#lossy — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #lossy, aggregated by home.social.
-
Exploring Hidden JPEG Features
https://fed.brid.gy/r/https://hackaday.com/2026/07/24/exploring-hidden-jpeg-features/
-
I am growing increasingly infuriated by the fact that just about anything that does anything with files, proprietary or not, automatically also applies lossy compression to said files without user prompting. Worse, sometimes this isn't even a setting the user can disable. #compression #lossy #enshittification #baddesign
-
I am growing increasingly infuriated by the fact that just about anything that does anything with files, proprietary or not, automatically also applies lossy compression to said files without user prompting. Worse, sometimes this isn't even a setting the user can disable. #compression #lossy #enshittification #baddesign
-
I am growing increasingly infuriated by the fact that just about anything that does anything with files, proprietary or not, automatically also applies lossy compression to said files without user prompting. Worse, sometimes this isn't even a setting the user can disable. #compression #lossy #enshittification #baddesign
-
Ni -14 #LUFS, ni -16. Tampoco -18, como lo propone #ReplayGain o la #EBU #R128.
La «zona segura» para normalización de audio digital con pérdidas (“ #lossy”) es -19 o -20 LUFS (especialmente en audio lossy con tasas de transferencias muy bajas, como 32 kbps en estéreo, o 16 kbps en monoaural, que es en donde suelen darse sutiles desperfectos en la ganancia del audio digital).
La exigencia de la #UniónEuropea se quedó corta. 🤭
#EBUR128 #EBU_R128 #LossyAudio #AudioDigital #DigitalAudio #Radios
-
Ni -14 #LUFS, ni -16. Tampoco -18, como lo propone #ReplayGain o la #EBU #R128.
La «zona segura» para normalización de audio digital con pérdidas (“ #lossy”) es -19 o -20 LUFS (especialmente en audio lossy con tasas de transferencias muy bajas, como 32 kbps en estéreo, o 16 kbps en monoaural, que es en donde suelen darse sutiles desperfectos en la ganancia del audio digital).
La exigencia de la #UniónEuropea se quedó corta. 🤭
#EBUR128 #EBU_R128 #LossyAudio #AudioDigital #DigitalAudio #Radios
-
Ni -14 #LUFS, ni -16. Tampoco -18, como lo propone #ReplayGain o la #EBU #R128.
La «zona segura» para normalización de audio digital con pérdidas (“ #lossy”) es -19 o -20 LUFS (especialmente en audio lossy con tasas de transferencias muy bajas, como 32 kbps en estéreo, o 16 kbps en monoaural, que es en donde suelen darse sutiles desperfectos en la ganancia del audio digital).
La exigencia de la #UniónEuropea se quedó corta. 🤭
#EBUR128 #EBU_R128 #LossyAudio #AudioDigital #DigitalAudio #Radios
-
Ni -14 #LUFS, ni -16. Tampoco -18, como lo propone #ReplayGain o la #EBU #R128.
La «zona segura» para normalización de audio digital con pérdidas (“ #lossy”) es -19 o -20 LUFS (especialmente en audio lossy con tasas de transferencias muy bajas, como 32 kbps en estéreo, o 16 kbps en monoaural, que es en donde suelen darse sutiles desperfectos en la ganancia del audio digital).
La exigencia de la #UniónEuropea se quedó corta. 🤭
#EBUR128 #EBU_R128 #LossyAudio #AudioDigital #DigitalAudio #Radios
-
Lossless converting the palleted PNG (at 100671 bytes ≈ 101 kB) to JPEG XL result in a file at 25178 bytes ≈ 25 kB.
Converting lossless to webp, using
cwebp -lossless -m 6 -z 9 -o e08463_353cfb9f6bf646159fdf9dc573b1080e.{webp,png}
the file is 85118 bytes ≈ 85 kB. Lossy webp (i.e. without `-lossless`) is 437632 bytes ≈ 438 kB
-
Lossless converting the palleted PNG (at 100671 bytes ≈ 101 kB) to JPEG XL result in a file at 25178 bytes ≈ 25 kB.
Converting lossless to webp, using
cwebp -lossless -m 6 -z 9 -o e08463_353cfb9f6bf646159fdf9dc573b1080e.{webp,png}
the file is 85118 bytes ≈ 85 kB. Lossy webp (i.e. without `-lossless`) is 437632 bytes ≈ 438 kB
-
Lossless converting the palleted PNG (at 100671 bytes ≈ 101 kB) to JPEG XL result in a file at 25178 bytes ≈ 25 kB.
Converting lossless to webp, using
cwebp -lossless -m 6 -z 9 -o e08463_353cfb9f6bf646159fdf9dc573b1080e.{webp,png}
the file is 85118 bytes ≈ 85 kB. Lossy webp (i.e. without `-lossless`) is 437632 bytes ≈ 438 kB
-
Lossless converting the palleted PNG (at 100671 bytes ≈ 101 kB) to JPEG XL result in a file at 25178 bytes ≈ 25 kB.
Converting lossless to webp, using
cwebp -lossless -m 6 -z 9 -o e08463_353cfb9f6bf646159fdf9dc573b1080e.{webp,png}
the file is 85118 bytes ≈ 85 kB. Lossy webp (i.e. without `-lossless`) is 437632 bytes ≈ 438 kB
-
Lossless converting the palleted PNG (at 100671 bytes ≈ 101 kB) to JPEG XL result in a file at 25178 bytes ≈ 25 kB.
Converting lossless to webp, using
cwebp -lossless -m 6 -z 9 -o e08463_353cfb9f6bf646159fdf9dc573b1080e.{webp,png}
the file is 85118 bytes ≈ 85 kB. Lossy webp (i.e. without `-lossless`) is 437632 bytes ≈ 438 kB
-
@OpenComputeDesign
Remember when we were talking about #audio #lossy #compression earlier?This is NUUUUUTS:
-
@OpenComputeDesign
Remember when we were talking about #audio #lossy #compression earlier?This is NUUUUUTS:
-
@OpenComputeDesign
Remember when we were talking about #audio #lossy #compression earlier?This is NUUUUUTS:
-
@OpenComputeDesign
Remember when we were talking about #audio #lossy #compression earlier?This is NUUUUUTS:
-
@OpenComputeDesign
Remember when we were talking about #audio #lossy #compression earlier?This is NUUUUUTS:
-
Since the #PNG spec is getting some updates, I was thinking... you know what would be sick? Baking classic #lossy compression right into PNG.
I don't mean #JPEG or anything like that. Not even the #LossyPNG scheme I enjoy using where I reduce the number of colors to reduce the entropy. That wouldn't need any changes to the spec.
I'm thinking #Amiga HAM modes! Instant 3:1 (well, 2.4:1) lossy compression with (hopefully) minimal artifacting. That would be SICK!
Makes me want to throw together a Python script or something to read PNG files, apply the HAM "compression," and then see how the output looks. It wouldn't provide much compression savings with regular PNG, it would need a change to the format. But since I never had an Amiga, I'm curious to see how it looks, and how well it might scale to modern image resolutions.
Conceptually, it's not much different than 3:1 horizontal chroma subsampling, but might actually look a lot better.
Quick explainer: instead of having a 24-bit-per-pixel image, you have a 10-bit-per-pixel image, where the first two bits are a mode selector, and the last 8 bits are the actual pixel data. The modes are Hold and Modify{R,G,B}. Modify{R,G,B} modify just the red, green, or blue component of the 24-bit color for that pixel with the 8 bits specified. The modify mode assigns that pixel to be equivalent to the 8-bit value from a pre-defined palette or some pre-defined bit-assignment mode like RRRGGGBB or something. So basically, the horizontal color resolution is divided by 3, but sometimes you get lucky and there's no noticeable loss.
-
Since the #PNG spec is getting some updates, I was thinking... you know what would be sick? Baking classic #lossy compression right into PNG.
I don't mean #JPEG or anything like that. Not even the #LossyPNG scheme I enjoy using where I reduce the number of colors to reduce the entropy. That wouldn't need any changes to the spec.
I'm thinking #Amiga HAM modes! Instant 3:1 (well, 2.4:1) lossy compression with (hopefully) minimal artifacting. That would be SICK!
Makes me want to throw together a Python script or something to read PNG files, apply the HAM "compression," and then see how the output looks. It wouldn't provide much compression savings with regular PNG, it would need a change to the format. But since I never had an Amiga, I'm curious to see how it looks, and how well it might scale to modern image resolutions.
Conceptually, it's not much different than 3:1 horizontal chroma subsampling, but might actually look a lot better.
Quick explainer: instead of having a 24-bit-per-pixel image, you have a 10-bit-per-pixel image, where the first two bits are a mode selector, and the last 8 bits are the actual pixel data. The modes are Hold and Modify{R,G,B}. Modify{R,G,B} modify just the red, green, or blue component of the 24-bit color for that pixel with the 8 bits specified. The modify mode assigns that pixel to be equivalent to the 8-bit value from a pre-defined palette or some pre-defined bit-assignment mode like RRRGGGBB or something. So basically, the horizontal color resolution is divided by 3, but sometimes you get lucky and there's no noticeable loss.
-
Since the #PNG spec is getting some updates, I was thinking... you know what would be sick? Baking classic #lossy compression right into PNG.
I don't mean #JPEG or anything like that. Not even the #LossyPNG scheme I enjoy using where I reduce the number of colors to reduce the entropy. That wouldn't need any changes to the spec.
I'm thinking #Amiga HAM modes! Instant 3:1 (well, 2.4:1) lossy compression with (hopefully) minimal artifacting. That would be SICK!
Makes me want to throw together a Python script or something to read PNG files, apply the HAM "compression," and then see how the output looks. It wouldn't provide much compression savings with regular PNG, it would need a change to the format. But since I never had an Amiga, I'm curious to see how it looks, and how well it might scale to modern image resolutions.
Conceptually, it's not much different than 3:1 horizontal chroma subsampling, but might actually look a lot better.
Quick explainer: instead of having a 24-bit-per-pixel image, you have a 10-bit-per-pixel image, where the first two bits are a mode selector, and the last 8 bits are the actual pixel data. The modes are Hold and Modify{R,G,B}. Modify{R,G,B} modify just the red, green, or blue component of the 24-bit color for that pixel with the 8 bits specified. The modify mode assigns that pixel to be equivalent to the 8-bit value from a pre-defined palette or some pre-defined bit-assignment mode like RRRGGGBB or something. So basically, the horizontal color resolution is divided by 3, but sometimes you get lucky and there's no noticeable loss.
-
Since the #PNG spec is getting some updates, I was thinking... you know what would be sick? Baking classic #lossy compression right into PNG.
I don't mean #JPEG or anything like that. Not even the #LossyPNG scheme I enjoy using where I reduce the number of colors to reduce the entropy. That wouldn't need any changes to the spec.
I'm thinking #Amiga HAM modes! Instant 3:1 (well, 2.4:1) lossy compression with (hopefully) minimal artifacting. That would be SICK!
Makes me want to throw together a Python script or something to read PNG files, apply the HAM "compression," and then see how the output looks. It wouldn't provide much compression savings with regular PNG, it would need a change to the format. But since I never had an Amiga, I'm curious to see how it looks, and how well it might scale to modern image resolutions.
Conceptually, it's not much different than 3:1 horizontal chroma subsampling, but might actually look a lot better.
Quick explainer: instead of having a 24-bit-per-pixel image, you have a 10-bit-per-pixel image, where the first two bits are a mode selector, and the last 8 bits are the actual pixel data. The modes are Hold and Modify{R,G,B}. Modify{R,G,B} modify just the red, green, or blue component of the 24-bit color for that pixel with the 8 bits specified. The modify mode assigns that pixel to be equivalent to the 8-bit value from a pre-defined palette or some pre-defined bit-assignment mode like RRRGGGBB or something. So basically, the horizontal color resolution is divided by 3, but sometimes you get lucky and there's no noticeable loss.
-
Since the #PNG spec is getting some updates, I was thinking... you know what would be sick? Baking classic #lossy compression right into PNG.
I don't mean #JPEG or anything like that. Not even the #LossyPNG scheme I enjoy using where I reduce the number of colors to reduce the entropy. That wouldn't need any changes to the spec.
I'm thinking #Amiga HAM modes! Instant 3:1 (well, 2.4:1) lossy compression with (hopefully) minimal artifacting. That would be SICK!
Makes me want to throw together a Python script or something to read PNG files, apply the HAM "compression," and then see how the output looks. It wouldn't provide much compression savings with regular PNG, it would need a change to the format. But since I never had an Amiga, I'm curious to see how it looks, and how well it might scale to modern image resolutions.
Conceptually, it's not much different than 3:1 horizontal chroma subsampling, but might actually look a lot better.
Quick explainer: instead of having a 24-bit-per-pixel image, you have a 10-bit-per-pixel image, where the first two bits are a mode selector, and the last 8 bits are the actual pixel data. The modes are Hold and Modify{R,G,B}. Modify{R,G,B} modify just the red, green, or blue component of the 24-bit color for that pixel with the 8 bits specified. The modify mode assigns that pixel to be equivalent to the 8-bit value from a pre-defined palette or some pre-defined bit-assignment mode like RRRGGGBB or something. So basically, the horizontal color resolution is divided by 3, but sometimes you get lucky and there's no noticeable loss.
-
Looking at all the personal data that I've accumulated over the years, it's sad to see how much of it is of bad quality due to #lossy #compression (esp. pictures).
Let me tell you that in 2025 it is reasonable to store -- and to some extend even send -- at least your pictures (via DNG) and your audio recordings (as FLAC) without loosing a single bit of information. It's mind-blowing when you think about how far we've come.
-
Looking at all the personal data that I've accumulated over the years, it's sad to see how much of it is of bad quality due to #lossy #compression (esp. pictures).
Let me tell you that in 2025 it is reasonable to store -- and to some extend even send -- at least your pictures (via DNG) and your audio recordings (as FLAC) without loosing a single bit of information. It's mind-blowing when you think about how far we've come.
-
Looking at all the personal data that I've accumulated over the years, it's sad to see how much of it is of bad quality due to #lossy #compression (esp. pictures).
Let me tell you that in 2025 it is reasonable to store -- and to some extend even send -- at least your pictures (via DNG) and your audio recordings (as FLAC) without loosing a single bit of information. It's mind-blowing when you think about how far we've come.
-
Looking at all the personal data that I've accumulated over the years, it's sad to see how much of it is of bad quality due to #lossy #compression (esp. pictures).
Let me tell you that in 2025 it is reasonable to store -- and to some extend even send -- at least your pictures (via DNG) and your audio recordings (as FLAC) without loosing a single bit of information. It's mind-blowing when you think about how far we've come.
-
Even though the blog stated a "--bitrate 6.7" command line, and the output log exhibited "6kbit/sec", it was only with 6.1 that I was able to get libopus 1.3.1 to generate a 1'448'253 album right under the 1'457'664 floppy limit.
Here is a similar 34'985 bytes OPUS track.
-
Even though the blog stated a "--bitrate 6.7" command line, and the output log exhibited "6kbit/sec", it was only with 6.1 that I was able to get libopus 1.3.1 to generate a 1'448'253 album right under the 1'457'664 floppy limit.
Here is a similar 34'985 bytes OPUS track.
-
Even though the blog stated a "--bitrate 6.7" command line, and the output log exhibited "6kbit/sec", it was only with 6.1 that I was able to get libopus 1.3.1 to generate a 1'448'253 album right under the 1'457'664 floppy limit.
Here is a similar 34'985 bytes OPUS track.
-
Even though the blog stated a "--bitrate 6.7" command line, and the output log exhibited "6kbit/sec", it was only with 6.1 that I was able to get libopus 1.3.1 to generate a 1'448'253 album right under the 1'457'664 floppy limit.
Here is a similar 34'985 bytes OPUS track.
-
Even though the blog stated a "--bitrate 6.7" command line, and the output log exhibited "6kbit/sec", it was only with 6.1 that I was able to get libopus 1.3.1 to generate a 1'448'253 album right under the 1'457'664 floppy limit.
Here is a similar 34'985 bytes OPUS track.
-
Your Text Needs More JPEG - We’ve all been victims of bad memes on the Internet, but they’re not all just bad ... - https://hackaday.com/2024/03/19/your-text-needs-more-jpeg/ #discretecosinetransform #fouriertransform #softwarehacks #compression #javascript #lossifizer #lossy #jpeg #text #dct #fft
-
Your Text Needs More JPEG - We’ve all been victims of bad memes on the Internet, but they’re not all just bad ... - https://hackaday.com/2024/03/19/your-text-needs-more-jpeg/ #discretecosinetransform #fouriertransform #softwarehacks #compression #javascript #lossifizer #lossy #jpeg #text #dct #fft
-
Your Text Needs More JPEG - We’ve all been victims of bad memes on the Internet, but they’re not all just bad ... - https://hackaday.com/2024/03/19/your-text-needs-more-jpeg/ #discretecosinetransform #fouriertransform #softwarehacks #compression #javascript #lossifizer #lossy #jpeg #text #dct #fft
-
Your Text Needs More JPEG - We’ve all been victims of bad memes on the Internet, but they’re not all just bad ... - https://hackaday.com/2024/03/19/your-text-needs-more-jpeg/ #discretecosinetransform #fouriertransform #softwarehacks #compression #javascript #lossifizer #lossy #jpeg #text #dct #fft
-
Your Text Needs More JPEG - We’ve all been victims of bad memes on the Internet, but they’re not all just bad ... - https://hackaday.com/2024/03/19/your-text-needs-more-jpeg/ #discretecosinetransform #fouriertransform #softwarehacks #compression #javascript #lossifizer #lossy #jpeg #text #dct #fft
-
I have a new favourite sticker. Whoever thought of making this, is a comedic genius.
#svg #sticker #lossy #compression -
I have a new favourite sticker. Whoever thought of making this, is a comedic genius.
#svg #sticker #lossy #compression -
I have a new favourite sticker. Whoever thought of making this, is a comedic genius.
#svg #sticker #lossy #compression -
I have a new favourite sticker. Whoever thought of making this, is a comedic genius.
#svg #sticker #lossy #compression -
Marina Galchenkova: Data reduction in serial #crystallography [#MX]
#Lossy #compression using non-uniform encoding: using 8 bit floating numbers increases compression ratios, while reconstruction / science still works – down to three significant bits.
-
Marina Galchenkova: Data reduction in serial #crystallography [#MX]
#Lossy #compression using non-uniform encoding: using 8 bit floating numbers increases compression ratios, while reconstruction / science still works – down to three significant bits.
-
Marina Galchenkova: Data reduction in serial #crystallography [#MX]
#Lossy #compression using non-uniform encoding: using 8 bit floating numbers increases compression ratios, while reconstruction / science still works – down to three significant bits.
