Graphics.CopyTexture

Parameters

Description

Copies pixel data from one texture to another.

This method copies pixel data from one texture to another on the GPU. If you set Texture.isReadable to true for both src and dst textures, the method also copies pixel data on the CPU.

If you set Texture.isReadable to false, CopyTexture is one of the fastest ways to copy a texture. But to use CopyTexture, the following must be the same in both the source and destination texture areas:

• Format. You can also use two compatible formats - for example, TextureFormat.ARGB32 and RenderTextureFormat.ARGB32.
• Size.
• RenderTexture.antiAliasing values, if the textures are render textures.

You might be able to copy between incompatible formats depending on your graphics API. For example, on some APIs you can copy between formats with the same bit width.

Depending on your graphics API, you might not be able to copy between different types of textures. For more information on compatibility, see SystemInfo.copyTextureSupport and CopyTextureSupport.

If src is a depth-only render texture, you must copy the whole texture, not part of it. A depth-only render texture has its color buffer set to a color format of None and its depth buffer set to a valid RenderTexture.depthStencilFormat.

When you use Texture2D.ignoreMipmapLimit, textures can have a variety of mipmap limit settings. This means you may not be able to copy from one mipmap level to another, because for example the source texture is limited to half resolution and the destination texture is limited to quarter resolution.

You can load Textures at different resolutions by using QualitySettings.masterTextureLimit. Note that this affects CopyTextures as you cannot copy a full mip between textures with different master texture limit values. If you need to copy between textures with a different master texture limit, use the region-based overload. The region-based overload adjusts the source rectangle based on the source texture's master texture limit. It also adjusts the destination offset based on the destination's master texture limit. For example, copying a 128x128 area from position 16,16 to position 32,32 results in the following behaviours in these example cases:

• When the master texture limit is set to 0: Unity performs all copies as expected.
• When the master texture limit is set to 2 and both textures are subject to the master texture limit: Unity adjusts the source rectangle to 32x32 and adjusts the offset to 4,4. Unity changes the destination offset to 8x8. Unity performs all copies as expected without awareness of the master texture limit setting.
• When the master texture limit is set to 2, the source is a regular texture, and the destination is an array texture (array textures are never subject to the master texture limit): Unity adjusts the source rectangle to 32x32 and adjusts the offset to 4,4. Unity doesn't change the destination offset, which remains at 32x32.

The following sample shows how you can copy to or from a Texture2DArray, when Texture2Ds are subject to a master texture limit different from 0:

using UnityEngine;

public class CopyTextureSample : MonoBehaviour
{
    // Sample to copy all available mips: inTex -> outArray (no master texture limits) -> outTex
    [SerializeField] Texture2D inTex;
    [SerializeField] Texture2DArray outArray;
    [SerializeField] Texture2D outTex;
    [SerializeField] bool considerMasterTextureLimits;

    void Start()
    {
        int width = inTex.width;
        int height = inTex.width;
        
        outArray = new Texture2DArray(width, height, 1, inTex.format, true);
        outTex = new Texture2D(width, height, inTex.format, true);

        if (!considerMasterTextureLimits)
        {
            // Texture2D -> Texture2DArray
            // Master Texture Limit: "1: Half Resolution" => copies into mips that are now too large; will copy each mip of inTex into a quarter of outArray
            for (int mip = 0; mip < inTex.mipmapCount; ++mip)
            {
                int copyWidth = width >> mip;
                int copyHeight = height >> mip;
                Graphics.CopyTexture(inTex, 0, mip, 0, 0, copyWidth, copyHeight, outArray, 0, mip, 0, 0);
            }

            // Texture2DArray -> Texture2D 
            // Master Texture Limit: "1: Half Resolution" => errors, since we try to copy into mips that are now too small
            for (int mip = 0; mip < outArray.mipmapCount; ++mip)
            {
                int copyWidth = width >> mip;
                int copyHeight = height >> mip;
                Graphics.CopyTexture(outArray, 0, mip, 0, 0, copyWidth, copyHeight, outTex, 0, mip, 0, 0);
            }
        }
        else // considering master texture limits
        {
            int masterTextureLimit = QualitySettings.masterTextureLimit;
            
            // Texture2D -> Texture2DArray
            // Master Texture Limit: "1: Half Resolution" => mip0 of outArray is not written to, other mips copy as expected
            //  (ALTERNATIVE: if outArray creation already considered masterTextureLimit for its dimensions,
            //   the CopyTexture call can ignore masterTextureLimit since the mips will line up again) 
            for (int mip = 0; mip < inTex.mipmapCount - masterTextureLimit; ++mip)
            {
                int copyWidth = width >> mip;
                int copyHeight = height >> mip;
                int srcMip = mip;
                int dstMip = mip + masterTextureLimit;
                Graphics.CopyTexture(inTex, 0, srcMip, 0, 0, copyWidth, copyHeight, outArray, 0, dstMip, 0, 0);
            }

            // Texture2DArray -> Texture2D 
            // Master Texture Limit: "1: Half Resolution" => mip0 of outArray is not copied (but outTex does not upload it to GPU anyway)
            for (int mip = masterTextureLimit; mip < outArray.mipmapCount; ++mip)
            {
                int copyWidth = width >> mip;
                int copyHeight = height >> mip;
                int srcMip = mip;
                int dstMip = mip - masterTextureLimit;
                Graphics.CopyTexture(outArray, 0, srcMip, 0, 0, copyWidth, copyHeight, outTex, 0, dstMip, 0, 0);
            }
        }
    }
}

Mipmap level arguments always apply to the texture as loaded under the current master texture limit. For example, a 256x256 texture with master texture limit set to 0 mip 1 refers to a 128x128 mip. However, if the mastertexture limit is set to 2 on that texture, mip 1 refers to a 32x32 mip. This means that in many cases when using CopyTexture you do not need to take the master texture limit into account in your calls. In less common calls (for example, copying from Texture2D to TextureArray) you do need to adjust for it. To copy textures in a cubemap array, calculate the destination element as 6 * cubemapIndex + faceIndex. As a result, the six faces from the cubemap at index 0 are elements 0,1,2... 5. The six faces from the cubemap at array index 1 are 6,7 .... 11 and so on.

Compressed texture formats add some restrictions to the CopyTexture with a region variant. For example, PVRTC formats are not supported since they are not block-based (for these formats you can only copy whole texture or whole mip level). For block-based formats (e.g. DXT, BCn, ETC), the region size and coordinates must be a multiple of compression block size (4 pixels for DXT).

Even if you set Texture.isReadable to true, this method doesn't copy pixel data on the CPU if you copy only a region of a compressed texture.

Don't use an Apply method such as Texture2D.Apply after CopyTexture, because you might copy old or undefined CPU texture data to the GPU.

Additional resources: CopyTextureSupport.