🔥 Stable Video 3D - Local Install Guide 🔥 SV3D
HTML-код
- Опубликовано: 20 мар 2024
- Stable Video 3D can create stunning 3D Videos and 3D Rotation Videos from a single image. This has much improved 3D shapes and better even Lighting that tries to avoid baked in lighting
#### Links from my Video ####
Support me and get my Workflow: / 100825665
Pastebin Workflow: pastebin.com/RuCD6BYe
Models: huggingface.co/stabilityai/sv3d
Nodes (use ComfyuI Manager for install instead)
github.com/kijai/ComfyUI-KJNodes
github.com/Fannovel16/ComfyUI...
#### Join and Support me ####
Buy me a Coffee: www.buymeacoffee.com/oliviotu...
Join my Facebook Group: / theairevolution
Joint my Discord Group: / discord
AI Newsletter: oliviotutorials.podia.com/new...
Support me on Patreon: / sarikas 
Хобби
I'd love to see them add the ability to create an STL file from these. It would be amazing to be able to go from a photo to a 3D model.
I'm going to try it soon but I'm 99% sure this should be able to be directly put into polycam and processed into a 3d model
this would be amazing, i want to print a lot of things
there are online versions of this already doing that
whats the point of 3d models? why not generate the video or image straight away?
But you are aware of 3D printing, game development or 3D animation? Or do youseriously not know the purpose of 3D models? @@f4ust85
This is much better than I expected.
Was wating for a Video like this, thanks a lot! :D
Honestly I am absolutely salivating for once SD3 comes out.
SD3 Fine tunes in addition to ControlNet and SV3D will save me so much time.
Even if I need to do some clean up in Blender.
While I'm completely amateur in 3d modeling I kinda think based ony limited experience that fixing and cleaning premade model often takes more time then starting from scratch.
@@wykydytron That depends what you want and what you already have.
Talking from experience here.
Excellent!
it really is amazing from one frame to do that
This sounds great. I may finally now be able to create a character, move it to give me "turnaround" images where they have the same clothing/costume, import them into Cartoon Animator 5, rig them, and do any sort of animation I want with them... well, that's the dream, we'll see if the reality lives up to it.
Did you try the model with images not on white studio background, Olivio?
@Olivio Sarikas so we dont get an fbx or obj? just a spining video?
Thanks for sharing the workflow JSON, "just worked" on macOS w/ Comfy :)
perfect :)
In the whimsical realm where Olivio Sarikas reigns supreme, culinary indulgence takes on a delightfully cosmic twist. Imagine a golden, buttery spread that not only tantalises the taste buds but also whispers eldritch secrets from forgotten civilisations. With each velvety bite, you'll savour the rich, creamy notes while embracing the playful absurdity of flavours intertwined with ancient lore. So, let us raise our knives and pay homage to the buttery brilliance of Olivio Sarikas, for in this delightfully mad world; gastronomic perfection meets cosmic notion in the most deliciously whimsical of dances.
How is it on human images for 3d rotation? Can it rotate a team player headshot per say? Something to try out
I wonder if there is an underlying 3D model generator!? The 3D model would be super useful!
There is a way to generate a mesh with it, but i don't know how yet
Make a video on mesh generation! thanks!
Try gaussian splatting (to get correct 3d coherence) then use photogrammetry.
@@timmygilbert4102u mean u do a gaussian splat first and then use that "bad" model for photogrammetry?
RIP 3d modeling jobs 🪦
I hated the way they were licensing things at first, but the more I think about it, the more I like it. It's a bit like the old shareware model... you get the fully unlocked code, but if you like it, you pay for it. I hope it works out for them. They're doing great stuff.
love this
Thank you for always providing the latest information and great lectures!!!
Honestly, this is all I need. Use this with Agisoft Metashape and now you can convert AI Shapes to 3D. Thanks for the video
is agisoft better than meshroom or gaussian splatting?
@@salomahal7287 I don't know, never tried them. I usually go with industry standard tools and Agisoft is on top
I recommend putting in video the system requirements for installs, perhaps in the description.I assume this won't run on lower vram setups?
It's unclear where the output of the GMFSS Fortuna VFI node goes, and the node does not seem to be present in the workflow you linked to. Could you please clarify here in the comments?
"Craft Reaper" will cry for sure...
Is Stability Ai operational after the rumored staff exodus ?
Any option to do it without comfy ui?
can we export it as OBJ or other 3D file formats ?
stable video 3d, I think it does not generate any 3d object, only the video of a 3d object
You sort of could with just the frames from the rotation, but the issue with it is that there are clearly consistency issues, which would create all kinds of artifacts in a 3D model, making them essentially unusable.
You sound like youve tried it, id like to see your results
Coooool!!
To create a functional 3D object, it’s essential to have additional images from various angles. Simply training a 2D based model or employing a 2D engine won’t suffice to comprehend the complexities of three-dimensional space.
Here are my suggestions:
-Introduce a new “3D LORA model” or Blender files supported.
-Implement support for transparent backgrounds
-The correct approach to image generation should transition from “3D to 2D, 3D to 3D.” For high-quality 2D/3D image creation, the generator should be designed to construct the object layer by layer, starting with the skeleton, then adding muscles (structure), and finally the skin (texture).
Ok so I have a plan to get to the andromeda galaxy. First we build a really big ship, then we go really fast and get there. It's not hard to think of a high level plan on how to do something. The devil is in the details. How do you begin to do the things you mentioned?
I would argue it's better to train on volumetric/point cloud information, then use any old software to convert to mesh
can we get a tut for forge or automatic 1111
For any other noobs out there (like me), take the pastebin file and make it a .json file and drag and drop into comfyui. I am reporting back that I was able to render a chair on a gtx 1060 ti 6gb 😁😁
How much v-ram does it need?
anyone have a link to how to use sv3d-u model ?
sir i got error message after install custom nodes it says "When loading the graph, the following node types not found : - VideoTriangleCFGGuidance - SV3D_Condtioning Nodes that have failed to load will show as red on the graph
You need to update your ComfyUI to current version and then the nodes will show up
how can i save output video?
can I save the generated video or not ?
let me know if you find a way how to save output video
Auto1111 version?
So this produces videos only of simple, rotating objects - no 3d models? 😐
That sounds exciting! "Stable Video 3D's latest update is a game-changer for creating stunning 3D videos. The improved 3D shapes and advanced lighting features truly elevate the visual experience, making it easier than ever to create professional-looking content from a single image
mine comes up out looking weird.
I hit the "Update All" button and broke my comfyUI. turns out rgthree was the culprit.
ComfyUI always gets stuff first? I've had this in Forge since it's release. It comes bundled with it.
Its possible to save it as 3d model?
there is a way to create 3D meshes with it, but i haven't found that out yet
Of course not. This is not for making 3D models...
I dont think so, but you can use photogrammetry software to generate models off the frames
@@OlivioSarikas I am already looking forward to your video on this topic ❤
How about automatic 1111
always coming a bit later when someone writes an extension for it
@@OlivioSarikas can't wait 🙌🏻
I THINK THIS GUY KNOWS BETTER THAN TO MENTION ''COMFY UI'' ANYWHERE IN HIS TITLE BECAUSE HE WOULDNT GET HALF AS MANY CLICKS
😂😂😂😂
Why are you all caps about ComfyUI?
It doesn't seem like you're really doing this locally. It seems like this just plugs you into their website remotely, and they control your output as they seem fit.
useless (unless you sell 3D models for printing on ETSY) - but very cool!
Architecture and Construction, Video Games, Film and Animation, VR, Product Design, Interior Design, but you know... useless, if you can't see past the end of your nose...
It's kind of useless if you can't create a 3d model from it. Maybe you can combine it with TripoSR later.
stable 0123
Looks neat, just rotating is not that great to be honest, I’d kinda like to see it do what I saw in other videos (The one where you have to have a subscription online) where the head moves up and down if it’s a person, something less mechanical.
Obviously only if this can be done off-line and locally.
"Amazing Title - Affinity Photo" 😅
Does it still show that to you? strange!
@@OlivioSarikas Yep, in the drop down notifications. I'll blame Firefox 😆
not mindblowing. this is just picture, not 3d obj
It's a shame that you pay for streams. I will not follow your videos and you. You were already earning enough from RUclips views!!!