diff --git a/README.md b/README.md index d094a6f..1205c09 100644 --- a/README.md +++ b/README.md @@ -242,19 +242,24 @@ Compressing and Unpacking .KTX2/.basis Files `basisu -xuastc_ldr_6x6 -quality 75 -effort 4 x.png` -An alias for `-xuastc_ldr_6x6` is `-ldr_6x6i` (where 'i'="intermediate"). All [14 standard ASTC block sizes](https://developer.nvidia.com/astc-texture-compression-for-game-assets) are supported, from 4x4-12x12: 4x4, 5x4, 5x5, 6x5, 6x6, 8x5, 8x6, 10x5, 10x6, 8x8, 10x8, 10x10, 12x10 and 12x12. The XUASTC LDR to BC7 transcoder has special optimizations for several common block sizes: 4x4, 6x6 and 8x6. When transcoding XUASTC LDR at these particular block sizes, most XUASTC blocks are *directly* transcoded to BC7, skipping the real-time analytical bc7f encoding step. +An alias for `-xuastc_ldr_6x6` is `-ldr_6x6i` (where 'i'="intermediate"). All [14 standard ASTC block sizes](https://developer.nvidia.com/astc-texture-compression-for-game-assets) are supported, from 4x4-12x12: 4x4, 5x4, 5x5, 6x5, 6x6, 8x5, 8x6, 10x5, 10x6, 8x8, 10x8, 10x10, 12x10 and 12x12. The **XUASTC LDR to BC7 transcoder has special optimizations for several common block sizes: 4x4, 6x6 and 8x6**. When transcoding XUASTC LDR at these particular block sizes, most XUASTC blocks are *directly* transcoded to BC7, skipping the real-time analytical bc7f encoding step. -More ASTC and XUASTC LDR specific options: +More XUASTC LDR specific options (most of these also apply to standard ASTC): - - The options `-xuastc_arith`, `-xuastc_zstd` (the default), and `-xuastc_hybrid` control the XUASTC LDR profile used. The arithmetic profile trades off transcoding throughput for 5-18% better compression vs. the Zstd profile, and the hybrid profile is a balance between the two. + - The options `-xuastc_arith`, `-xuastc_zstd` (the default), and `-xuastc_hybrid` control the **XUASTC LDR profile used**. The arithmetic profile trades off transcoding throughput for 5-18% better compression vs. the Zstd profile, and the hybrid profile is a balance between the two. - - `-ts` or `-srgb` enables the sRGB profile (the default), and `-tl` or `-linear` enables the linear profile. + - `-ts` or `-srgb` enables the **sRGB profile (the default)**, and `-tl` or `-linear` **enables the linear profile**. - - `-weights X Y Z W` sets the unsigned channel error weights, used to favor certain channels during compression. + - `-weights X Y Z W` sets the unsigned integer **channel error weights**, used to favor certain channels during compression. - - Another set of XUASTC specific options overrides the windowed RDO behavior (windowed RDO is a separate and optional perceptual optimization vs. Weight Grid DCT): `-xy` enables and `-xyd` disables windowed RDO. By default, if Weight Grid DCT is not enabled (i.e. `-quality` isn't specified, or its set to 100), windowed RDO is disabled. Windowed RDO is automatically enabled if the quality level is less than 100, unless `-xyd` is specified. Also see the tool's help text for additional windowed RDO options: `-ls_min_psnr`, `-ls_min_alpha_psnr`, '-ls_thresh_psnr`, '-ls_thresh_alpha_psnr`, etc. + - Another set of XUASTC specific options overrides the **windowed RDO behavior** (windowed or bounded RDO is a separate and optional perceptual optimization vs. Weight Grid DCT): `-xy` enables and `-xyd` disables windowed RDO. By default, if Weight Grid DCT is not enabled (i.e. `-quality` isn't specified, or is set to 100), windowed RDO is disabled. Windowed RDO is automatically enabled if the quality level is less than 100, unless `-xyd` is specified. Also see the tool's help text for additional windowed RDO options: `-ls_min_psnr`, `-ls_min_alpha_psnr`, '-ls_thresh_psnr`, '-ls_thresh_alpha_psnr`, etc. - `-xs` disables 2-3 subset usage and `-xp` disables dual plane usage (slightly higher compression and faster transcoding to BC7). + - `-higher_quality_transcoding`: Permits slower but higher quality transcoding + - `-no_deblocking`: Disables all adaptive deblocking on larger ASTC block sizes (faster) + - `-force_deblocking`: Always use adaptive deblocking filter, even for block sizes <= 8x6 (slower) + - `-stronger_deblocking`: Use stronger deblocking when it's enabled (same performance) + - `-fast_xuastc_ldr_bc7_transcoding` and `-no_fast_xuastc_ldr_bc7_transcoding`: Controls faster direct XUASTC->BC7 transcoding (defaults to enabled, which is slightly lower quality) - To compress an LDR sRGB image to a standard ASTC LDR 6x6 .KTX2 file, using effort level 4 (valid effort levels 0-10):