About the progress of still images encoding with Daala (i.e. an alternative to JPEG and WebP):
http://people.xiph.org/~xiphmont/demo/daala/update1.shtml
http://people.xiph.org/~xiphmont/demo/daala/update1.shtml
Presentation by Timothy Terriberry about Daala at Linux.conf.au 2015:
Looks like Daala release will be sometime in 2016.
Looks like Daala release will be sometime in 2016.
Last edited:
@sidspyker : Can you please rename the thread into something like "Video and audio codecs and tools".
By the way, are you familiar with Bink audio and video? Why is anyone using it instead of let's say VP9 and Opus? I red somewhere that the Linux version of Tex Murphy - Tesla Effect is delayed because they can't make Bink2 work with Unity on Linux. My question was - What?.. Why do they use that altogether?
By the way, are you familiar with Bink audio and video? Why is anyone using it instead of let's say VP9 and Opus? I red somewhere that the Linux version of Tex Murphy - Tesla Effect is delayed because they can't make Bink2 work with Unity on Linux. My question was - What?.. Why do they use that altogether?
Last edited:
There is no good reason for using Bink other than Bink is exactly a tool that a developer can use to deliver full-motion video using the same toolset on Windows, Mac OS X, and consoles all the way back to the original Xbox. That is, when you are developing software that you intend to sell for a profit, actually a damned good reason.
I don't really see benefits of Bink when other and better decoders can work on all the same platforms as well. And in this case they even can work with Unity better. Original Xbox? You mean one before 360? I don't think anyone is concerned about targeting that.
Last edited:
Can you get VP9 today and provide a substantiated claim that it will run on multiple current gaming platforms and provide tech support on demand for it? No.
Developers who are in this to make a profit and answerable to officers and directors are constrained to those tools that meet their needs now and that they can obtain now and support now. Not whenever an organization of volunteers finishes software that takes somebody else's needs into account.
Developers have been using Bink and RAD Tools since 1999, on more than 6,000 titles, and Smacker since 1994. That's a lot of successful experience for a competitor that is not yet available and does not meet their needs to overcome.
The argument that nobody writes for original Xbox anymore is true but irrelevant. The point is rather that media prepared for Bink going back that far, for that matter going back to 1999, is still usable with modern Bink, without time, effort, and money spent on conversion that does not generate revenue.
Last edited:
I see no reason why VP9 can't be available on lot's of platforms (through ffmpeg for instance). It's a stable codec and I bet has way better technical support than Bink can ever have. If there are any bugs with it, they are fixed. You can apply those patches right away since you have the code. Try reporting bugs in Bink and wait for 2 years for them to fix it?
The point about Xbox is indeed irrelevant. It can go as far as it wants but no sane development would worry about archaic museum type platforms.
About costs - paying codec licenses is always expensive for any commercial project. And the bigger is its scale, the bigger is the toll. So it doesn't make any sense to ever use proprietary codecs when free ones especially surpass them technically. A lot of developers actually realize that and use Opus and VP8 (now VP9) just fine. Example with CDPR was already brought above. For audio they used Vorbis and CELT (Opus predecessor) in Witcher 1 and 2 respectively. Not sure what they used for video. So it's quite natural to expect them to use free codecs in TW3 as well.
The point about Xbox is indeed irrelevant. It can go as far as it wants but no sane development would worry about archaic museum type platforms.
About costs - paying codec licenses is always expensive for any commercial project. And the bigger is its scale, the bigger is the toll. So it doesn't make any sense to ever use proprietary codecs when free ones especially surpass them technically. A lot of developers actually realize that and use Opus and VP8 (now VP9) just fine. Example with CDPR was already brought above. For audio they used Vorbis and CELT (Opus predecessor) in Witcher 1 and 2 respectively. Not sure what they used for video. So it's quite natural to expect them to use free codecs in TW3 as well.
Last edited:
VP9? The decoder will kill the CPU on the consoles. Hell even PC CPUs aren't doing so good with it presently.
---------- Updated at 07:32 AM ----------
It's not like games haven't ever used 'free' codecs, I remember NFS Most Wanted(2005) used VP6 for example. Bink provides crossplatform support, relatively easy tools that give back a decent quality and most importantly, decoding doesn't destroy the CPU(VP9 does).
The only way something like that could run on weaker CPUs would be having a dedicated H/W decoder.
---------- Updated at 07:32 AM ----------
It's not like games haven't ever used 'free' codecs, I remember NFS Most Wanted(2005) used VP6 for example. Bink provides crossplatform support, relatively easy tools that give back a decent quality and most importantly, decoding doesn't destroy the CPU(VP9 does).
The only way something like that could run on weaker CPUs would be having a dedicated H/W decoder.
Didn't VP9 try to reduce algorithmic complexity? But I didn't really compare it to VP8 so not sure how much more demanding it is. May be they are more bent on stronger compression at the cost of computational resources. But it can have various parameters to adjust it probably (balancing it more to faster decoding and larger size). If VP9 is too expensive, VP8 can be more suitable for consoles.
Daala for example aims to reduce computational cost, it's one of the critical goals for the project. Hardware decoding of course helps, but I doubt any hardware has dedicated Bink decoders. On the other hand recent Nvidia and Intel chips already provide VP9 hardware decoders if I'm not mistaken. Not sure about AMD though.
Daala for example aims to reduce computational cost, it's one of the critical goals for the project. Hardware decoding of course helps, but I doubt any hardware has dedicated Bink decoders. On the other hand recent Nvidia and Intel chips already provide VP9 hardware decoders if I'm not mistaken. Not sure about AMD though.
Last edited:
Where have you heard that current CPUs provide hardware support for VP9? Qualcomm and nVidia have VP8 support, and MediaTek has a VP9 chip, but Intel?
Unfortunately, there's a lot of industry that doesn't really care whether a technology is freely licensed. The cost of licensing is negligible compared to the cost of engineering and the cost of delays. And HEVC (H.265) burns VP9 on performance and has already been adopted for 4K Blu-Ray, Windows 10, and a few others.
Unfortunately, there's a lot of industry that doesn't really care whether a technology is freely licensed. The cost of licensing is negligible compared to the cost of engineering and the cost of delays. And HEVC (H.265) burns VP9 on performance and has already been adopted for 4K Blu-Ray, Windows 10, and a few others.
I didn't really test it, since I'm not using Intel GPU, but here is what I saw:
http://techreport.com/news/27677/new-intel-igp-drivers-add-h-265-vp9-hardware-decode-support
If that's a pure driver change, then how is that hardware accelerated? And if not, it means that hardware has some features to support it and driver just enables them. No idea what they mean by "partial" support there. May be something similar to ARM Neon?
---------- Updated at 03:54 AM ----------
HEVC doesn't surpass it on performance, they improve decoders rather in parallel. HEVC is adopted more at this stage simply because of the MPEG-LA cartel reach and passive approach that Google took to defend VPx against patent attacks. It's too late to fix things from PR perspective (see that presentation about Daala about this). That's why Xiph and Mozilla are taking another approach and proactively defending their codecs.
http://techreport.com/news/27677/new-intel-igp-drivers-add-h-265-vp9-hardware-decode-support
If that's a pure driver change, then how is that hardware accelerated? And if not, it means that hardware has some features to support it and driver just enables them. No idea what they mean by "partial" support there. May be something similar to ARM Neon?
---------- Updated at 03:54 AM ----------
HEVC doesn't surpass it on performance, they improve decoders rather in parallel. HEVC is adopted more at this stage simply because of the MPEG-LA cartel reach and passive approach that Google took to defend VPx against patent attacks. It's too late to fix things from PR perspective (see that presentation about Daala about this). That's why Xiph and Mozilla are taking another approach and proactively defending their codecs.
Last edited:
The way I read that is they have instructions that are useful in HEVC, VP9, and some of OpenCL and OpenGL processing, and the driver adds support for functions that can use these instructions. Not that they have put a decoder targeted to any specific media technology in the GPU.
It may be similar to AMD NEON without being quite so purpose-built as NEON is.
And HEVC has repeatedly exceeded VP9 performance in testing conducted by multiple laboratories. Not by a small difference, either. Accusations that this is done to further the ends of a cartel lack support. VP9 is simply too little, too late.
Any testing reflects decoders which they used at the time, decoders performance isn't set in stone, so I don't expect HEVC to have any advantage here, they are quite comparable in potential (most of the ideas they use are the same) so decoders won't be very different from each other when they'll reach their full optimization. Its main advantage isn't technical at all and that one isn't fixable at this point. At least not easily.
Last edited:
They are going to race each other. It's good to separate the codec (i.e. spec) and encoders / decoders. Codec can be fixed, while the later can have better and better implementations. As I said, I don't see HEVC having any better potential than VP9. They are pretty similar in many aspects (i.e. codecs are similar) so their encoders / decoders can't have a significantly different theoretical performance ceiling as well. Daala is another story since it's conceptually a very different codec to begin with, so it can actually surpass them in performance quite a lot.
Also, there is some development on VP10 already.
Yeah it's just that with H264 it became synonymous with x264 because that was the leading encoder and easily the best so it's a force of habit 
Anyway, x265 I mean is doing pretty damn great right now. It obviously is more complex and takes more time than x264 to encode the same clips and the results aren't "definitely" better but it's getting there.
VP9 on the other hand, the encoder is slow(both official or otherwise) as hell and the quality, well I haven't seen a big difference on the current one.
Anyway, x265 I mean is doing pretty damn great right now. It obviously is more complex and takes more time than x264 to encode the same clips and the results aren't "definitely" better but it's getting there.
VP9 on the other hand, the encoder is slow(both official or otherwise) as hell and the quality, well I haven't seen a big difference on the current one.
Some info on Intel's hardware VP8 support in Broadwell:
http://www.phoronix.com/scan.php?page=news_item&px=MTYzNDA
http://www.phoronix.com/scan.php?page=news_item&px=MTc0NDc
It's mostly about VA-API, but it mentions Broadwell as supporting it (may be same partial support like above).
---------- Updated at 05:54 PM ----------
Here is another source: http://wccftech.com/intel-broadwell-significant-graphic-enhancements-haswell/
http://www.phoronix.com/scan.php?page=news_item&px=MTYzNDA
http://www.phoronix.com/scan.php?page=news_item&px=MTc0NDc
It's mostly about VA-API, but it mentions Broadwell as supporting it (may be same partial support like above).
---------- Updated at 05:54 PM ----------
Here is another source: http://wccftech.com/intel-broadwell-significant-graphic-enhancements-haswell/
So VP8 support is on track. Nothing about VP9 yet though.
A short script to convert all audio in the current directory (with given extension passed as parameter) to Opus (mostly for music, bitrate is set to 140 Kbps). Note: metadata is ignored unless the source if FLAC.
all_to_opus.sh
all_to_opus.sh
Code:
#!/bin/bash
ext="$1"
if [ -z "$ext" ]
then
echo "Error, no source extension provided!"
exit 1
fi
mkfifo wavpipe
for source in *.${ext}
do
if [[ "$ext" != "wav" ]] && [[ "$ext" != "flac" ]]
then
mpv -quiet -no-video -vo null -ao pcm:waveheader:file="wavpipe" "$source" &
else
cat "$source" > wavpipe &
fi
opusenc --bitrate 140 wavpipe "${source%.*}.opus"
done
rm wavpipe
Last edited:
A simple way to extract audio from Cyberia blobs (this ignores all markers - just splits the file by the header assuming there is nothing in between). A better way would be to write a proper parser in some normal language like C++:
extract_voc.sh
extract_voc.sh
Code:
#!/bin/bash
source="$1"
source_size=$(stat -c %s "$source")
offsets=( $(grep --only-matching --byte-offset --binary --text Creative "$source" | cut -d : -f 1) )
for ((i=0; i<${#offsets[@]}; i++))
do
left=${offsets[$i]}
if ((i == ${#offsets[@]} - 1))
then
right=$source_size
else
right=${offsets[$((i+1))]}
fi
dd skip=$left count=$(($right - $left)) if=$source of=${source%.*}_${i}.voc bs=1
done
Last edited: