Tone-mapping in transcoding HDR for playback on SDR screens??

[Abobader 2942]

7 hours ago, RanmaCanada said:

How many more chances do they get?  Multiple times in this thread they have been abusive to anyone who dares go against their uninformed opinion.  I myself have been angry at things that have happened, but I've never gone as far as this person has with their disrespect and personal attacks.  They also continue to sow misinformation and discord.

We do not show what we do, that user got warning from the mod's team already.

[Sammy 735]

Enough argumentum ad hominem..

That's a serious fallacy in my book. I'll listen (read) any argument based in facts but as soon as the personal attacks come out I tune out, no matter how valid the argument.

[cptlores 23]

On 1/27/2021 at 6:15 PM, softworkz said:

My personal view for that is: when you don't have a HDR display for the primary viewing setup, then it's stupid, to acquire HDR content and rely on your media server's tone mapping.

Are you serious? One of the main uses for Emby is to share a media library to multiple users. And the live transcoding adapting to each users equipment/bandwidth is maybe the main selling point of Emby. Your precious 4K HDR library is essentially worthless without transcoding and tone mapping, the day you want to watch something on the mobile while traveling etc.

 

[Sammy 735]

True. I watch 4k HDR on my HT with a ShieldTV without transcoding just fine but if I want to watch it in another room without more expensive equipment and transcoding is required it is pretty washed out. For my family that uses my server over the internet they have no choice but to transcode so they can never get the full glory or 4k HDR, not that they could get true 4k but at least they should get a picture that has proper tonal mapping.

[Sammy 735]

BTW, this thread as a week shy of three years old and the issue persists.

[ebr 14910]

37 minutes ago, cptlores said:

Are you serious? One of the main uses for Emby is to share a media library to multiple users. And the live transcoding adapting to each users equipment/bandwidth is maybe the main selling point of Emby. Your precious 4K HDR library is essentially worthless without transcoding and tone mapping, the day you want to watch something on the mobile while traveling etc.

 

I agree but you have to understand the limitations of the technology and adapt as well.

24 minutes ago, Sammy said:

BTW, this thread as a week shy of three years old and the issue persists.

And, as has been noted multiple times now, the technology simply hasn't been there to do this with any real chance of success for most people - until now...  Now, we are close (but it still isn't going to work on some low powered systems).

[Sammy 735]

@ebr

Thanks for the clarification.

Is this maybe related to the new hardware requirement for AV-1 laid out by Google for new Google TV (bye bye AndroidTV) Devices?

https://www.xda-developers.com/google-requires-new-android-tv-av1-video-decoding/

[CBers 6771]

9 hours ago, Sammy said:

Google TV (bye bye AndroidTV)

You don't want Google TV, even though it's coming 

The home screen is too cluttered under Google TV, whereas it's very minimalist on Android TV. 

I much prefer the home screen on my Shield than I do on the Chromecast Google TV. 

Thanks. 

 

[niallobr 9]

Just dropping in to say thanks for the work on bringing this to Emby. I have some friends watching HDR content and it needs to be transcoded and then looks washed out.

Yes it can take a while for some features to arrive in Emby but always find you guys open to improvements and solutions. You tend to focus in more useful areas than the Plex team also, in my opinion. Also appreciate there are many things being added into Emby at the moment (I was so happy to get subtitle offsets and the support for Live TV groups in the latest beta). Looking forward to this one.

[RanmaCanada 344]

On 29/01/2021 at 17:31, Sammy said:

BTW, this thread as a week shy of three years old and the issue persists.

That's because it's not exactly easy to do.  As I've mentioned already, Intel has been working on it themselves for years and will finally have a full hardware solution with their 11th series or processors.  Nvidia and AMD haven't even bothered to address it as they appear to not GAF.  The current plugins that the community has created don't always work, and everyone always bitches and complains towards said devs who created these plugins in their spare time.  

As for 4k material, it should NEVER be transcoded, whether is HDR or DolbyVision.  You keep a separate library for your 4k content, and your non 4k content.  If you already have the space for 4k content, then you obviously have the space for regular content and not having it is just you being lazy.  Having your 4k content in a separate library will also allow you to block certain users who attempt to play those files on their potato systems and thus complaining to you that their movie looks like garbage while the transcode is bringing your server to its knees because of their stupidity.  

[RDHoworth 6]

I think that you are now being rather arrogant. It is not being Lazy to want to keep only one copy of a film as there are many reasons for wanting to do this. Just because it is not your idea of best practice is not a valid reason for your response.

[RanmaCanada 344]

1 hour ago, RDHoworth said:

I think that you are now being rather arrogant. It is not being Lazy to want to keep only one copy of a film as there are many reasons for wanting to do this. Just because it is not your idea of best practice is not a valid reason for your response.

It's not being arrogant, it's being practical.  As I mentioned, this has not been feasible until just recently (just like DolbyVision, which is another story of user ignorance), and yet none of you in here seem to understand the technology behind it, nor the work that is involved.  It takes a minimum of 17k passmark score to transcode a 4k file down to 1080p.  That is before tonemapping gets involved.  Do you really think it's proper to waste that amount of power so someone can watch a movie on their phone?  In the long run, it will cost you less to buy more HD space to keep your libraries separate than it would to transcode them every single time a user with a non compliant device attempts to play them back.  Intel will have a full hardware solution with their 11th gen chips.  Software solutions will still take more power than most people have bothered to put into their Emby servers.  Even if we look at Plex, the only answer they have with 100% support is in LInux.  Every other OS is done in software, meaning no NVENC, no VCN/VCE no Quicksync.  So unless you have a balls to the wall server, or running Linux, tonemapping is not going to be feasible for you.  Which means about 90% of the people who run Emby will not even be able to use this feature once it's implemented, because let's face it, most people have their server running on potatos.  Let's also not forget that there is still no standard for tonemapping, and that the current plugins don't work 100% of the time.  So when tone mapping brings your server to its knees, and your colours look off, what are you going to do?  Oh that's right, everyone will moan and complain that the software is garbage because you users don't understand the complexities behind it all.  

 

 

[Carlo 4330]

Emby's tone mapping solution should run faster on Windows than other "solutions" if things go well.

Even with GPU support because of the IO involved with passing things back and forth across the bus from CPU to GPU there will be a limit that you don't see with normal transcoding that can take place strictly in the GPU.

The reality of the situation in real life is for most users is going to be limited tone mapping streams concurrently. Just throwing a number out as an example that may not play out in real life but to make a point, say your system can transcode/tone map 2 streams at the same time.

If you can support 2 streams and only share to family members this might be just fine, but if you share to a half dozen people or more outside your house and plan on having a large 4K library without 1080 backups things may not work well for you if your hardware can only support 2 of these streams at one time.

Not trying to rain on anyone's parade but there will be limits to what can be done because of the way hardware currently works.  This is something @RanmaCanadaunderstands and is trying to explain.  Tone Mapping isn't going to be a "fix all 4K solution" for large amounts of people short of having things like 11th Gen Intel or equiv Nvidia GPU that can do this processing on chip FULLY.

The industry is progressing but it actually is still pretty early for this type of thing.  Emby will deliver a solution and will keep updating it as technology allows but it's not wise to think this is going to allow you to have only 4K HDR media and stream to 6 of your friends at the same time like you can easily now with 1080 content.

[RDHoworth 6]

I repeat that this is still being arrogant. Why should you define what is reasonable or not? If a user wants to run this on a Threadripper  running Linux just so that can watch a tone mapped film on a phone. (I realize I am exaggerating). Then that is reasonable for them. 

Some users may want to run Emby on high end hardware, or need to, and we should not take the position that this is not reasonable. If the Emby team took the position that this functionality is not worth developing due to a perceived lack of interest or expectation of user take up, then that is different, but we should know this so that a choice can be made on whether to stay with Emby or not.

[Luke 37060]

5 minutes ago, RDHoworth said:

I repeat that this is still being arrogant. Why should you define what is reasonable or not? If a user wants to run this on a Threadripper  running Linux just so that can watch a tone mapped film on a phone. (I realize I am exaggerating). Then that is reasonable for them. 

Some users may want to run Emby on high end hardware, or need to, and we should not take the position that this is not reasonable. If the Emby team took the position that this functionality is not worth developing due to a perceived lack of interest or expectation of user take up, then that is different, but we should know this so that a choice can be made on whether to stay with Emby or not.

No, we haven't taken that position, and as previously mentioned the feature will be landing on the beta channel for testing very soon. Thanks.

[pwhodges 1528]

5 minutes ago, RDHoworth said:

I repeat that this is still being arrogant. Why should you define what is reasonable or not?

They are not - the currently available hardware is, and Emby have repeatedly said they are working to make this as efficient as they can.  Complaining because they aren't planning to make it better than currently physically possible is a bit unreasonable, I feel.

Paul

[RDHoworth 6]

@cayars, our replies crossed. I really do understand what you are saying, and thank you for being clear. My point is not that I expect to be able to run this without needing to consider very real limitations and set expectations accordingly. We should not criticize people that want to do this, even if "better" methods should be adopted. None of us can identify all use cases, and we really should not try to get users to justify them. It will be up to the users of Emby to either fully use the new feature, and make the investment required, or not, and adopt alternatives.

To be very clear, I applaud the effort of the developers, and offer no criticism of them and deeply respect the obvious technical abilities being demonstrated. I only object to the attitudes of some to try so force their views on others.

[vdatanet 1549]

8 minutes ago, cayars said:

Emby's tone mapping solution should run faster on Windows than other "solutions" if things go well.

Wishing it works well on other platforms like Linux too. I hope you have not focused your efforts solely on Windows.

[Carlo 4330]

Emby is all in on this now and moving forward.  Don't know how much more we can say except, just wait and see what it can do for you on your system.

How many simultaneous tone mapped transcodes you NEED SIMULTANIOUSLY will play into the hardware needed or what strategy is best for you (transcode vs 1080 versions). Everyone has different needs, but I think it's safe to say for "casual use" Emby should have something that will work for those with proper hardware.

For someone like myself who only shares with immediate family (inside and outside home) this solution should hopefully work well but I wouldn't expect it to work shared to a half dozen people all in need of tone mapped transcoding at the same time.

[Carlo 4330]

3 minutes ago, vdatanet said:

Wishing it works well on other platforms like Linux too. I hope you have not focused your efforts solely on Windows.

Nope, not at all.  We're trying to support best we can the different platforms Emby runs on!

[Xorp 47]

On 1/29/2021 at 5:31 PM, Sammy said:

BTW, this thread as a week shy of three years old and the issue persists.

It's also extremely bizarre and embarrassing Jellyfin beat Emby to this feature. For those that don't know, Jellyfin is a Emby knockoff community project using years old Emby code as the base. Yet they seem to have way more rapid feature development than the flagship version.

[Carlo 4330]

It's a half @ss implementation that barely works.  Have you read the forums and all the people it doesn't work for (most)?

Try it yourself and see if you can get it to work. Then take note of what speed it does.  If you get 1.2X speed you can mostly support one transcode and 2 will kill your system so neither works correctly.

People running Plex and JellyFin aren't jumping and down with joy about how good it works unless you're one of the few with high end hardware with just the right OS.

[vdatanet 1549]

30 minutes ago, Xorp said:

It's also extremely bizarre and embarrassing Jellyfin beat Emby to this feature. For those that don't know, Jellyfin is a Emby knockoff community project using years old Emby code as the base. Yet they seem to have way more rapid feature development than the flagship version.

To make a garden, you need more than a couple of flowers. Look at their client apps and you will see the difference.

[softworkz 3335]

1 hour ago, Xorp said:

It's also extremely bizarre and embarrassing Jellyfin beat Emby to this feature. For those that don't know, Jellyfin is a Emby knockoff community project using years old Emby code as the base. Yet they seem to have way more rapid feature development than the flagship version.

You are joking. What they have is something that we would never publish. Their software tone mapping is so slow that they are not offering it. Their Nvidia tone mapping is slow and they have no solution for QuickSync.

[cryzis 7]

21 minutes ago, softworkz said:

You are joking. What they have is something that we would never publish. Their software tone mapping is so slow that they are not offering it. Their Nvidia tone mapping is slow and they have no solution for QuickSync.

I have been watching this thread for a few years, I was one the first people to get HW tone mapping working within jellyfin as soon as the patches were merged. What you are saying is incorrect. Nvidia tone mapping is in no way "slow". My P2000 could handle 5 simultaneous 4k HDR transcodes. It was VRAM limited (1GB per stream, OpenCL needed this for best performance). As for Intel hardware VAAPI is supported, and the developer has merged FFmpeg patches to get the tonemap_vaapi filter working in KBL+. This allows for complete tone mapping within CPU HW. The OpenCL solution works rather well. You are correct native quick sync support is not there yet but is being actively developed.

From the looks of things it does appear you have been leapfrogged with respect to this feature. I have faith that your solution will be performant and produce great results, perhaps you have figured out some great ways to optimize the pipeline. I am looking forward to testing out what you release, but at least back up what you are saying when you make claims to what is currently out there.