Alice Lee Poetics (李 诗诗)

Results 94 comments of Alice Lee Poetics (李 诗诗)

    Note that for browsers security would be of extraordinary consideration.     Alike the reasoning: [ https://github.com/richtr/NoSleep.js/issues/157#issuecomment-1529149759](https://github.com/richtr/NoSleep.js/issues/157#issuecomment-1529149759)     "Browsers for security considerations may choose to only support a limited set of medias to...

    Is "`window.define = undefined`" really important premise..?     If so, maybe add it into Eruda?

    Also for usability concern:     The "margin-bottom" of "<body>" should be temporarily overridden: to compensate Eruda panels' blockage.

    Probably relevant:     [ https://github.com/libjxl/libjxl/issues/1245#issuecomment-2607267203](https://github.com/libjxl/libjxl/issues/1245#issuecomment-2607267203)     [ https://github.com/libjxl/libjxl/issues/3921#issuecomment-2513471925](https://github.com/libjxl/libjxl/issues/3921#issuecomment-2513471925)     ("compression strategies")

    Generally RGB one would be smaller, but not always. (bug?)     Though for most cases the dummy Alpha wouldn't make much difference. (< 1 KiB; scarcely exceeds 500 B)     ----     In...

    For the RGB / RGBA issue, likely unusual.     For the output size: JXL seems to be suboptimal dealing with contiguous color regions (large in particular).     It also appeared inferior (comparing...

    The output size affairs (vs. WebP Lossless) should be probably another issue.     However I doubt the necessity of > 8bpc (bit-per-component):     https://trac.ffmpeg.org/ticket/11107#comment:1     (off-topic for this issue)     The Alpha quirkiness would...

    I guess it's the handling on contiguous color regions.     Selecting few representative cases tends be more effective than vast testing over random input.

    More exemplar case for this issue:     https://github.com/MasterInQuestion/attach/raw/main/libjxl/libjxl/3921/poise.webp     (~ 438.98 KiB; WebP Lossless: 725x725)     Reproducible from "-e 9" to "-e 11".     "e9": 362,818 B     "e9a": 360,955 B     "e10": 363,068 B     "e10a":...

    I noted a core difference between "e9" "e10" is the so called "streaming" thing:     [ https://github.com/libjxl/libjxl/issues/3517#issuecomment-2482078563](https://github.com/libjxl/libjxl/issues/3517#issuecomment-2482078563)     .     Theoretically the more context available, the better estimation possible.     The additional settings likely...