Re: [FFmpeg-user] Requesting colormatrix advice
Does anyone else have any idea what Paul's even talking about, at this point? P On Tuesday, 7 November 2023 at 18:41:55 GMT, Paul B Mahol wrote: On Tue, Nov 7, 2023 at 6:58 PM Carl Zwanzig wrote: > On 11/7/2023 10:05 AM, Paul B Mahol wrote: > > Documentation is there and its 99.99% correct and maintained, but as > usual > > you misinterpret it almost always via your logic fallacies. > > Tell us what's wrong. > > I never assume anything without sufficient facts and data I can confirm again and again using proven solutions and procedures. > z! > ___ > ffmpeg-user mailing list > ffmpeg-user@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-user > > To unsubscribe, visit link above, or email > ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". > ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Seeking Efficient Video Redaction Using FFMPEG with Limited Memory Usage
Paul B Mahol (12023-11-07): > Do you have actual proof for such claims? Libavfilter is at least linear in the number of filters for each single step of the filtering. That means a filter graph with 2000 filters will take at least 200 times as much power than a graph with 10 filters. Really a lot more, because larger graphs tend to be deeper too. This is not unavoidable, this part would be rather easy to optimize. -- Nicolas George signature.asc Description: PGP signature ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
Michael Koch (12023-11-07): > That's a good summary of the current state. I did also try to understand > this stuff some time ago, without success. Except for this: > > Interactions between all of this are not documented anywhere. This is completely untrue, and Phil has been familiar with FFmpeg for enough time that we can consider it intentional: it is a lie. Video manipulation is a difficult topic, there are lots of ways of doing similar things that will yield subtly different results. The ability to chose the right one comes with experience and effort. Unfortunately, both Phil and Mark have repeatedly shown they had no intention of making any effort themselves, their policy is to expect somebody to do the job for them gratis and insult the project when that does not happen. Regards, -- Nicolas George signature.asc Description: PGP signature ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] question about opencl use in decoders/encoders
ср, 8 нояб. 2023 г., 00:36 Dennis Mungai : > On Tue, 7 Nov 2023 at 00:55, Andrew Randrianasulu > > wrote: > > > FFmpeg today support some opencl filters (but I was unable to use them > > freely in Cinelerra-gg probably because they require not just > > initialization of hw device but also manual up and download of data ...) > > > > but I wonder if same opencl infra can be used in decoders and encoders > too? > > > > > > > https://cml.news/g/cml-raw-log-hdr/topic/braw_codec_analysis/25749037?p=,,,20,0,0,0::recentpostdate%2Fsticky,,,20,2,0,25749037 > > > > apparently those big player codecs use it, so it should be possible > > technically, just at some dev time cost ... > > > > there is also some prores gpu accelerated decoder (closed source, win) > > > > https://mogurenko.com/2020/11/15/cuda-optimized-prores-decoder/ > > > > and of course old standby mjpeg: > > > > https://github.com/nctu-pp/mjpeg-encoder > > > > (opencl part does not seems to work for me, but may be it basically > > artefact of running in 32bit x86 mode , I filled mesa bug about it > > > > > Hello Andrew, > > I hope this email finds you well. > In FFmpeg, OpenCL is primarily utilized in specific filters, a list of them > which you can verify in your build(s) with OpenCL enabled via : > > ffmpeg -filters | grep opencl > > On the encoding side, to the best of my knowledge, the libx264 encoder > wrapper implements lookahead (LA) bits that can be offloaded to an OpenCL > capable GPU. > yeah, something to test! thanks for reminder. Some users have reported potential performance gains in production, even > upstream where there's still consensus on whether this feature should be > deprecated. > See this merge request's comments on the same: > https://code.videolan.org/videolan/x264/-/merge_requests/66 > > Separate from FFmpeg: > > 1. There are also implementations of VP9 encoders utilizing OpenCL-based > offloading, such as the VP9 encoder by Ittiam systems: > > https://www.ittiam.com/resource/ittiam-enables-opencl-based-vp9-encode-for-mobile-and-consumer-applications/ > , > whose code can be viewed here https://github.com/ittiamvpx/libvpx nice find, it even builds like this on my end LDFLAGS="-lOpenCL" setarch i686 ./configure --enable-opencl after I fixed little 32bit configure bug: -check_cpp < 2. A VP8 OpenCL-based encoder: https://github.com/Aazmp/vp8oclenc > 3. And other derivatives of (1) above: > https://github.com/awatry/libvpx.opencl > > I had also posted on this mailing list earlier regarding OpenCL somewhere > in ~2018, may require verification for correctness: > https://ffmpeg.org/pipermail/ffmpeg-user/2018-April/039610.html > > Hope that helps :-) > surely! ___ > ffmpeg-user mailing list > ffmpeg-user@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-user > > To unsubscribe, visit link above, or email > ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". > ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] question about opencl use in decoders/encoders
On Tue, 7 Nov 2023 at 00:55, Andrew Randrianasulu wrote: > FFmpeg today support some opencl filters (but I was unable to use them > freely in Cinelerra-gg probably because they require not just > initialization of hw device but also manual up and download of data ...) > > but I wonder if same opencl infra can be used in decoders and encoders too? > > > https://cml.news/g/cml-raw-log-hdr/topic/braw_codec_analysis/25749037?p=,,,20,0,0,0::recentpostdate%2Fsticky,,,20,2,0,25749037 > > apparently those big player codecs use it, so it should be possible > technically, just at some dev time cost ... > > there is also some prores gpu accelerated decoder (closed source, win) > > https://mogurenko.com/2020/11/15/cuda-optimized-prores-decoder/ > > and of course old standby mjpeg: > > https://github.com/nctu-pp/mjpeg-encoder > > (opencl part does not seems to work for me, but may be it basically > artefact of running in 32bit x86 mode , I filled mesa bug about it > Hello Andrew, I hope this email finds you well. In FFmpeg, OpenCL is primarily utilized in specific filters, a list of them which you can verify in your build(s) with OpenCL enabled via : ffmpeg -filters | grep opencl On the encoding side, to the best of my knowledge, the libx264 encoder wrapper implements lookahead (LA) bits that can be offloaded to an OpenCL capable GPU. Some users have reported potential performance gains in production, even upstream where there's still consensus on whether this feature should be deprecated. See this merge request's comments on the same: https://code.videolan.org/videolan/x264/-/merge_requests/66 Separate from FFmpeg: 1. There are also implementations of VP9 encoders utilizing OpenCL-based offloading, such as the VP9 encoder by Ittiam systems: https://www.ittiam.com/resource/ittiam-enables-opencl-based-vp9-encode-for-mobile-and-consumer-applications/, whose code can be viewed here https://github.com/ittiamvpx/libvpx 2. A VP8 OpenCL-based encoder: https://github.com/Aazmp/vp8oclenc 3. And other derivatives of (1) above: https://github.com/awatry/libvpx.opencl I had also posted on this mailing list earlier regarding OpenCL somewhere in ~2018, may require verification for correctness: https://ffmpeg.org/pipermail/ffmpeg-user/2018-April/039610.html Hope that helps :-) ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On Tue, Nov 7, 2023 at 8:01 PM Carl Zwanzig wrote: > On 11/7/2023 10:10 AM, Paul B Mahol wrote: > > I never assume anything without sufficient facts and data I can confirm > > again and again using proven solutions and procedures. > > And yet, you seem to know enough to tell people that they're wrong, and do > so without telling them exactly what's wrong and how to correct it. Sounds > like that's being "More extremely lazy". > I answered all relevant sane questions. If something is still unclear ask again. > > "but as usual you misinterpret it almost always via your logic fallacies." > Missing was an explanation of what those supposed logic fallacies are and > how they're false. > > "I can answer it, but I guess OP will listen to your usual misinformation." > How do you know they will? Is it actually "misinformation" (still not > answered)? Why didn't you answer it correctly in the first place? > > > Oh, and don't bother again saying that I (or Mark, or Phil, etc) should be > banned. > You are just adding not very useful noise. > > z! > ___ > ffmpeg-user mailing list > ffmpeg-user@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-user > > To unsubscribe, visit link above, or email > ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". > ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On 11/7/2023 10:10 AM, Paul B Mahol wrote: I never assume anything without sufficient facts and data I can confirm again and again using proven solutions and procedures. And yet, you seem to know enough to tell people that they're wrong, and do so without telling them exactly what's wrong and how to correct it. Sounds like that's being "More extremely lazy". "but as usual you misinterpret it almost always via your logic fallacies." Missing was an explanation of what those supposed logic fallacies are and how they're false. "I can answer it, but I guess OP will listen to your usual misinformation." How do you know they will? Is it actually "misinformation" (still not answered)? Why didn't you answer it correctly in the first place? Oh, and don't bother again saying that I (or Mark, or Phil, etc) should be banned. z! ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Seeking Efficient Video Redaction Using FFMPEG with Limited Memory Usage
вт, 7 нояб. 2023 г., 20:57 Carl Zwanzig : > On 11/7/2023 10:00 AM, Paul B Mahol wrote: > > I can answer it, > > Then DO SO! There's no reason to hold back. > well, I am not sure if anything can pull 2000 filters ON TOP of each other (src-> f1 -> f2 -> f3 ... -> f2000 ->end) but applying dynamic selection of some (software, single input/output, same kind) filters pissible over timeline in CinelerraGG https://www.cinelerra-gg.org/ Paul also did make another GUI for complex filtergraphs (second link) https://github.com/RomanSichkaruk/FFmpegGUI https://github.com/richardpl/lavfi-preview > > but I guess OP will listen to your usual misinformation. > > What is misinformation? Tell us. If you can't or won't, then maybe what I > said is actually correct. > > z! > ___ > ffmpeg-user mailing list > ffmpeg-user@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-user > > To unsubscribe, visit link above, or email > ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". > ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On Tue, Nov 7, 2023 at 6:58 PM Carl Zwanzig wrote: > On 11/7/2023 10:05 AM, Paul B Mahol wrote: > > Documentation is there and its 99.99% correct and maintained, but as > usual > > you misinterpret it almost always via your logic fallacies. > > Tell us what's wrong. > > I never assume anything without sufficient facts and data I can confirm again and again using proven solutions and procedures. > z! > ___ > ffmpeg-user mailing list > ffmpeg-user@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-user > > To unsubscribe, visit link above, or email > ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". > ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On Tue, Nov 7, 2023 at 6:57 PM wrote: > On 11/7/23 12:42, Michael Koch wrote: > > ... I did also try to understand this stuff some time ago, without > success. > > > > Michael > > Well, you're just stupid, Michael. Phil, too. Me, too. > More extremely lazy, for not providing small sample(s). > > Hahahahaha...BoH > > ___ > ffmpeg-user mailing list > ffmpeg-user@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-user > > To unsubscribe, visit link above, or email > ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". > ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On 11/7/2023 10:05 AM, Paul B Mahol wrote: Documentation is there and its 99.99% correct and maintained, but as usual you misinterpret it almost always via your logic fallacies. Tell us what's wrong. z! ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On 11/7/23 12:42, Michael Koch wrote: ... I did also try to understand this stuff some time ago, without success. Michael Well, you're just stupid, Michael. Phil, too. Me, too. Hahahahaha...BoH ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Seeking Efficient Video Redaction Using FFMPEG with Limited Memory Usage
On 11/7/2023 10:00 AM, Paul B Mahol wrote: I can answer it, Then DO SO! There's no reason to hold back. but I guess OP will listen to your usual misinformation. What is misinformation? Tell us. If you can't or won't, then maybe what I said is actually correct. z! ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On Tue, Nov 7, 2023 at 6:53 PM Phil Rhodes via ffmpeg-user < ffmpeg-user@ffmpeg.org> wrote: > On Tuesday, 7 November 2023 at 17:28:58 GMT, Paul B Mahol < > one...@gmail.com> wrote: > > > You are deeply confused, and make assumptions from no knowledge and from > no> actual facts at all. > Oh, okay - are you telling us the options I listed aren't part of ffmpeg? They're all listed in the documentation, on pages such as > https://trac.ffmpeg.org/wiki/colorspace and > https://ffmpeg.org/ffmpeg-filters.html > Perhaps the documentation is incorrect? > Documentation is there and its 99.99% correct and maintained, but as usual you misinterpret it almost always via your logic fallacies. > - P > ___ > ffmpeg-user mailing list > ffmpeg-user@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-user > > To unsubscribe, visit link above, or email > ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". > ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Seeking Efficient Video Redaction Using FFMPEG with Limited Memory Usage
On Mon, Nov 6, 2023 at 9:13 AM sreekar manchukonda wrote: > > Hello, > > I'm seeking help on how to utilize ffmpeg to apply an extensive number of > filters—specifically over 2000, each filter including a crop, boxblur, and > overlay—to a video over an hour long. The intention is to redact certain > sections of the input at various times without consuming excessive memory. > > Could you advise if there's a method to manage thread pooling to limit the > number of concurrent threads to prevent memory overflow? > > Also, I'm looking for a strategy to process these threads in a sequence, > incorporating the output directly into the final video, while periodically > clearing the memory utilized by intermediate filter operations. Presently, > memory is only freed after the complete processing concludes. > > I would greatly appreciate an efficient approach to this. > > Thank you. > ___ > ffmpeg-user mailing list > ffmpeg-user@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-user > > To unsubscribe, visit link above, or email > ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". Since you seem to have different operations happening on different parts of the timeline (based on assumption that running 2000 filters on the same frame is knda excessive), you can run different processes to operate on segments maybe 5-10 minutes long outputting to an intermediate mostly lossless format. You can merge them back together to a 1 hour long output and transcode to whatever final format just once. -Srikanth ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On Tuesday, 7 November 2023 at 17:28:58 GMT, Paul B Mahol wrote: > You are deeply confused, and make assumptions from no knowledge and from no> > actual facts at all. Oh, okay - are you telling us the options I listed aren't part of ffmpeg? They're all listed in the documentation, on pages such as https://trac.ffmpeg.org/wiki/colorspace and https://ffmpeg.org/ffmpeg-filters.html Perhaps the documentation is incorrect? - P ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Seeking Efficient Video Redaction Using FFMPEG with Limited Memory Usage
On Tue, Nov 7, 2023 at 6:46 PM Carl Zwanzig wrote: > On 11/7/2023 9:40 AM, Paul B Mahol wrote: > > Do you have actual proof for such claims? > > I do not need to "prove" my opinion. You're free to agree or not, and if > you > feel it's not correct or appropriate to provide other information. (I do > know from my own experience that what I think the OP was taking about is > fairly easy in OpenShot. You can accept that or not.) > > Or, _you_ could actually try to help the OP and answer their questions. > You > have not done so, and seldom do at all. > I can answer it, but I guess OP will listen to your usual misinformation. > > z! > ___ > ffmpeg-user mailing list > ffmpeg-user@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-user > > To unsubscribe, visit link above, or email > ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". > ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Seeking Efficient Video Redaction Using FFMPEG with Limited Memory Usage
On 11/7/2023 9:40 AM, Paul B Mahol wrote: Do you have actual proof for such claims? I do not need to "prove" my opinion. You're free to agree or not, and if you feel it's not correct or appropriate to provide other information. (I do know from my own experience that what I think the OP was taking about is fairly easy in OpenShot. You can accept that or not.) Or, _you_ could actually try to help the OP and answer their questions. You have not done so, and seldom do at all. z! ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
Am 07.11.2023 um 18:23 schrieb Phil Rhodes via ffmpeg-user: So let me get this straight. There's a command line option called colorspace. There's a filter called colorspace which has options including "space," "trc" (meaning "transfer characteristics") and "primaries," as well as several other options which do not actually have anything to do with colourspace. There's a filter called setparams which has options including “color_primaries,” “color_trc” (meaning “colour transform” this time) and “colorspace.” This filter is known to be buggy. Several other filters (including scale and zscale) have options to set input and output colour spaces, which might complicate things further. Interactions between all of this are not documented anywhere. Has it occurred to anyone that this is getting a bit out of hand, and regardless the knowledge level of the user, it creates a situation where things are likely to be the wrong colour? That's a good summary of the current state. I did also try to understand this stuff some time ago, without success. Michael ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Seeking Efficient Video Redaction Using FFMPEG with Limited Memory Usage
On Tue, Nov 7, 2023 at 6:31 PM Carl Zwanzig wrote: > On 11/6/2023 9:13 AM, sreekar manchukonda wrote: > > > I'm seeking help on how to utilize ffmpeg to apply an extensive number of > > filters—specifically over 2000, each filter including a crop, boxblur, > and > > overlay—to a video over an hour long. > > This is a single video, being processed once, with 2000 filters being > applied? That in itself sounds like a recipe for disaster, and that a more > purpose-built editing system would do a better job (e.g. openshot or > shotcut, both are open-source). You may be able to create an EDL (Edit > Decision List) from your filters to drive the editor software. > Do you have actual proof for such claims? > > If you have storage space, you could also use uncompressed intermediate > files and successively run smaller groups of filters on them. > > Later, > > z! > ___ > ffmpeg-user mailing list > ffmpeg-user@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-user > > To unsubscribe, visit link above, or email > ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". > ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Seeking Efficient Video Redaction Using FFMPEG with Limited Memory Usage
On 11/6/2023 9:13 AM, sreekar manchukonda wrote: I'm seeking help on how to utilize ffmpeg to apply an extensive number of filters—specifically over 2000, each filter including a crop, boxblur, and overlay—to a video over an hour long. This is a single video, being processed once, with 2000 filters being applied? That in itself sounds like a recipe for disaster, and that a more purpose-built editing system would do a better job (e.g. openshot or shotcut, both are open-source). You may be able to create an EDL (Edit Decision List) from your filters to drive the editor software. If you have storage space, you could also use uncompressed intermediate files and successively run smaller groups of filters on them. Later, z! ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On Tue, Nov 7, 2023 at 6:24 PM Phil Rhodes via ffmpeg-user < ffmpeg-user@ffmpeg.org> wrote: > So let me get this straight. > There's a command line option called colorspace. > There's a filter called colorspace which has options including "space," > "trc" (meaning "transfer characteristics") and "primaries," as well as > several other options which do not actually have anything to do with > colourspace. > There's a filter called setparams which has options > including “color_primaries,” “color_trc” (meaning “colour transform” this > time) and “colorspace.” This filter is known to be buggy. > Several other filters (including scale and zscale) have options to set > input and output colour spaces, which might complicate things further. > Interactions between all of this are not documented anywhere. > Has it occurred to anyone that this is getting a bit out of hand, and > regardless the knowledge level of the user, it creates a situation where > things are likely to be the wrong colour? > You are deeply confused, and make assumptions from no knowledge and from no actual facts at all. setparams filter only changes metadata fields which is obvious from its name, it does not change actual pixels at all and is not buggy. > - Phil > ___ > ffmpeg-user mailing list > ffmpeg-user@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-user > > To unsubscribe, visit link above, or email > ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". > ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
So let me get this straight. There's a command line option called colorspace. There's a filter called colorspace which has options including "space," "trc" (meaning "transfer characteristics") and "primaries," as well as several other options which do not actually have anything to do with colourspace. There's a filter called setparams which has options including “color_primaries,” “color_trc” (meaning “colour transform” this time) and “colorspace.” This filter is known to be buggy. Several other filters (including scale and zscale) have options to set input and output colour spaces, which might complicate things further. Interactions between all of this are not documented anywhere. Has it occurred to anyone that this is getting a bit out of hand, and regardless the knowledge level of the user, it creates a situation where things are likely to be the wrong colour? - Phil ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On Tue, Nov 7, 2023 at 5:03 PM wrote: > On 11/7/23 10:27, Paul B Mahol wrote: > > On Tue, Nov 7, 2023 at 3:09 PM wrote: > > > >> On 11/7/23 08:39, Paul B Mahol wrote: > >>> On Tue, Nov 7, 2023 at 2:22 PM wrote: > >>> > On 11/7/23 06:55, Phil Rhodes via ffmpeg-user wrote: > > On Tuesday, 7 November 2023 at 11:48:24 GMT, < > markfilipak.i...@gmail.com> wrote: > > > >> I think it sets the color tags on the _input_ > > I'm not so sure ... > > I tried it. The saturation was still there. > > What exactly you tried? > >> > >> set ARGS=-map 0:v -map 0:a:1 -map 0:a:0 -vf > >> > setparams=range=tv:color_primaries=smpte170m:color_trc=smpte170m:colorspace=smpte170m,separatefields,setdar=16/9 > >> > >> -c:v libx265 -x265-params crf=18 -c:a copy > >> ffmpeg -ss 10:37.846 -to 11:09.335 -i "%SOURCE%" %ARGS% "%TARGET%" > >> > >> It seems to me that if 'setparams' is marking tags in the SOURCE instead > >> of the encoder, it would > >> not be in '-vf' and would not require transcoding. It would be a > '-bsf:v'. > >> So, I don't know what > >> 'setparams' does. Apparently nothing. > >> > > > > That is a bug. > > What is a bug, Paul? Is it a bug that '-vf > setparams==smpte170m' doesn't affect the > encoder? Or, on the other hand, is it a bug that '-vf > setparams==smpte170m' doesn't > affect the decoded stream as though it was in the source's colormatrix? As > Phil pointed out, the > documentation is ambiguous regarding whether it applies to the source or > to the encoding, so the bug > is also ambiguous, eh? > > > Workaround is to use: -colorspace -color_range etc output options > directly. > > As I pointed out, libx265 can't do anything with colormatrix values of > 'unknown'. In the face of > 'unknown', it can only encode to bt709 without changing the chroma because > it can't convert from > 'unknown', eh? In other words, I need a way of tricking libx265 into > thinking that the source is > 'smpte170m' so it will convert it to bt709. Is '-vf > setparams==smpte170m' supposed > to do that? Who knows? For sure, -colorspace -color_range etc. don't do > that. They affect only the > encoder. They are not workarounds. libx265 can't convert 'unknown' to > anything. It can only follow > the rules of Blu-ray. So I can't do an is-this-source-actually-smpte170m > test. Do you get it my friend? > > You are deeply confused. To tag input or output you need to use "-colorspace bt601" and similar as input or output option, here input option means option goes before input and output option means it goes after encoder selection and before output file name. setparams filter is only currently useful if you want to convert/transcode it from bt601 to bt709 in which case you than tag it with that filter. If you do not want to transcode than there are bitstream filters for h264/hevc/prores/etc to set some metadata too including colour primaries. colormatrix filter is not recommended for usage at all. > Regards -- Mark. > ___ > ffmpeg-user mailing list > ffmpeg-user@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-user > > To unsubscribe, visit link above, or email > ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". > ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On 11/7/23 10:27, Paul B Mahol wrote: On Tue, Nov 7, 2023 at 3:09 PM wrote: On 11/7/23 08:39, Paul B Mahol wrote: On Tue, Nov 7, 2023 at 2:22 PM wrote: On 11/7/23 06:55, Phil Rhodes via ffmpeg-user wrote: On Tuesday, 7 November 2023 at 11:48:24 GMT, < markfilipak.i...@gmail.com> wrote: I think it sets the color tags on the _input_ I'm not so sure ... I tried it. The saturation was still there. What exactly you tried? set ARGS=-map 0:v -map 0:a:1 -map 0:a:0 -vf setparams=range=tv:color_primaries=smpte170m:color_trc=smpte170m:colorspace=smpte170m,separatefields,setdar=16/9 -c:v libx265 -x265-params crf=18 -c:a copy ffmpeg -ss 10:37.846 -to 11:09.335 -i "%SOURCE%" %ARGS% "%TARGET%" It seems to me that if 'setparams' is marking tags in the SOURCE instead of the encoder, it would not be in '-vf' and would not require transcoding. It would be a '-bsf:v'. So, I don't know what 'setparams' does. Apparently nothing. That is a bug. What is a bug, Paul? Is it a bug that '-vf setparams==smpte170m' doesn't affect the encoder? Or, on the other hand, is it a bug that '-vf setparams==smpte170m' doesn't affect the decoded stream as though it was in the source's colormatrix? As Phil pointed out, the documentation is ambiguous regarding whether it applies to the source or to the encoding, so the bug is also ambiguous, eh? Workaround is to use: -colorspace -color_range etc output options directly. As I pointed out, libx265 can't do anything with colormatrix values of 'unknown'. In the face of 'unknown', it can only encode to bt709 without changing the chroma because it can't convert from 'unknown', eh? In other words, I need a way of tricking libx265 into thinking that the source is 'smpte170m' so it will convert it to bt709. Is '-vf setparams==smpte170m' supposed to do that? Who knows? For sure, -colorspace -color_range etc. don't do that. They affect only the encoder. They are not workarounds. libx265 can't convert 'unknown' to anything. It can only follow the rules of Blu-ray. So I can't do an is-this-source-actually-smpte170m test. Do you get it my friend? Regards -- Mark. ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On Tue, Nov 7, 2023 at 3:09 PM wrote: > On 11/7/23 08:39, Paul B Mahol wrote: > > On Tue, Nov 7, 2023 at 2:22 PM wrote: > > > >> On 11/7/23 06:55, Phil Rhodes via ffmpeg-user wrote: > >>>On Tuesday, 7 November 2023 at 11:48:24 GMT, < > >> markfilipak.i...@gmail.com> wrote: > >>> > I think it sets the color tags on the _input_ > >>> I'm not so sure ... > >> > >> I tried it. The saturation was still there. > >> > >> What exactly you tried? > > set ARGS=-map 0:v -map 0:a:1 -map 0:a:0 -vf > setparams=range=tv:color_primaries=smpte170m:color_trc=smpte170m:colorspace=smpte170m,separatefields,setdar=16/9 > > -c:v libx265 -x265-params crf=18 -c:a copy > ffmpeg -ss 10:37.846 -to 11:09.335 -i "%SOURCE%" %ARGS% "%TARGET%" > > It seems to me that if 'setparams' is marking tags in the SOURCE instead > of the encoder, it would > not be in '-vf' and would not require transcoding. It would be a '-bsf:v'. > So, I don't know what > 'setparams' does. Apparently nothing. > That is a bug. Workaround is to use: -colorspace -color_range etc output options directly. > > ___ > ffmpeg-user mailing list > ffmpeg-user@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-user > > To unsubscribe, visit link above, or email > ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". > ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On 11/7/23 08:39, Paul B Mahol wrote: On Tue, Nov 7, 2023 at 2:22 PM wrote: On 11/7/23 06:55, Phil Rhodes via ffmpeg-user wrote: On Tuesday, 7 November 2023 at 11:48:24 GMT, < markfilipak.i...@gmail.com> wrote: I think it sets the color tags on the _input_ I'm not so sure ... I tried it. The saturation was still there. What exactly you tried? set ARGS=-map 0:v -map 0:a:1 -map 0:a:0 -vf setparams=range=tv:color_primaries=smpte170m:color_trc=smpte170m:colorspace=smpte170m,separatefields,setdar=16/9 -c:v libx265 -x265-params crf=18 -c:a copy ffmpeg -ss 10:37.846 -to 11:09.335 -i "%SOURCE%" %ARGS% "%TARGET%" It seems to me that if 'setparams' is marking tags in the SOURCE instead of the encoder, it would not be in '-vf' and would not require transcoding. It would be a '-bsf:v'. So, I don't know what 'setparams' does. Apparently nothing. ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On Tue, Nov 7, 2023 at 2:22 PM wrote: > On 11/7/23 06:55, Phil Rhodes via ffmpeg-user wrote: > > On Tuesday, 7 November 2023 at 11:48:24 GMT, < > markfilipak.i...@gmail.com> wrote: > > > >> I think it sets the color tags on the _input_ > > I'm not so sure ... > > I tried it. The saturation was still there. > > What exactly you tried? > Giving up for lack of information. > > Thanks folks -- Mark. > > ___ > ffmpeg-user mailing list > ffmpeg-user@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-user > > To unsubscribe, visit link above, or email > ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". > ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On 11/7/23 06:55, Phil Rhodes via ffmpeg-user wrote: On Tuesday, 7 November 2023 at 11:48:24 GMT, wrote: I think it sets the color tags on the _input_ I'm not so sure ... I tried it. The saturation was still there. Giving up for lack of information. Thanks folks -- Mark. ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
I don't think you need to do this for a second - what's on the disc will be
709. That's how blu-rays work, unless they're the oddball Sony 4K thing.
The whole purpose of colour grading is to put whatever comes off the film
scanner into 709.
P
On Tuesday, 7 November 2023 at 12:04:34 GMT, Rob Hallam
wrote:
On Mon, 6 Nov 2023 at 03:42, wrote:
> The Question:
> Do you have any advice regarding colormatrix?
> Thanks for any advice for relieving a headache
> --Mark.
If you haven't already, it might be worth tackling this from another
direction. Look at the credits for the remaster, figure out who was
involved* in doing the actual work, and shoot them an email or
message?
"Hi, I have a copy of the {WHICHEVER} bluray, can I pick your brain
about the colorspace / film digitizing† process you used? (if it
wasn't you, who would you recommend I ask)"
Their answers might make it easier to figure out which dials and
levers to adjust, colorspace or otherwise, if your 601-to-709
experiment doesn't produce acceptable results. They might be
interesting from a process perspective regardless.
Cheers,
Rob
*†: I can think of a case of a 90s Paramount series that was shot on
film, but which also used a modest amount of CGI to augment the
remastering of VFX- whatever CGI they used presumably was done to
match the film in color-grading terms. If this case is the same as
your case, perhaps Wade Felker or Dylan Hucklesby might be good places
to start. Good luck!
___
ffmpeg-user mailing list
ffmpeg-user@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-user
To unsubscribe, visit link above, or email
ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
___
ffmpeg-user mailing list
ffmpeg-user@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-user
To unsubscribe, visit link above, or email
ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On Mon, 6 Nov 2023 at 03:42, wrote:
> The Question:
> Do you have any advice regarding colormatrix?
> Thanks for any advice for relieving a headache
> --Mark.
If you haven't already, it might be worth tackling this from another
direction. Look at the credits for the remaster, figure out who was
involved* in doing the actual work, and shoot them an email or
message?
"Hi, I have a copy of the {WHICHEVER} bluray, can I pick your brain
about the colorspace / film digitizing† process you used? (if it
wasn't you, who would you recommend I ask)"
Their answers might make it easier to figure out which dials and
levers to adjust, colorspace or otherwise, if your 601-to-709
experiment doesn't produce acceptable results. They might be
interesting from a process perspective regardless.
Cheers,
Rob
*†: I can think of a case of a 90s Paramount series that was shot on
film, but which also used a modest amount of CGI to augment the
remastering of VFX- whatever CGI they used presumably was done to
match the film in color-grading terms. If this case is the same as
your case, perhaps Wade Felker or Dylan Hucklesby might be good places
to start. Good luck!
___
ffmpeg-user mailing list
ffmpeg-user@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-user
To unsubscribe, visit link above, or email
ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On Tuesday, 7 November 2023 at 11:48:24 GMT, wrote: > I think it sets the color tags on the _input_ I'm not so sure, it specifically says "marks interlace and color range for the output frames." It then goes on to imply the filter does a bit more than set just interlacing and colour range flags, since it seems to support at least three different types of colour primary specification, but we're forced to guess. P ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On 11/7/23 06:29, Phil Rhodes via ffmpeg-user wrote: Well, yeah, but we've been discussing x265, too, so it seemed worth asking. I presume you're talking about this setparams filter: https://ffmpeg.org/ffmpeg-filters.html#setparams-1 There aren't any examples of usage there, but it seems to do nothing but set the flags on the output data. I think it sets the color tags on the _input_. That gives a starting point for ffmpeg to convert from -- convert from BT601 to BT709. That has to be better than convert from unknown to BT709, eh? I'll give it a try on my saturated video. Thank you, Paul. -- Mark. ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
Well, yeah, but we've been discussing x265, too, so it seemed worth asking. I presume you're talking about this setparams filter: https://ffmpeg.org/ffmpeg-filters.html#setparams-1 There aren't any examples of usage there, but it seems to do nothing but set the flags on the output data. I'm not sure how that would work out with various output file formats and codecs, which may have various different ways of including that information. I'm not clear what the function of "color transfer," "color primaries" and "colorspace" is in that case. P On Tuesday, 7 November 2023 at 10:52:53 GMT, Paul B Mahol wrote: On Tue, Nov 7, 2023 at 11:40 AM Phil Rhodes via ffmpeg-user < ffmpeg-user@ffmpeg.org> wrote: > > > On Tuesday, 7 November 2023 at 10:30:33 GMT, Paul B Mahol < > one...@gmail.com> wrote: > > Use setparams filter to properly tag inputs colorspace related metadata. > > Y'know, I've been dealing with this stuff for twenty-plus years, and I'm > not very clear on exactly what you're trying to suggest Mark does here. > What software are you referring to? Is it documented anywhere? Are there > examples? > This is mailing list about FFmpeg software. > P > ___ > ffmpeg-user mailing list > ffmpeg-user@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-user > > To unsubscribe, visit link above, or email > ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". > ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
[FFmpeg-user] Converting 25fps 'PAL'-fast to 24fps cinema
The Windows CLI script below should be easy enough for Linux users to read. It converts 'PAL'-fast (or PAL speedup if you like) back to cinema. It is tested and known good. set ITSSCALE=-itsscale 1.0416667 set SETTS=-bsf:v setts=time_base=1/9:pts=N*9/24:dts=N*9/24 set ATEMPO=-af atempo=24/25 set CODE=-c:v copy -c:a ac3 -c:s copy ffmpeg %ITSSCALE% -i %SOURCE% -map 0 %SETTS% %ATEMPO% %CODE% -dn -r 24 %TARGET% exit Notes: 1.0416667 = 25/24. FFmpeg reports 25fps TARGET when it's actually 24fps. FFprobe & MPV both report 24fps. Missing %ITSSCALE% causes subtitles to be early. Missing %SETTS% provokes bogus 'Timestamps are unset' + 'pts has no value' warnings. 'N/TS/24' (instead of 'N*9/24') provokes 'Non-monotonous DTS in output stream'. Missing %ATEMPO% causes MPV's clock to speed up. A 20:08 video actually ends in 19:20! Missing '-r 24' causes the TARGET to be 25fps despite the PTS changes. The notes contain a lot of interesting stuff that could be/should be discussed. For example: Why would fast audio (no atempo) cause MPV's clock to run hot? Why/how is there no pitch change? Why is ffmpeg reporting 24fps video as 25fps? More... ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On Tue, Nov 7, 2023 at 11:40 AM Phil Rhodes via ffmpeg-user < ffmpeg-user@ffmpeg.org> wrote: > > > On Tuesday, 7 November 2023 at 10:30:33 GMT, Paul B Mahol < > one...@gmail.com> wrote: > > Use setparams filter to properly tag inputs colorspace related metadata. > > Y'know, I've been dealing with this stuff for twenty-plus years, and I'm > not very clear on exactly what you're trying to suggest Mark does here. > What software are you referring to? Is it documented anywhere? Are there > examples? > This is mailing list about FFmpeg software. > P > ___ > ffmpeg-user mailing list > ffmpeg-user@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-user > > To unsubscribe, visit link above, or email > ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". > ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On Tuesday, 7 November 2023 at 10:30:33 GMT, Paul B Mahol wrote: > Use setparams filter to properly tag inputs colorspace related metadata. Y'know, I've been dealing with this stuff for twenty-plus years, and I'm not very clear on exactly what you're trying to suggest Mark does here. What software are you referring to? Is it documented anywhere? Are there examples? P ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On Tue, Nov 7, 2023 at 11:21 AM wrote: > On 11/7/23 04:08, Phil Rhodes via ffmpeg-user wrote: > > I have no idea how colour handling in x265 works so I can't really > advise on specifically how to set it up. Possibly it's documented > somewhere. One of the problems with this (which comes up in some high end > post production software quite often) is that it's sometimes not very clear > whether we're specifying what we expect the input data to be, or what we > expect it to be converted to. > > I thought only I had that problem. :-) > > > For instance, I'm not sure what's meant by "transfer" in this context. > > Perfect example. > > >> It's 1920x1080. I 'heard' of "studio swing". What is it relative to > what x265 expects (above)? Do you know? > > I'm possibly guilty of indulging in jargon here; by "studio swing" I'm > referring to a reduced data range often from 16-235 for the luminance > channel. > > The preferred term appears to be 'limited range'. > > >> I feel a headache coming on. Let's assume ffprobe is correct and > colormatrix is undefined on the discs > > What's on the disc will be in Rec. 709. > > I don't think so. If that were true, then what's on disc and what's in the > MP4 would both play > accurately and without saturation. I'm pretty sure they're not BT709. > > > I honestly don't know what data ranges blu-rays use. > > Limited. It's silly -- more legacy analog TV hangover -- but it's limited. > The MPEG folks just can't > seem to wean themselves away from TV-on-disc. > > > It's often fairly clear if you can look at it and it looks all lifted, > that'll be why. > > "lifted"? > > > Unfortunately there are a lot of moving parts here. > > I may be naive but I don't think so. X265 has presets that appear to cover > all commercial media. How > it covers is apparently a matter of some conjecture. But really, all that > color adaptation stuff, > like to accommodate phosphors in CRTs, needs to be done BY THE TV, not by > the media. There should be > one set of colors and they should be RGB full range with unlimited gamut. > The TV makers know what > the TVs are capable of and can do dot-by-dot fix up. > > You're off the hook, Phil. Thanks for your thoughts. I appreciate them. -- > Mark. > Use setparams filter to properly tag inputs colorspace related metadata. > > ___ > ffmpeg-user mailing list > ffmpeg-user@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-user > > To unsubscribe, visit link above, or email > ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". > ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
On 11/7/23 04:08, Phil Rhodes via ffmpeg-user wrote: I have no idea how colour handling in x265 works so I can't really advise on specifically how to set it up. Possibly it's documented somewhere. One of the problems with this (which comes up in some high end post production software quite often) is that it's sometimes not very clear whether we're specifying what we expect the input data to be, or what we expect it to be converted to. I thought only I had that problem. :-) For instance, I'm not sure what's meant by "transfer" in this context. Perfect example. It's 1920x1080. I 'heard' of "studio swing". What is it relative to what x265 expects (above)? Do you know? I'm possibly guilty of indulging in jargon here; by "studio swing" I'm referring to a reduced data range often from 16-235 for the luminance channel. The preferred term appears to be 'limited range'. I feel a headache coming on. Let's assume ffprobe is correct and colormatrix is undefined on the discs What's on the disc will be in Rec. 709. I don't think so. If that were true, then what's on disc and what's in the MP4 would both play accurately and without saturation. I'm pretty sure they're not BT709. I honestly don't know what data ranges blu-rays use. Limited. It's silly -- more legacy analog TV hangover -- but it's limited. The MPEG folks just can't seem to wean themselves away from TV-on-disc. It's often fairly clear if you can look at it and it looks all lifted, that'll be why. "lifted"? Unfortunately there are a lot of moving parts here. I may be naive but I don't think so. X265 has presets that appear to cover all commercial media. How it covers is apparently a matter of some conjecture. But really, all that color adaptation stuff, like to accommodate phosphors in CRTs, needs to be done BY THE TV, not by the media. There should be one set of colors and they should be RGB full range with unlimited gamut. The TV makers know what the TVs are capable of and can do dot-by-dot fix up. You're off the hook, Phil. Thanks for your thoughts. I appreciate them. -- Mark. ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
Re: [FFmpeg-user] Requesting colormatrix advice
I have no idea how colour handling in x265 works so I can't really advise on specifically how to set it up. Possibly it's documented somewhere. One of the problems with this (which comes up in some high end post production software quite often) is that it's sometimes not very clear whether we're specifying what we expect the input data to be, or what we expect it to be converted to. For instance, I'm not sure what's meant by "transfer" in this context. > It's 1920x1080. I 'heard' of "studio swing". What is it relative to what x265 > expects (above)? Do you know? I'm possibly guilty of indulging in jargon here; by "studio swing" I'm referring to a reduced data range often from 16-235 for the luminance channel. > I feel a headache coming on. Let's assume ffprobe is correct and colormatrix >is undefined on the discs What's on the disc will be in Rec. 709. I honestly don't know what data ranges blu-rays use. It's often fairly clear if you can look at it and it looks all lifted, that'll be why. Unfortunately there are a lot of moving parts here. P ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
