Loading...
> >

why always jpg, why not .png? higher quality?

2»

Comments

  • RedzRedz Posts: 1,459

    Iray's default thresholds for texture compression are just 512 and 1024 pixels. Given that the standard for skin textures is now 4096, and as changing this render-time compression is an 'advanced' feature in the render settings, my suspicion has been that the majority would not benefit from the file bloat of lossless images. But if, as others say, the tiff and png files are compressed more efficiently, then that could be a game changer. I for one would love to seen some experimental comparisons, both for quality, render time and vram consumption. I suppose there is a history of small is good, and my vram is already easily overloaded. If a single character's textures will overload that, then I'd rather have the jpg. 

  • LyonessLyoness Posts: 1,632
    Redz said:

    Iray's default thresholds for texture compression are just 512 and 1024 pixels. Given that the standard for skin textures is now 4096, and as changing this render-time compression is an 'advanced' feature in the render settings, my suspicion has been that the majority would not benefit from the file bloat of lossless images. But if, as others say, the tiff and png files are compressed more efficiently, then that could be a game changer. I for one would love to seen some experimental comparisons, both for quality, render time and vram consumption. I suppose there is a history of small is good, and my vram is already easily overloaded. If a single character's textures will overload that, then I'd rather have the jpg. 

    oh. I bump those up to 2048 and 4096 because the compression can cause artifacts that I don't want in my pictures.

  • RedzRedz Posts: 1,459
    Lyoness said:
    Redz said:

    Iray's default thresholds for texture compression are just 512 and 1024 pixels. Given that the standard for skin textures is now 4096, and as changing this render-time compression is an 'advanced' feature in the render settings, my suspicion has been that the majority would not benefit from the file bloat of lossless images. But if, as others say, the tiff and png files are compressed more efficiently, then that could be a game changer. I for one would love to seen some experimental comparisons, both for quality, render time and vram consumption. I suppose there is a history of small is good, and my vram is already easily overloaded. If a single character's textures will overload that, then I'd rather have the jpg. 

    oh. I bump those up to 2048 and 4096 because the compression can cause artifacts that I don't want in my pictures.

    How many users are even aware that you can do this though? And if you have a lower end system or a big scene, as opposed to a simple one-character closeup, then your memory will quickly choke and the render time slow to a crawl. I think a set of texture compression presets would be very helpful in these situations. Even better if you could choose which objects get compression and which don't. 

  • Takeo.KenseiTakeo.Kensei Posts: 1,303
    Redz said:
    Lyoness said:
    Redz said:

    Iray's default thresholds for texture compression are just 512 and 1024 pixels. Given that the standard for skin textures is now 4096, and as changing this render-time compression is an 'advanced' feature in the render settings, my suspicion has been that the majority would not benefit from the file bloat of lossless images. But if, as others say, the tiff and png files are compressed more efficiently, then that could be a game changer. I for one would love to seen some experimental comparisons, both for quality, render time and vram consumption. I suppose there is a history of small is good, and my vram is already easily overloaded. If a single character's textures will overload that, then I'd rather have the jpg. 

    oh. I bump those up to 2048 and 4096 because the compression can cause artifacts that I don't want in my pictures.

    How many users are even aware that you can do this though? And if you have a lower end system or a big scene, as opposed to a simple one-character closeup, then your memory will quickly choke and the render time slow to a crawl. I think a set of texture compression presets would be very helpful in these situations. Even better if you could choose which objects get compression and which don't. 

    http://blog.irayrender.com/post/54506874080/saving-on-texture-memory

  • MistaraMistara Posts: 38,675
    edited September 2016

    enlightened culd like made a product with pngs, could like market as UHD,

    as an experiment release a standard and a UHD version of a product and see what happens?

    would be like the difference of standard dvd and blu-ray,
    vhs or betamax  lol

    tv these days are like4k, the new xbox one s supports uhd resolution

    Post edited by Mistara on
  • LyonessLyoness Posts: 1,632
    MistyMist said:

    enlightened culd like made a product with pngs, could like market as UHD,

    as an experiment release a standard and a UHD version of a product and see what happens?

    yeah. I dunno about that.  to do all that extra work and only have a couple of them sell.... that's a potentially major time suck.

    I'd have to think about it and see what Daz thought about it. 

    I just checked the size difference and it's around 150% bigger and it's going to be pretty taxing on the systems.  Include that with the HD morphs I make... hmmmm.... dunno. 
    AND, that's not even including the HAIR and clothes.  I don't know if we're ready.  I'll ask though

     

  • ToborTobor Posts: 2,300
    edited September 2016
    Redz said:

    Iray's default thresholds for texture compression are just 512 and 1024 pixels. Given that the standard for skin textures is now 4096, and as changing this render-time compression is an 'advanced' feature in the render settings, my suspicion has been that the majority would not benefit from the file bloat of lossless images.

    Apart from any influence of the alpha channel bits, a 4096x4096 JPG decompressed in memory is the the same as a PNG, BMP, or TIF of the same dimension. Compression in the *saved* file format, lossy or lossless, has no bearing in how it's unpacked into RAM. 

    Texture compression in Iray works in reverse to what most people think: the setting is meant to allow you to disable the compression for textures under the specified dimensional size, and it's the small files where the feature most often needs to be disabled. It works to reduce the memory footprint of the scene database, the same as down-rezing texture manually. Someone else posted the seminal blog post on this feature, which shows a comparison. The feature really is independent from the choice of JPG, PNG, TIF, and any lossless or lossy compressions. It's just about dimensional size of the texture.

    Post edited by Tobor on
  • WendyLuvsCatzWendyLuvsCatz Posts: 40,081

    BTW Carrara 8.5 uses mipmapping as does iClone so textures are reduced the further away they are from the camera anyway, I am not sure about iray but know 3Delight runs tdkmake that puts the textures into a temporary tiff format too, so even if one had large png files the render engine does not necessarily use them as is, however the first two engines do see the alpha channel in png thus not requiring an opacity mask. 

  • Cris PalominoCris Palomino Posts: 12,480
    Jimbow said:
    Male-M3dia said:

    PNG also use a larger amount of memory and resources. This would become a major issue on machines with less memory or if you have limited memory like using Iray with a GPU.

    Iray converts the textures for rendering. If anything, jpegs slow things down. Uncompressed image formats seem to work better.

    Is this based on experience?  Where did you find information on this?  Thanks.

  • BejaymacBejaymac Posts: 1,942
    mjc1016 said:

    Several years ago, I ran a series of render tests (3Delight) and TIFs were actually the FASTEST to render...the reason is that 3DL didn't need to spend time coverting them, just rewriting the header info.  So even though they were the largest files, they rendered the quickest.  A few seconds to a few minutes depending on the number/size/shot etc.  Every renderer has a 'native' format for texture files and if the images are already in that format and don't need to be converted there will be a speed increase.  Most of them are using some uncompressed format.  The conversion process is going to result in some amount of quality loss...whether it is acceptible or even noticeable is the big question.

    That said, for most things, the end results were often not that different to the naked eye.  Yes on closeups there were noticeable differences, but 'average' shots, nope.

    Also for working with the images...either editing/customizing or just creating different resolutions the uncompressed (tif/bmp) files are much better.

    Not entirely correct, 3DL doesn't use any of the texture formats you can plug into a surface, it converts them all into it's own format, the TDL, granted it is a TIF format but not the same type most would use, it's a compressed/compacted version that also generates a full set of mipmaps.

    Used a 4kx4k TIF one time during testing, can't remember the size of the actual TIF (was a while ago) but I got a TDL just under 4 MB back.

  • nicsttnicstt Posts: 11,715
    edited September 2016

    I'm not seeing any  almost no difference in memory usage through GPUz; either at 512/1024 or 2048/4096. One G3F figure, hair and a swimsuit, there is just over 2150MB in use; ran it a few times and closed and opened Studio, and the difference is about 50MB.

    The only time I'm seeing a significant difference is when I change the Medium to be 4096 - the same as the high, then I see about 700MB difference.

    Post edited by nicstt on
  • McGyverMcGyver Posts: 7,085
    edited September 2016
    MistyMist said:

    downscaling is the easy part

    is the filterforge resource jpg or something noiseless?

    Filter Forge offers the ability to render out your image in several formats... In addition to .jpeg you can create BMP, TIFF, TGA, PNG, OpenEXR, and PFM.

    Or do you mean is it based off of a JPEG image (as in the results are based off the input of a starting image) or more like a procedural texture (works independently to produce an image) or more like Vector graphics?

    In which case... It can use/alter/distort a JPEG much like Photoshop filters do and it can generate an image all on its own without the need for an image to be input... Basically procedural, though I don't know if that's textbook procedural.

    Filter Forge does not do vector graphics... But I think Genetica might.

    Post edited by McGyver on
  • TaozTaoz Posts: 10,258
    LeatherGryphon said:

     Even one with lots of solid color space which GIF is supposedly good at.

    Not sure what you mean here, GIF is only 8 bit (256 colors)? And normally takes up quite a lot of space compared to a JPG of the same pixel size.

  • ToborTobor Posts: 2,300
    Cris Palomino said:
    Jimbow said:

    Iray convers the textures for rendering. If anything, jpegs slow things down. Uncompressed image formats seem to work better.

    Is this based on experience?  Where did you find information on this?  Thanks.

    Many graphics programs will internally use run length encoding to store images in memory It's not a "format" per se, but just a way to pack bits. Lossy compression that causes artifacts will result in less efficient encoding -- even the all-white areas can have artifact noise in them, and every piece of noise will affect the encoding efficiency. How much depends on the image, and the nature of any artifacts.  I don't know for certain Iray supports this type of encoding, but it's common in .graphics programs, even the simpler ones. Iray's more robust texture compression saves even more RAM, and this is a conversion made on-the-fly.

  • TaozTaoz Posts: 10,258

    I'd rather have files that are too large as they can always de

    Lyoness said:
    Redz said:

    Iray's default thresholds for texture compression are just 512 and 1024 pixels. Given that the standard for skin textures is now 4096, and as changing this render-time compression is an 'advanced' feature in the render settings, my suspicion has been that the majority would not benefit from the file bloat of lossless images. But if, as others say, the tiff and png files are compressed more efficiently, then that could be a game changer. I for one would love to seen some experimental comparisons, both for quality, render time and vram consumption. I suppose there is a history of small is good, and my vram is already easily overloaded. If a single character's textures will overload that, then I'd rather have the jpg. 

    oh. I bump those up to 2048 and 4096 because the compression can cause artifacts that I don't want in my pictures.

    JPG artifacts are terrible to look at, so I definitely prefer a larger file size than artifacts. Quality comes first, IMO.

  • LeatherGryphonLeatherGryphon Posts: 12,091
    edited September 2016
    Taozen said:
    LeatherGryphon said:

     Even one with lots of solid color space which GIF is supposedly good at.

    Not sure what you mean here, GIF is only 8 bit (256 colors)? And normally takes up quite a lot of space compared to a JPG of the same pixel size.

    GIFs are good for cartoon work or images with a limited color palette.  Not good at high detail or subtle color gradiations.

    https://websitespeedexperts.com/advantages-and-disadvantages-of-gif-images/

    GIFs are an indexed color scheme.  Each pixel is represented by an 8-bit number that identifies one of 256 colors from the palette and you can choose whatever colors you want in the palette but only as many as can be represented by the size of the index.  There are different versions of the GIF format that permit smaller index numbers (eg: 7-bit, 6-bit, 5-bit, 4-bit, 3-bit, 2-bit) and I think (but am not positive) versions that permit larger indicies up to 24-bit but basically GIF has traditionally had an 8-bit index to a 256 color palette for any size image.  You can get subtle color variations if you have a limited range of colors in your palette. (eg: mostly shades of reds and oranges)

    https://en.wikipedia.org/wiki/GIF

    A good image editor will let you specify the size of the index and the colors in your palette in your GIF outputs.

    Post edited by LeatherGryphon on
Sign In or Register to comment.