't blame the model, look at your prompt.

5 It does the old vag in balls thing. You know a little xtra hole in the middle. Didn't notice it much in beta 4. It was a big issue in ye olde wan2.2 aio ones by phr00t. I assume it mixes the two concepts (i.e. related loras) merging together to some extend. It was also not an issue in the LTX2.3 version of 10eros. Which was actually better at male focussed stuff than for example dasiwa.

Beta 5 It also does naked often not completely without reminders. Probably especially on the male side. And that is definitely better in beta4.

As for blaming prompts: The BIG advantage of H3 over LTX2.3 and even 2.5 is that it does NOT need extreme prompt babysitting. Not just for nswf just in general getting a working result without a book sized prompt. And H3 does that very well. And beta4 also. So the blaming of the prompt for issues with 5 feels weird (i only read it here, haven't looked at the actual post or its context.).
that said, If there are any triggers or phrases that help prevent certain issues or help, that would be ok (if we know them).

i also hope someone makes a similar project for ltx2.5 It is not as good as h3 but it is SO much faster and can do longer.

5 It does the old vag in balls thing. You know a little xtra hole in the middle. Didn't notice it much in beta 4. It was a big issue in ye olde wan2.2 aio ones by phr00t. I assume it mixes the two concepts (i.e. related loras) merging together to some extend. It was also not an issue in the LTX2.3 version of 10eros. Which was actually better at male focussed stuff than for example dasiwa.

Beta 5 It also does naked often not completely without reminders. Probably especially on the male side. And that is definitely better in beta4.

As for blaming prompts: The BIG advantage of H3 over LTX2.3 and even 2.5 is that it does NOT need extreme prompt babysitting. Not just for nswf just in general getting a working result without a book sized prompt. And H3 does that very well. And beta4 also. So the blaming of the prompt for issues with 5 feels weird (i only read it here, haven't looked at the actual post or its context.).
that said, If there are any triggers or phrases that help prevent certain issues or help, that would be ok (if we know them).

i also hope someone makes a similar project for ltx2.5 It is not as good as h3 but it is SO much faster and can do longer.

if you want long H3 videos -> 1min+ , look for the Continuum plugin\workflow on civitai