Poor Quality Export From Moho 13.5

[BigBoiiiJones]

Why is the export quality in Moho so poor compared to Clip Studio Paint? The only reason I can think of is maybe DPI quality and Moho simply lacks that so if that's a case maybe Moho should implement a DPI feature since most anime studios work at 192DPI to 300DPI at 1080p/720p

This drawing was made in Clip Studio 3840x2160 at 1200dpi but I even downscaled the canvas to 1080p but kept the 1200dpi and there was no quality loss. On the Moho Render I put a native Moho vector next to the hair and its also low quality. Both images were exported as PNGS. The Moho Project Settings is set to 4K. I don't know why Moho's AA and render quality is kind of bad and any other animation software doesn't suffer from this. I think Moho could benefit from a DPI setting because this would give the render engine more samples to work with when applying AA.

I made another post on here about bad AA and I feel this may be tied to the DPI in some way and the way Moho handles renders. I really hope in a future version of Moho (hopefully version 14) we see a huge improvement to the renderer and how AA is handled. This is not good results and I hope to see future improvement.

[synthsin75]

DPI is generally a print media thing. Moho is made to make videos.
If you render to PNG sequence, you shouldn't have any problem with the AA. But making that into a video will require a video codec, which could degrade the quality, depending on the codec you use.

[BigBoiiiJones]

DPI is generally a print media thing. Moho is made to make videos.
If you render to PNG sequence, you shouldn't have any problem with the AA. But making that into a video will require a video codec, which could degrade the quality, depending on the codec you use.

This was exported from Moho's renderer as a PNG sequence. This usually doesn't happen with up close shots mostly mid to far shots with Moho. I noticed CACANi and OpenToonz has a DPI setting as well. I think its because these software have Japanese productions in mind and the line widths are considerably smaller. When I lower the DPI of CSP to 72DPI I get a similar picture quality but maybe slightly better compared to Moho.

Here is 72dpi in Clip and it is slightly lower res in export with AA. The pixels are more apparent and aren't as smoothed. It seems Moho is doing some additional compression. It looks like Moho is applying some sort of sharpness filter.


Just want to add I'm doing anime so I'm working with videos I understand DPI is mostly used for "prints" but I feel like in CSP or these other software it may contribute to something else. Most of the time LO will be printed out so a animation director could do a check pass as well so its still relevant for Japanese animation productions in that sense but in CSP it does seem to affect my overall quality in some way which is odd. Even if it doesn't play much of a role and I'm hallucinating there is no question that Moho's render is very poor compared to other software.

[BigBoiiiJones]

Before extra troubleshooting I have extra smooth image and AA enabled in the exporter.

[synthsin75]

I'm not seeing any degradation when I import your first image and export a PNG in Moho.

[BigBoiiiJones]

I'm not seeing any degradation when I import your first image and export a PNG in Moho.

I'm on mobile now but this doesn't happen with objects that are considered closeups it mostly happens with shots from far shots and mid shots when the line width is small. The pictures uploaded are from a character sheet I did for study and it's a mid shot I zoomed in 600% with windows photo app and screencapped it to showcase the artifacts and pixelation. It gets worse with lower resolutions as well like 1080p or 720p. So with you taking this super upclose shot of the image the artifacts won't replicate that way. When I get back I'll upload the original and basically do a 4K render and then change the projecy properties to 1080p and do additional render and you'll see these artifacts.

[slowtiger]

Unless you're working in print you don't need to know about DPI at all, because a DPI setting only makes sense in bringing any bitmap to paper: it tells the printer how many pixels it should spread over a certain distance on paper.

The only thing which matters in video production is the dimensions of the project/video. HD 1920 x 1020 px is the same in all programs on all systems.

Let's say your Moho project is in HD (1920 x 1020 px), and you import an image. If the image is 1920 x 1020 px as well, it will fit the project window perfectly. That is because Moho imports artwork as 1:1, each pixel of artwork is one pixel in the project window. If your image is smaller, say 640 x 480 px, it will be smaller than the project window. If your image is larger, it will spill over the project window. Moho imports 1:1.

Other software interpret the imported artwork, like a video editor which assumes that anything you import, no matter whether it's 4K or HD or from a shitty webcam, shall fill the project window completely. And even other software may give you the choice between 1:1 or interpreting, like TVPaint. Our example picture of 640 x 480 px will be enlarged and spread over the 1920 x 1020 px of my FinalCut video project. Of couse such an enlarged image will look poorly between genuine HD scenes because it was scaled up. The other way round, importing a larger image into a smaller project, will show no difference between this and other HD images.

So the only size which really matters is the dimensions of your Moho project file. If it's 1920 x 1020 px you just need to take care that any imported background image is at least the same size, or even bigger if you plan to zoom in.

Now why do japanese studios still bother with DPI and fuck up everything? That's a bit of history of technique and style.

The first japanese softwares which could colour scanned drawings worked very simple: the drawings were 1-bit, having only black or white pixels. An image like that is easy to fill with colours, but all the lines will have jaggy corners. So they developed a workflow where each drawing was scanned in very large, then colourized, and then exported to a much smaller size so the pixels were interpreted/interpolated: the average of 2 or 4 or 9 pixels gave the value of 1 output pixel. If the drawing already had very thin lines, they got even thinner in the process, becoming less than a pixel wide and not black anymore, but somehow grayish.

Nowadays software can easily colourize drawings with anti-aliased lines, like 8-bit scans, and there's no need to scale down the coloured drawing to smoothen the lines. But in digital animation it is very hard (to be exact: it's impossible) to draw a less-than-1-pixel-wide line: that's why animators work in larger dimensions and scale down the artwork to get these very thin lines of that style which still is in demand.

I imagine that, at some point in history, some programmer with a background in printing wrote the import/export part of japanese animation software, and they just re-used the interpolation part of the printing process, spreading pixels on paper, to deal with the scaling. Unfortunately they also put the now meaningless DPI setting into the interface. So instead of TVPaint's appropriate choice between "import 1:1" and "scale to fit" they included a DPI number setting - which makes no sense between digital image files.

And because japanese companies are more than conservative in terms of changing their proven workflow, this has never been questioned nor revised. Instead, they force foreign programmers to include useless DPI numbers into their software as well.

Instead of asking for DPI settings you should understand the meaning of production dimensions, broadcast dimensions, and how to calculate line widths. Fortunately the latter is only necessary for bitmaps. For vector artwork created in Moho you just use the line width which looks good, and it will get rendered perfectly in any output dimensions.

[BigBoiiiJones]

Unless you're working in print you don't need to know about DPI at all, because a DPI setting only makes sense in bringing any bitmap to paper: it tells the printer how many pixels it should spread over a certain distance on paper.

The only thing which matters in video production is the dimensions of the project/video. HD 1920 x 1020 px is the same in all programs on all systems.

Let's say your Moho project is in HD (1920 x 1020 px), and you import an image. If the image is 1920 x 1020 px as well, it will fit the project window perfectly. That is because Moho imports artwork as 1:1, each pixel of artwork is one pixel in the project window. If your image is smaller, say 640 x 480 px, it will be smaller than the project window. If your image is larger, it will spill over the project window. Moho imports 1:1.

Other software interpret the imported artwork, like a video editor which assumes that anything you import, no matter whether it's 4K or HD or from a shitty webcam, shall fill the project window completely. And even other software may give you the choice between 1:1 or interpreting, like TVPaint. Our example picture of 640 x 480 px will be enlarged and spread over the 1920 x 1020 px of my FinalCut video project. Of couse such an enlarged image will look poorly between genuine HD scenes because it was scaled up. The other way round, importing a larger image into a smaller project, will show no difference between this and other HD images.

So the only size which really matters is the dimensions of your Moho project file. If it's 1920 x 1020 px you just need to take care that any imported background image is at least the same size, or even bigger if you plan to zoom in.

Now why do japanese studios still bother with DPI and fuck up everything? That's a bit of history of technique and style.

The first japanese softwares which could colour scanned drawings worked very simple: the drawings were 1-bit, having only black or white pixels. An image like that is easy to fill with colours, but all the lines will have jaggy corners. So they developed a workflow where each drawing was scanned in very large, then colourized, and then exported to a much smaller size so the pixels were interpreted/interpolated: the average of 2 or 4 or 9 pixels gave the value of 1 output pixel. If the drawing already had very thin lines, they got even thinner in the process, becoming less than a pixel wide and not black anymore, but somehow grayish.

Nowadays software can easily colourize drawings with anti-aliased lines, like 8-bit scans, and there's no need to scale down the coloured drawing to smoothen the lines. But in digital animation it is very hard (to be exact: it's impossible) to draw a less-than-1-pixel-wide line: that's why animators work in larger dimensions and scale down the artwork to get these very thin lines of that style which still is in demand.

I imagine that, at some point in history, some programmer with a background in printing wrote the import/export part of japanese animation software, and they just re-used the interpolation part of the printing process, spreading pixels on paper, to deal with the scaling. Unfortunately they also put the now meaningless DPI setting into the interface. So instead of TVPaint's appropriate choice between "import 1:1" and "scale to fit" they included a DPI number setting - which makes no sense between digital image files.

And because japanese companies are more than conservative in terms of changing their proven workflow, this has never been questioned nor revised. Instead, they force foreign programmers to include useless DPI numbers into their software as well.

Instead of asking for DPI settings you should understand the meaning of production dimensions, broadcast dimensions, and how to calculate line widths. Fortunately the latter is only necessary for bitmaps. For vector artwork created in Moho you just use the line width which looks good, and it will get rendered perfectly in any output dimensions.

The problem is I was using Moho as part of my work for a Japanese studio for freelance work and they require DPI settings so the animation director could print and make proper corrections for this specific production, so this does not scale well for that and the print isn't to good either and you cant look at these two examples and say the Moho result gives best results compared to CSP. There is clear artifacts and pixelization in the AA within Moho. I know production dimensions and such and can calculate widths that isn't the problem here. If you're working with Moho on something Anime related the AA does terrible with smaller line widths with far and mid shots. I wasn't initially asking for DPI settings completely I was just wondering if DPI had an effect on said thing and suggested it would be nice if Moho had a setting for it. Not that the setting would solve the issue. After rereading my post I can see how it can come off that way I completely worded it wrong so my bad but isn't what I meant. I think the issue is how Moho handles AA and non AA images in general. The renderer could use a overhaul/upgrade in general because even with no AA with widths under 2 become dotted and don't even connect properly I even made a discussion about it and Greenlaw in fact agreed that it was a bug with Moho. So we cant even disable AA and use our own with AE plugins like OLM or PSOFT or F's.

[slowtiger]

OK, I understand the hassle of dealing with a demanding studio overseas ...

- As for a required DPI setting: this only makes sense in image files anyway, you could easily set up a batch job in Photoshop to open each image and save it with a differend DPI setting (without changing the artwork's dimensions).

- As for thin vector lines rendered poorly in Moho: I don't use these, so I wasn't aware of that problem. My solution would be to render in double dimensions, then re-scale in Photoshop or anything else.

Good luck!