Author
|
Message
|
sonic7
|
sonic7
Posted 2 Years Ago
|
Group: Forum Members
Last Active: 2 Years Ago
Posts: 1.7K,
Visits: 19.4K
|
Not sure if this has been mentioned on the forum, but in case it hasn't - check out what this guy is doing ... Here's just one of his videos:
--------------------------------------------------------------------------------------------------------------------------------------------------------
Please be patient with me ..... I don't always 'get it' the first time 'round - not even the 2nd time! :( - yikes! ... ● MSI GT72VR Laptop, i7 7700HQ 4-Core 3.8 GHz 16GB RAM; Nvidia 1070, 8GB Vram ● iClone-7.93 ● 3DXChange Pipeline 7.81 ● CC-3 Pipeline 3.44 ● Live Face ● HeadShot ● Brekel Pro-Body ● Popcorn FX ● iRAY ● Kinect V2 ● DaVinci Resolve17 ● Mixcraft 8.1
|
|
|
Kelleytoons
|
Kelleytoons
Posted 2 Years Ago
|
Group: Forum Members
Last Active: Last Year
Posts: 9.2K,
Visits: 22.1K
|
Actually RL kind of "started" with this very process - CrazyTalk was kind of the cornerstone of this, using still images and animating them, decades ago (RL kind of had a twin path, with CT and iClone, but unless I'm misremembering, CT came first). So - old tech, nothing new here (perhaps just done a lot more polished, but it was kind of amazing what CT could do back before 3D animation was even a possibility).
Alienware Aurora R16, Win 11, i9-149000KF, 3.20GHz CPU, 64GB RAM, RTX 4090 (24GB), Samsung 870 Pro 8TB, Gen3 MVNe M-2 SSD, 4TBx2, 39" Alienware Widescreen Monitor Mike "ex-genius" Kelley
|
|
|
sonic7
|
sonic7
Posted 2 Years Ago
|
Group: Forum Members
Last Active: 2 Years Ago
Posts: 1.7K,
Visits: 19.4K
|
Yeah - when I see 'young' guys (ie: sharp minds) jumping in and out of various software's like this it both fascinates AND unsettles me. Fascination because of the growing options and pathways becoming available to create visual stories. But unsettling (for me) because it makes me question my current approach. And I've always 'preferred' using a single software (in order to get to know it well) - even though some would say that's unrealistic. I do use Blender 'a bit' to supplement iClone. It's natural to want to use the 'easiest' approach given how involved film making is.
--------------------------------------------------------------------------------------------------------------------------------------------------------
Please be patient with me ..... I don't always 'get it' the first time 'round - not even the 2nd time! :( - yikes! ... ● MSI GT72VR Laptop, i7 7700HQ 4-Core 3.8 GHz 16GB RAM; Nvidia 1070, 8GB Vram ● iClone-7.93 ● 3DXChange Pipeline 7.81 ● CC-3 Pipeline 3.44 ● Live Face ● HeadShot ● Brekel Pro-Body ● Popcorn FX ● iRAY ● Kinect V2 ● DaVinci Resolve17 ● Mixcraft 8.1
|
|
|
AutoDidact
|
AutoDidact
Posted 2 Years Ago
|
Group: Forum Members
Last Active: 3 days ago
Posts: 2.1K,
Visits: 13.6K
|
Companies Like Daz3D ,who’s main focus is still image rendering, will likely be put out of business by AI art generation, within five years. Some of their user base is already publicly announcing their migration. It will take a bit longer for usable animation and unlimited camera angles etc. but People will soon have an alternative to buying into expensive software and hardware eco system like Iclone as well.
|
|
|
Sophus
|
Sophus
Posted 2 Years Ago
|
Group: Forum Members
Last Active: Last Week
Posts: 230,
Visits: 2.6K
|
Some days ago Corridor Crew released a short film where they filmed themselves before a green screen and then used Stable Diffusion to change the art style of the recordings into an anime. The impressive part is, that they found a way to get a somewhat stable image generation, at least for this style. They used Dreambooth to train a model of their faces, then stabilised the noise seed during image creation and later used some anti flicker filters in After Effects. The backgrounds were generated in Stable Diffusion as well. To get the church backgrounds consistent looking they used an Unreal Engine asset as an input image. So, it's not only image based but movie based. Pretty impressive stuff, but I'm not sure this will work with more realistic looking styles. But you could also use DAZ or Reallusion characters to train the model and transfer the style instead of real human actors. https://www.youtube.com/watch?v=GVT3WUa-48Y
|
|
|
sonic7
|
sonic7
Posted 2 Years Ago
|
Group: Forum Members
Last Active: 2 Years Ago
Posts: 1.7K,
Visits: 19.4K
|
Although things are heading in this AI direction, it still has a way to go before it could supplant the use of *real* 3D Avatars, Sets and Props - though things seem to be accelerating fast. For now, I guess *some* aspects of these newer approaches could be used to *supplement* the current 3D method of film making. I'm thinking mainly of the depth map approach this guy demonstrated. For example, in shots that have only a small camera movement (like establishing shots / 'one-off' shots). It might save on both the cost and design-time of using actual sets/props, but could still be lit as actual 3D mesh. The *design* could instead be done in 2D then displaced into 3D using the depth map approach. I'm wondering, after using the "displacement modifier" (in Blender) to create the 3D look, whether that can be exported out of Blender as actual 3D mesh for importing into iClone? Someone might know about this. Another way of supplementing iClone film production might be to use the "I-Id animation software" mentioned for creating background *extras* in scenes where the lower resolution wouldn't matter since the characters would occupy only a fraction of the frame. Edid: - Just looked up "I-Id" and, (unfortunately) the *cost* for the small time solo filmaker could well be off-putting. ($50/month if wanting 'commercial' use, and that's for just 15 minutes). 93% of original size (was 540x16) - Click to enlarge
--------------------------------------------------------------------------------------------------------------------------------------------------------
Please be patient with me ..... I don't always 'get it' the first time 'round - not even the 2nd time! :( - yikes! ... ● MSI GT72VR Laptop, i7 7700HQ 4-Core 3.8 GHz 16GB RAM; Nvidia 1070, 8GB Vram ● iClone-7.93 ● 3DXChange Pipeline 7.81 ● CC-3 Pipeline 3.44 ● Live Face ● HeadShot ● Brekel Pro-Body ● Popcorn FX ● iRAY ● Kinect V2 ● DaVinci Resolve17 ● Mixcraft 8.1
|
|
|
AutoDidact
|
AutoDidact
Posted 2 Years Ago
|
Group: Forum Members
Last Active: 3 days ago
Posts: 2.1K,
Visits: 13.6K
|
Yeah that D-ID service is clearly targeted toward corporate clients. I quickly burned though my 14 day trial and made a few videos where I generated an AI character on another site and used it with D-ID for easy talking head/commentary clips. I have CTA4 I wish it were this easy to make talking heads from any front facing image with out all of the manual set up required.
|
|
|
Kelleytoons
|
Kelleytoons
Posted 2 Years Ago
|
Group: Forum Members
Last Active: Last Year
Posts: 9.2K,
Visits: 22.1K
|
I saw the Corridor Crew video where they processed their live action stuff and I got really excited because I wanted to do the same with iClone rendered footage. But - WAY too much work (even though they talk about "democratizing" the animation process, it still requires quite a team and quite expensive equipment and software). The Stable Diffusion app they used, DreamStudio, is "okay" - I've had fun playing with it but it's nowhere near ready to handle video (again, they did it with $$$ - no one man studio could possibly do the same). Plus - the process they used was for a VERY specific animation style with hardly any background movement (still images animated only). I think we are DECADES away from being able to take our rendered stuff and actually process it in a way that makes it any better than the so-so filters any program offers today (I won't be alive, for sure). So - it's interesting. Even this guy's "animation" is VERY limited - sort of panning around the talking head. I can't see much practical use except as a one-off. We already have MUCH better tools and ability to generate MUCH better looks inside of iClone itself, with or without post processing. And doing it even easier (assuming you have the necessary mocap tools - and, yes, they do cost $$ but if this old man can afford them then almost anyone can).
Alienware Aurora R16, Win 11, i9-149000KF, 3.20GHz CPU, 64GB RAM, RTX 4090 (24GB), Samsung 870 Pro 8TB, Gen3 MVNe M-2 SSD, 4TBx2, 39" Alienware Widescreen Monitor Mike "ex-genius" Kelley
|
|
|
sonic7
|
sonic7
Posted 2 Years Ago
|
Group: Forum Members
Last Active: 2 Years Ago
Posts: 1.7K,
Visits: 19.4K
|
Yeah - Personally, the animation using D-ID doesn't overly grab me, but I'm wondering about the Blender 3D displacement approach using depth maps. - I checked, and apparently it IS possible to convert the Blender projected displacements into actual 3D mesh for export. I see two advantages and one disadvantage. + Greater creative freedom in generating *any* imaginable background using 2D artwork as a starting point. + Cost savings on 3D asset purchases. - But, it involves multiple steps 1) Create the artwork, 2) Generate the depth maps, 3) Layout in Blender and export mesh (assuming iClone usage). I guess you could get 'similar' results using a couple of iClone's planes to force a "shift in perspective" during a camera move.
--------------------------------------------------------------------------------------------------------------------------------------------------------
Please be patient with me ..... I don't always 'get it' the first time 'round - not even the 2nd time! :( - yikes! ... ● MSI GT72VR Laptop, i7 7700HQ 4-Core 3.8 GHz 16GB RAM; Nvidia 1070, 8GB Vram ● iClone-7.93 ● 3DXChange Pipeline 7.81 ● CC-3 Pipeline 3.44 ● Live Face ● HeadShot ● Brekel Pro-Body ● Popcorn FX ● iRAY ● Kinect V2 ● DaVinci Resolve17 ● Mixcraft 8.1
|
|
|
AutoDidact
|
AutoDidact
Posted 2 Years Ago
|
Group: Forum Members
Last Active: 3 days ago
Posts: 2.1K,
Visits: 13.6K
|
I think we are DECADES away from being able to take our rendered stuff and actually process it in a way that makes it any better than the so-so filters Daz recently updated their EULA to put restrictions on using Daz Assets in AI generators. This is a clear indication that the Daz/poser and perhaps Iclone community does NOT seem understand that this technology is creating a separate standalone eco system that wont depend on users buying & owning 3D assets and expensive software and rendering them to be “filtered” or “converted” after the fact.. They already have the ability create consistent Characters from nothing or custom trained self inserted styles images. Now they can Re-POSE them for consistent sequential art such as graphic novels/comics full animation frames is not that far off.
Yes ATM ,you still need a beefy GPU PC for home spun custom AI Data training, but it will only be a short time before all the tedious labor that the Corridoor crew Did, is packaged into a back ended subscription service for both laypeople hobbyists and pros with budgets. I predict the the current business models of companies like Daz & Poser will be the first to die and the animation tech will threaten Reallusion sooner than later.
|
|
|