RGBA something wrong

[HoNgocQuocSang2721]

This is png with alpha (RGBA)

when use toktx , the image.ktx2 not contains alpha (RGB)
Please contact solution for me @@!

[MarkCallow]

What was the exact command line you used? What version of the KTX software? What is the VkFormat of the image.ktx2 you created?

toktx is deprecated but it should still work. Please try its replacement ktx create.

[HoNgocQuocSang2721]

I clone github with new version and run commandline here by chatgpt :
toktx --uastc --alpha texture.ktx2 texture.png (i remember that)
And result return texture.png not contains Alpha
"Could you provide me with the command line or setting format in .lib / dll to execute it successfully?" @@!

[HoNgocQuocSang2721]

Website I try convert : https://www.3dpea.com/en/convert/PNG-to-KTX2
Website view: https://www.creators3d.com/online-viewer
you can test when convert by website 3dpea and KTX-Software.
And view file ktx between them:
this picture use 3dpea

this picture use KTX-Software

and here is code, i use c# format for ktx2 :

START CODE ===================================================

public static byte[]? ConvertToKTX2(byte[] imageData, int width, int height)
{
Console.WriteLine("width: " + width);
Console.WriteLine("height: " + height);
if (imageData.Length != width * height * 4)
{
Console.WriteLine("****************************************");
}

   var createInfo = new ktxTextureCreateInfo
   {
       glInternalformat = 0x8058,
       // VK_FORMAT_R8G8B8A8_UNORM
       vkFormat = 37,
       baseWidth = (uint)width,
       baseHeight = (uint)height,
       baseDepth = 1,
       numDimensions = 2,
       numLevels = 1, 
       numLayers = 1,
       numFaces = 1,
       isArray = ktx_bool_t.KTX_FALSE,
       generateMipmaps = ktx_bool_t.KTX_FALSE,
   };

   IntPtr ptrToKtxTexture2 = Marshal.AllocHGlobal(IntPtr.Size);
   Marshal.WriteIntPtr(ptrToKtxTexture2, IntPtr.Zero);


   ktx_error_code_e resultCode = KTXBindingsFunctions.ktxTexture2_Create(createInfo, ktxTextureCreateStorageEnum.KTX_TEXTURE_CREATE_ALLOC_STORAGE, ptrToKtxTexture2);

   if (resultCode != ktx_error_code_e.KTX_SUCCESS)
   {
       Console.WriteLine("Failed to create a empty texture to host input image data. " + resultCode);
       return null;
   }
   else
   {
       IntPtr texptr = Marshal.ReadIntPtr(ptrToKtxTexture2);

       ktxTexture2 managedTexture = ktxTexture2.ktxTexture2_MarshalFromPointer(texptr);
       managedTexture.vtbl.SetImageFromMemory(texptr, 0, 0, 0, imageData, imageData.Length);


       ktxBasisParams compressParams = new ktxBasisParams
       {
           structSize = (uint)Marshal.SizeOf<ktxBasisParams>(),
           uastc = ktx_bool_t.KTX_TRUE, 
           threadCount = 14,
           compressionLevel = 2,
           qualityLevel = 255,
           normalMap = ktx_bool_t.KTX_FALSE,
           preSwizzle = ktx_bool_t.KTX_FALSE,
           verbose = ktx_bool_t.KTX_FALSE
       };


       IntPtr paramsptr = Marshal.AllocHGlobal(Marshal.SizeOf(compressParams));
       Marshal.StructureToPtr(compressParams, paramsptr, false);



       resultCode = KTXBindingsFunctions.ktxTexture2_CompressBasisEx(texptr, paramsptr);


       if (resultCode == ktx_error_code_e.KTX_SUCCESS)
       {

           IntPtr dataSizePtr = Marshal.AllocHGlobal(IntPtr.Size);
           Marshal.WriteIntPtr(dataSizePtr, IntPtr.Zero);
           IntPtr bufferPtr = Marshal.AllocHGlobal(IntPtr.Size);
           Marshal.WriteIntPtr(bufferPtr, IntPtr.Zero);

           resultCode = managedTexture.vtbl.WriteToMemory(texptr, bufferPtr, dataSizePtr);
           if (resultCode != ktx_error_code_e.KTX_SUCCESS)
           {

               Console.WriteLine("Texture write failed.");
               managedTexture.vtbl.Destroy(texptr);
               Marshal.FreeHGlobal(paramsptr);
               Marshal.FreeHGlobal(bufferPtr);
               Marshal.FreeHGlobal(dataSizePtr);
               return null;
           }
           int dataSize = Marshal.ReadInt32(dataSizePtr);
           IntPtr unmangedBufferPtr = Marshal.ReadIntPtr(bufferPtr);
           byte[] ktx2Bytes = new byte[dataSize];

           Marshal.Copy(unmangedBufferPtr, ktx2Bytes, 0, dataSize);
           Marshal.FreeHGlobal(paramsptr);
           Marshal.FreeHGlobal(bufferPtr);
           Marshal.FreeHGlobal(dataSizePtr);
           managedTexture.vtbl.Destroy(texptr);

           return ktx2Bytes;
       }
       else
       {
           Marshal.FreeHGlobal(paramsptr);
           managedTexture.vtbl.Destroy(texptr);
           return null;
       }
   }

}

END CODE ===================================================

Please check for me! Thanks so much @@!

Example.zip

[MarkCallow]

It looks to me as though the alpha channel values are all 1.0. In such cases the Basis Universal encoder, which you are invoking via --uastc or compressParams.uastc = KTX_TRUE, removes the alpha channel because it is not serving any purpose. This is important for the ETC1S/BasisLZ case because an alpha channel doubles the number of blocks needed. For UASTC it is not so important but that is the way the encoder is written.

If you make sure there is at least one alpha value < 1.0 then the channel will not be removed.

The --alpha option has not served any purpose since the first version of the software. Do not use it.

[HoNgocQuocSang2721]

But I used the website https://www.3dpea.com/en/convert/PNG-to-KTX2 to convert and the result was as I expected. So how do I use CommandLine to get the result like that? Please check the code above, I have given a specific example. When running png format images, there is no problem, only the image with alpha is stuck.

[MarkCallow]

Sorry for the delay in my reply. Apple Mail has stopped flagging unread messages in my GitHub mailbox for some reason and I only saw your message today.

In the example.zip you posted, use KTX-Software.ktx2, which is in a UASTC format, actually does have an alpha channel as you can see in the image below:

The color gradient you can see is from the vertex colors of the quad which is being textured. If you aren't seeing the alpha channel either your app is transcoding it to a format without alpha or has some other display problem.

As for use website 3dpea.ktx2 that file is an invalid KTX v2 file that is nevertheless accepted by libktx. More on the invalidity later. Here is the rendering from the same application I used above:

Notice this has a grey background between the wires. The 3dpea website appears to have written an alpha channel where every texel has the value 1.0 so you are seeing the background of the image between the wires. For sure there are no transparent texels in this image. Because of invalidity I mentioned, I cannot use our ktx extract tool to extract the image to a PNG file for closer examination of the alpha channel so I added a few lines of code to my loader to check all the alpha values. They are 1.0.

About the invalidity, the image is float format (R32G32B32A32) which is fine but the DFD has many incorrect values in each of the 4 samples (in this case a sample corresponds to a channel or component). Here is the output from ktx validate

Validation failed

error-6028: Invalid sample in basic DFD block. The samples must match the expected samples of the VkFormat.
    DFD block #1 sample #1 qualifierLinear in basic DFD block is 1 but the expected value is 0 for VK_FORMAT_R32G32B32A32_SFLOAT.
error-6028: Invalid sample in basic DFD block. The samples must match the expected samples of the VkFormat.
    DFD block #1 sample #1 sampleLower in basic DFD block is 4294967295 but the expected value is 3212836864 for VK_FORMAT_R32G32B32A32_SFLOAT.
error-6028: Invalid sample in basic DFD block. The samples must match the expected samples of the VkFormat.
    DFD block #1 sample #1 sampleUpper in basic DFD block is 1 but the expected value is 1065353216 for VK_FORMAT_R32G32B32A32_SFLOAT.
error-6028: Invalid sample in basic DFD block. The samples must match the expected samples of the VkFormat.
    DFD block #1 sample #2 bitOffset in basic DFD block is 4 but the expected value is 32 for VK_FORMAT_R32G32B32A32_SFLOAT.
error-6028: Invalid sample in basic DFD block. The samples must match the expected samples of the VkFormat.
    DFD block #1 sample #2 qualifierLinear in basic DFD block is 1 but the expected value is 0 for VK_FORMAT_R32G32B32A32_SFLOAT.
error-6028: Invalid sample in basic DFD block. The samples must match the expected samples of the VkFormat.
    DFD block #1 sample #2 sampleLower in basic DFD block is 4294967295 but the expected value is 3212836864 for VK_FORMAT_R32G32B32A32_SFLOAT.
error-6028: Invalid sample in basic DFD block. The samples must match the expected samples of the VkFormat.
    DFD block #1 sample #2 sampleUpper in basic DFD block is 1 but the expected value is 1065353216 for VK_FORMAT_R32G32B32A32_SFLOAT.
error-6028: Invalid sample in basic DFD block. The samples must match the expected samples of the VkFormat.
    DFD block #1 sample #3 bitOffset in basic DFD block is 8 but the expected value is 64 for VK_FORMAT_R32G32B32A32_SFLOAT.
error-6028: Invalid sample in basic DFD block. The samples must match the expected samples of the VkFormat.
    DFD block #1 sample #3 qualifierLinear in basic DFD block is 1 but the expected value is 0 for VK_FORMAT_R32G32B32A32_SFLOAT.
error-6028: Invalid sample in basic DFD block. The samples must match the expected samples of the VkFormat.
    DFD block #1 sample #3 sampleLower in basic DFD block is 4294967295 but the expected value is 3212836864 for VK_FORMAT_R32G32B32A32_SFLOAT.
error-6028: Invalid sample in basic DFD block. The samples must match the expected samples of the VkFormat.
    DFD block #1 sample #3 sampleUpper in basic DFD block is 1 but the expected value is 1065353216 for VK_FORMAT_R32G32B32A32_SFLOAT.
error-6028: Invalid sample in basic DFD block. The samples must match the expected samples of the VkFormat.
    DFD block #1 sample #4 bitOffset in basic DFD block is 12 but the expected value is 96 for VK_FORMAT_R32G32B32A32_SFLOAT.
error-6028: Invalid sample in basic DFD block. The samples must match the expected samples of the VkFormat.
    DFD block #1 sample #4 qualifierLinear in basic DFD block is 1 but the expected value is 0 for VK_FORMAT_R32G32B32A32_SFLOAT.
error-6028: Invalid sample in basic DFD block. The samples must match the expected samples of the VkFormat.
    DFD block #1 sample #4 sampleLower in basic DFD block is 4294967295 but the expected value is 3212836864 for VK_FORMAT_R32G32B32A32_SFLOAT.
error-6028: Invalid sample in basic DFD block. The samples must match the expected samples of the VkFormat.
    DFD block #1 sample #4 sampleUpper in basic DFD block is 1 but the expected value is 1065353216 for VK_FORMAT_R32G32B32A32_SFLOAT.
warning-5004: Non-conformant texture file accepted by libktx.
    KTX 2.0 file does not conform to the specification but it is currently accepted by libktx.

Please report this to www.3dpea.com as well as reporting their failure to preserve the alpha data. The KTXwriter field of use website 3dpea.ktx2 says KTXwriter: three.js 155 so you might want to report the problem to three.js as well.

[MarkCallow]

Please report this to www.3dpea.com as well as reporting their failure to preserve the alpha data. The KTXwriter field of use website 3dpea.ktx2 says KTXwriter: three.js 155 so you might want to report the problem to three.js as well.

Have you reported the issue?

[HoNgocQuocSang2721]

My current solution is that when the image has Alpha, it will be converted to png and normal images will use ktx2

[MarkCallow]

My current solution is that when the image has Alpha, it will be converted to png and normal images will use ktx2

Why? I have demonstrated that the UASTC .ktx2 file you created has alpha. If you aren't seeing it in your app, it is, as I wrote, because you are transcoding to an RGB format or have a problem in your rendering pipeline.