My God... what is wrong with you guys?
Such extreme views on some software. I've never seen anything quite like this before, except maybe when dealing with actual fake scam applications. That's really why you think VEAI is?
I use Hybrid, VEAI, Avisynth, Adobe Premiere, and Davinci Resolve. Everything has its use and place.
To each his own I guess...
+ Reply to Thread
Results 541 to 565 of 565
Please, paint on any picture the place where logo was, shouldn't be hard for you to spot the blurry distracting artifacts.
For the love of God, upload images on the VideoHelp's own image server, it's not that hard and it makes em actually load
Here is a random logo removal, a last one I did
FYI, there are three basic logo removal methods, two of them has nothing to do with motion, so looks like you know only about one of them.
Still waiting for your MS Paint skills painting the place where logo was, use the forensic filtering if you need.
Attached the clip with motion.
Attached the clip with motion.
I had no doubt about your results, even without watching the video instead of the pictures. LS probably wanted to stress about the fact that showing the full video is more convincing than static images.
(1) Cel animation is far easier to work with, be it encoding, or restorations. In this exact case, the logo can be consider unwanted noise, and attempted removal is a restoration.
(2) The excess grain and errant noise makes it easy to hide flaws. This is an old restoration trick, adding less-bad noises to conceal imperfect removal of others.
Now, if Voodoo here wants to impress me, I'll find him a far more typical clip of a logo that needs removal. Not a cartoon, not a faint watermark, not animation. I'll probably be wasting my time, but I at least want to provide a nice long varied clip. Then it's a case of "STFU or prove your magic method".
Logo removal is one of those idiotic irksome claims that goes back decades. After a while, years, literal decades, you get tired of it.
LS probably wanted to stress about the fact that showing the full video is more convincing than static images.
-- and that brings us full circle to the topic of this thread: Topaz, artifacts, crappy upscale/sharpening, and mostly-BS "AI" used for marketing to the lemming masses.
Anyway, progress was made, he found bunch of excuses instead of claiming that all logos are irremovable.
Examples of "expert" advises:
Maybe here you can show your MS Paint skills: Attached non-animation video with various scenes, slow, fast, horizontal & vertical movement.
Hint: Size of a Ford Focus logo.
You had two options here:
(1) Discuss current logo removal methods (using your InpaintDelogo, initial released in 2019, not really mature until later releases), which may make past years/decades of advice less accurate (albeit still not really wrong).
(2) Be a troll/dick.
You chose the latter.
So I decline further interaction.
But if that Fleischer sample truly is a before/after (and you're not just trolling), then it is interesting. I'd need to see the scripts/whatever, and be able to recreate myself. And then attempt on projects where I've wanted to remove logos for years/decades (and nothing was ever satisfactory, not even with InpaintDelogo).
I see logo ghosts on the live action sample, and it's noticeable on an HDTV. Also bad encoding glitches (too much noise, not enough bitrate). You're hiding leftover removal in noise, I'm not fooled, that's a standard trick. Sometimes it works well, sometimes laughably not at all.
This side conversation is derailing the thread.
There is nothing fundamentally new in the past ~decades, so there is nothing new in InpaintDelogo too, just various minor tweaks.
Same you could do ~decades ago, but obliviously you prefer to troll.
The Fleischer clip is fairly straightforward. Use the features of the logo against itself.
Threshold (of local contrast) sense edge detection, with possible median, to create a mask, then apply simple darken/dyn range reduction to the area convered by the mask, then downstream 2nd mask of edge+surrounding border to apply minor edge blend blur. Slight addition of gaussing noise and subsequent re-encode would basically hide any remaining incongruities.
For the record, poisondeathray, StainlessS, jagabo and johnmeyer impress me greatly (and some others I'm surely forgetting off-hand here), I'm sometimes in awe of what they've done. So it's not like I can't be impressed. And I think you (lollo) are quite competent at Avisynth.
Topaz doesn't impress me in the least, hatchet job of upscale+sharpen.
InpaintDelogo is more advanced that previous attempts to de-logo, but it's not flawless magic. Voodoo needs to take credit for what he's accomplished (kudos!), but not try to overstate the ability of his tool, nor be a tool himself.
Avisynth has so many possibilities, and nobody can know everything about it, even though some of us try. I know some of this has gotten better in recent past years, but I've not been able to keep up to date on it.
Generally with "logo removal" the post-removal attempts to hide are hatchet jobs. (Most often, the removal attempt itself is a hatchet job.) And indeed, the live action sample above has many tell-tale artifacts of this nature, especially viewed large. But with toons being easier, any slight improvements in methods might make it more believable at a non-small (preview) size. Or at least easier to ignore.
Noting that almost no logos are grades of transparent these days, so it has limited values. Even with the current-best methods, most logo "removals" still create distracting blurry blobs. (I never understand the willful ignorance to state the obvious here. It ain't removed.)
I've seen actual frame-by-frame repair, and I'm sure you have as well. It can actually remove harsh problems, and that would include logo-type noise (constant film issues frame-to-frame, types of cyclical damage). This is probably something I would have gotten into at work, had health not cut my career short. The software was starting to become more viable, and the need greater where I was (as we'd started to enter uglier sources, and the set-aside projects).
I'm not seeing what's humorous here.
I was impressed by the showed results, you replied you're not. Fine for me, but then when you changed your "tone" I had to smile
The funny thing is that I also think that logo removal is difficult and produces somehow contradictory results, and VoodooFX also agree with us
But I also thinks that all efforts from doom9's avisynth developers, such as "master" VoodooFX, deserve more consideration and support!
Logos tend to be located in corners, and not a simple transparency. Unless the content is very cartoon-like (large area of solid colors), those can't be removed without artifacts. And those artifacts are often more distracting than the logo. At least the logo was static, the blurry blob is an obnoxious amoeba moving on the screen.
But don't feed me dog shit and call it pudding
All started with:
In any case, VoodooFX does not need me to defend him, I was just impressed by (some of) his results and your comments. I stop here!
I spent $200 dollars just to find out I had to spend the last 5 days learning how to ivtc. Starting with Stargate SG1 S1E1.
With that being said. I'm still going to include VEAI in my workflow.
Which, during that 5 days, turned into DVDfab to VOB > DGIndex (force film) > Virtualdub2 > (5 days learning ivtc techniques) > Learned the SeperateFields() command > Realized there are Blended frames that are resulting in residual chroma ghosting at certain scene changes that no amount of ivtc tweaking is getting rid of. Thinking an override file is the only way to resolve..
Then decided to just go back to handbraking a file out, throwing it at VEAI, and being done with it. Because the thought of trying to jump through that many hoops for... 220+ episodes. Is not sustainable.
With that being said. Where do I find the repository of scripts built by this community to fix all these mistakes in common DVD series such as Stargate, so I don't have to jump through all these hoops to get a proper detelecined source to work with, free of blended frames.
That's what this place really needs. Community guides for fixing common Sci-Fi DVD series.
Where the first post is always updated with the most up-to-date method of acquiring the best possible detelcined source possible for throwing at programs like VEAI or for further refinement in Vdub/Synth, etc.
That might make the thought of spending more time with more advanced solutions reasonable to consider. Because I'm not writing out override files for 220 episodes by myself. Less time spent on fixing this stuff could mean more time spent learning to use filters to make your claims about topaz come true.
Forgot the most important part of the post. Getting knowledgable on Hybrid/Avisynth/VirtualDub2, is like playing whack-a-mole. And that's just to IVTC. Nevermind how many dead end web searches a person will go through to make those programs try to match or exceed topaz on a grand scale such as 220+ episodes.
You have no idea how hard it was to just to find this post. https://forum.videohelp.com/threads/401305-TIVTC-settings. while having come across it once before but forgetting to bookmark it. It took luck to stumble across it again.
Last edited by MrEos; 28th Sep 2022 at 23:16.
Then decided to just go back to handbraking a file out, throwing it at VEAI, and being done with it.
For the time effort involved in a Handbrake > VEAI
vs learning the more complicated methods.
Somewhat worth it.
It came down to spending the 100+ for the blu-ray set, or spending 200 for VEAI and having something to throw at a whole collection of dvd's after that.
I can see jumping down to Vdubing the more favored episodes. But, for a collection that is likely to be re-ripped at some point in the future when technology advanced. I didn't see the point in trying to perfect it on my own. That's why I just went back to batch converting with handbrake it's like 99 percentile issues from handbraking vs trying to TFM().TDecimate() and figuring out if DGindex should honor pulldown or force film. etc.
The whack-a-mole game of learning it is giving me gray hairs.
And I hate that for as long as everyone has been at this. That one single thread I linked to about sums up the total combined effort to bring forth a solution to fix any specific telecined DVD source. Like the internet is just going to waste... Like I'm starting from scratch instead of having something to work off that has had 1,000's of people working on over the years. Not including the plugins of course.
If I can thank VEAI for anything, it was to send me on a quest to make a proper progressive source backup. But it's not a chore for the faint of heart.
is very close to what I settled on for now, until I feel like playing whack-a-mole again.
Last edited by MrEos; 29th Sep 2022 at 02:21.
Can you please stick to the topic of this thread and not bash each other over whose logo removal is best?I think,therefore i am a hamster.
Logo not mentioned in last 3 posts (not including yours), and not after 19Sept.
For those who wanna test various super resolution "AI" (= neural networks) models, we just found this interesting repo (and relative colab):
Repository to use super resolution models and video frame interpolation models and also trying to speed them up with TensorRT. This repository contains the fastest inference code that you can find, at least I am trying to archive that. Not all codes can use TensorRT due to various reasons, but I try to add that if it works. Further model architectures are planned to be added later on.
Currently working networks:
- ESRGAN with rlaphoenix/VSGAN and HolyWu/vs-realesrgan
- RealESRGAN / RealESERGANVideo with xinntao/Real-ESRGAN and rlaphoenix/VSGAN
- RealESRGAN ncnn with styler00dollar/realsr-ncnn-vulkan-python and media2x/realsr-ncnn-vulkan-python
- Rife4 with HolyWu/vs-rife
- RIFE ncnn with styler00dollar/VapourSynth-RIFE-ncnn-Vulkan and HomeOfVapourSynthEvolution/VapourSynth-RIFE-ncnn-Vulkan
- SwinIR with HolyWu/vs-swinir
- Sepconv (enhanced) with sniklaus/revisiting-sepconv
- EGVSR with Thmen/EGVSR and HolyWu/vs-basicvsrpp
- BasicVSR++ with HolyWu/vs-basicvsrpp
- RealBasicVSR with ckkelvinchan/RealBasicVSR
- RealCUGAN with bilibili/ailab
- FILM with google-research/frame-interpolation
- PAN with zhaohengyuan1/PAN
- IFRNet with ltkong218/IFRNet
- M2M with feinanshan/M2M_VFI
- IFUNet with 98mxr/IFUNet
- eisai with ShuhongChen/eisai-anime-interpolator
- SCUNet with cszn/SCUNet
- GMFupSS with 98mxr/GMFupSS
- ST-MFNet with danielism97/ST-MFNet
- VapSR with zhoumumu/VapSR
PLEASE DON'T ABUSE.
Last edited by forart.it; 13th Nov 2022 at 04:08.