AI text2img img2vid Tensorart guides, links, loras & prompts. [DO NOT POST REAL PEOPLE'S IMAGES OR RESULTS]

Hi, im asking if there are any minimax v2v tensor? If there are, do they work well to make nude video (like wan vace)?
 
thankss, but what i mean is the result of the end of frmae are always too contrast, u try put thr prompt for the static contrast but still same
 
The fluctuating credit pricing is concerning. Seen prices change while setting up generation plus sometimes they state that something costs 30 credits but then the 50 is spent by the time its done
 
anyone know where i can find loras for minimax? i tried searching tensor for them but only a few came up. i cant even find Obscura's loras when using the searchbar
 
Would you consider doing something that is basically your H3 Templates now, but with the option for multiple optional LORAs? I was noticing how much the deepthroat one was helping for BJ related stuff, but then when I tried to do one with sex, it was having difficulty listening to the prompt without the sex Lora, but then the ass and vag looked messed up without the hole helper LORA.

Whether you do something like this or not, thanks in advance and for all you've done already.
 
Played around way to much today and used the following prompt in the following links:

1. https://tensor.art/template/1035639150751013424
2. https://tensor.art/template/1035495390679449302
3. https://tensor.art/template/1036260303953736295

(guessing these are different because of the LORAs (pretty sure that's the case) but I'm still new.

Prompt 1
**Shot 1** Photorealistic midday sunny beach photography, the young woman with long brunette hair shown in preserves her exact facial features, body, small light blue bikini top that is fully covering her small breasts, small light blue bikini bottoms, is lying on her stomach, propped up on her elbows, holding an open book almost flat on the sand, sunbathing on a sandy beach, tight medium camera shot 10 degrees above eye level. **Beat 1 (0-2s):** Woman looks to the camera and smiles after seeing the camera. **Beat 2 (2-3s):** The woman turns her head 15 degrees with a cautious expression to the right and looks past the camera. **Beat 3 (3-4s):** The woman cautiously turns her head 30 degrees to the right and looks past the camera. **Beat 4 (4-6s):** The woman looks back to the camera with a mischievous smile. She closes her book and places in on the ground on the left. **Beat 5 (6-10s):** With her left elbow still on the ground, the woman props herself a little higher, slightly turning toward the camera. With her right hand, she reaches over to the top of the bikini covering her left breast and pulls it down to reveal her left breast and nipple, natural breast bounce. **Beat 6 (10-11s):** Her left elbow remains on the ground. With her right hand, the woman puts her bikini top back in place over her left breast and brings her right hand back in front of her and settles back into her original pose. **Beat 7 (11-13s):** The woman picks up the book, opens it with both hands, and laughs as she looks down at the open book. **Beat 8 (13-14s):** slow fade. **Camera work:** Smooth tracking shot that maintains consistent distance and angle throughout. Professional cinematography, no crash-zooms, temporal consistency, natural color, cinematic golden-hour lighting, crisp textures, clean edges, fine surface detail, high microcontrast, realistic shading, accurate tone mapping, smooth gradients, realistic highlights, detailed fabric and hair, sharp but natural. **overall_soundscape:** Soft ocean waves and light coastal breeze under everything. **non_diegetic_music:** None.

Prompt 2
**Shot 1** Photorealistic midday sunny beach photography, the young woman with long brunette hair shown in preserves her exact facial features, body, small light blue bikini top, small light blue bikini bottoms, is lying on her back on the sand, sunbathing on a sandy beach, eyes closed, relaxed expression, hands at her sides. Woman is lying with her feet toward the water. Camera is positioned further in land, a foot above the woman’s head, looking down over her body, to toward the ocean past the woman’s feet. Tight medium camera shot. **Beat 1 (0-2s):** Woman sighs and reaches to adjust her bikini bottoms. **Beat 2 (2-4s):** the Woman pulls the front of her bikini bottoms three inches away from her body exposing her shaved vagina from above. **Beat 3 (4-5s):** The woman puts her bikini bottoms back into place, covering her vagina. **Beat 5 (5-6s):** The woman returns her left hand to her side and rubs her right hand across her stomach before returning it to her side. **Beat 6 (6-8s):** The woman tilts her head slightly to the camera, opens her yes to look at the camera, and smiles. **Beat 6 (8-9s):** slow fade. **Camera work:** Smooth tracking shot that maintains consistent distance and angle throughout. Professional cinematography, no crash-zooms, temporal consistency, natural color, cinematic golden-hour lighting, crisp textures, clean edges, fine surface detail, high microcontrast, realistic shading, accurate tone mapping, smooth gradients, realistic highlights, detailed fabric and hair, sharp but natural. **overall_soundscape:** Soft ocean waves and light coastal breeze under everything. **non_diegetic_music:** None.

Had to do about 7 takes each until I was happy.

Link 1 is my favorite as it preserves the faces and bodies the best.
Link 3 follows the prompt very well but, while the picture is great, the faces look similar.
Link 2 takes a lot of credits but has a smoother finish than link 3 with a similar face issue.

Thank you to all that shared the links, my problems are surely user error. In the end, playing with tensor because I wanted to use it for a project. In the end, very time consuming with a pretty steep learning curve (the project is SFW, longer videos). In the end though, this is a bit of fun.

I've took advice and borrowed from so many of you on here. Your work is incredible and thank you.
 
Just added a second lora option for you. Not totally tested so if it fails let me know.
 
Just done a couple new gens, tried using the All in One in combination with hole helper. So far so good.

I've had massive variation across different prompts and starting images with how high I should do the Lora Strength. Seems like with H3, going at like .8 or above makes the prompts seem more accurate and makes stuff like genitals not look as messed up, but might sacrifice face accuracy possibly (not exactly sure)... sticking to like .6 or under seems mostly decent.

But I've done as high as like 1.2 loras on some, and the face still looked fairly accurate. Could just be a luck of the draw with the seeds and stuff sometimes or just having better starting images than others.

The gens I tried with 2 LORAs seemed not so good at 1.0 strength, but better at 0.5. But I've only just begun using the 2 loras option. Thanks again!

I know (or at least assume) the LORAs themselves are just something you implement on your templates and not tweak them directly, so I'm just speaking generally to share my experience so far in case it might help other as well.

I was happy to see when I popped over to Civit where I haven't been in forever, that guy K3nk is back at the forefront of some H3 LORAs. I used to follow him back in the day when I ran stuff locally and he had a lot of good shit for sex related LORAs, looks like he's still at it!
 
Correct for the most part. The minimax tool uses a built in Turbo lora set at 1 str. Everything else for the loras used is the default you'd find for that lora anywhere except for a few tools. If I name it, heavily imply it, etc then the lora isn't modified. Some tools do have a blend though.

I haven't done a ton of minimax testing tbh but what I've noticed is that ideal strength seems to be lower compared to wan. 0.5-0.8 range before it affects the face.
 
Anybody knows a template in tensor.art that supports more than 3 loras just like tensohub does?? I can't get the epic bukkare result I'm getting on tensorhub with the WAN2.2 Lightning 30FPS - IMAGE TO VIDEO - ADVANCED template which only supports 2 Loras...
 
Seedream 5.0 pro edit, best fucking thing that will happen in your life and cost .08 i think 😂 on wavespeed, add this prompt “Realistic portrait of a person with radiant, even-toned skin, professional lighting, studio background” and it will look like some professional shit
 
I may need to cut out the 2nd lora unfortunately if it stays above 50 for long

Also
https://tensor.art/template/990243449892513685

I need a couple beta testers before fully releasing this and adding more options. It is intended to be all in one for wan and h3. Note that you can change the lora by setting using custom on the lora selector and setting it to the ID. Universal lora is DR34ML4Y at 0.6 for WAN and HM AIO at 0.5 str for H3.
 
I was having tremendous results last night with the double LORA template of yours. I had the free accounts on the initial run, but then I was doing a bunch on the premium account, including going past my daily credits.

The entire time I was using it, it had its estimated cost around 39-48. I only saw it over 50 earlier in the day when I wasn't using my premium account at the time. Just saying that to push back that the adding of the 2nd LORA made the cost go up. I mean, it might have, but just adding in my anecdotal account that I didn't notice it.
 
It's already down so I'm not too concerned. As long as it doesn't stay above for long periods of time we're good.
 
guys I don't understand the following:
there are loras like this one https://tensor.art/models/1019865687935438101 that have been published fairly recently it seems and have been run. But the prompt would require to write words that tensor will stop, even using the bypass.

So what's the way to use stuff like this?

EDIT: okay I've figured out. From what I've understand one has to use them in the comfyflow in tensor art
 
You must log in or register to reply here.
Back
Top