Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
What happens if you export with software encoder? Is it now 44MB?

If I export with same settings in software rendering then render time is 6 seconds and file size is 44.7MB. So Premiere's estimation is always for software rendering export.

Visually hardware and software rendering is identical for this example.
 
If I export with same settings in software rendering then render time is 6 seconds and file size is 44.7MB. So Premiere's estimation is always for software rendering export.

Visually hardware and software rendering is identical for this example.
I must be a little skeptical when I see a file size as 44MB vs 29MB. Obviously their is more data in the 44MB version using the same codec. Are you certain the quality between the two are the same? Because it seems that by having two different sized exports using the same codec would vary in quality. Perhaps the software codec is heavy or perhaps the hardware export is too lean?
 
I must be a little skeptical when I see a file size as 44MB vs 29MB. Obviously their is more data in the 44MB version using the same codec. Are you certain the quality between the two are the same? Because it seems that by having two different sized exports using the same codec would vary in quality. Perhaps the software codec is heavy or perhaps the hardware export is too lean?
I'm just about to give the above a test for comparison. I agree @gomakoto - if a file is smaller in size, there is 100% less data in that file. Two files identical files exported with the same codec/format should be identical file sizes with hardware and software encoding. One shouldn't produce a difference so I would consider this a fault/bug relating to M1/M1 Pro/M1 Max. It's worth noting that when I did the comparative export on my windows desktop with software and hardware encoding the files were the same size and so was the quality. The difference in file size/quality was a direct result of hardware encoding with the M1 Pro.
 
@birmexer

Result so you can try the same thing and verify.

Latest Premiere 2022.

Taking the first three 4K cllps from the link above. Putting them on a timeline.

Three videos together are 22 seconds long.

Added the ALEXA Default LogC2Rec709 LUT.

Exported using h.264, YouTube1080p HD Preset.

Estimated file size was 44MB.

Export render time was 2 and a half seconds on M1 Max 32 core.

Exported video was 29.7MB, less than the estimated size.

Quality is perfect. Not a single pixel wrong, no artefacts.
Thanks for doing this! I'm going to give it a go as soon as I can see the link to the files on the forum. It's interesting you say there is no visual difference as you would expect a visual difference when two files encoded at the same bitrate are different in size.

It's worth noting for the comparison, I exported the same project I exported from the Mac M1 Pro on my Windows desktop with both software/hardware encoding and the file was identical - no compression artefacts, blurriness or pixelation like what happens when exporting the identical project on M1 Pro using hardware encoding.
 
  • Like
Reactions: metapunk2077fail
I must be a little skeptical when I see a file size as 44MB vs 29MB. Obviously their is more data in the 44MB version using the same codec. Are you certain the quality between the two are the same? Because it seems that by having two different sized exports using the same codec would vary in quality. Perhaps the software codec is heavy or perhaps the hardware export is too lean?

Notice I carefully say 'Visually hardware and software rendering is identical for this example.'

Because two videos or images can look 'visually identical' to the eyes even if one has more data than the other. For example a jpeg saved with Quality level 8 in Photoshop looks identical to Quality level 12, but the file size is smaller. You can save a tiff as a jpeg and they can visually look identical also.

The important thing is that there are no visible artefacts and that the colour grading has exported correctly.
 
Thanks for doing this! I'm going to give it a go as soon as I can see the link to the files on the forum. It's interesting you say there is no visual difference as you would expect a visual difference when two files encoded at the same bitrate are different in size.

Visual differences and artefacts would probably become apparent at a much smaller file size.

When you start off with ProRes 444 or any log profile you have so much freedom for grading and for exporting to smaller file sizes without losing quality.

If you start of with h264 as your source files then there is much more chance for error. Any small colour grading or export option can have artefacts or banding.
 
Visual differences and artefacts would probably become apparent at a much smaller file size.

When you start off with ProRes 444 or any log profile you have so much freedom for grading and for exporting to smaller file sizes without losing quality.

If you start of with h264 as your source files then there is much more chance for error. Any small colour grading or export option can have artefacts or banding.
Sorry I should have added I shoot in log with Canon EOS R - but the file is still a MP4 with H264 when I look at the metadata. I'm not that clued up about codecs in truth!

I hear what you're saying but why would the files be same with software/hardware encoding on my Windows desktop or my Intel iMac compared with significantly different files for software/hardware encoding on the M1 Pro? It's only when I tested on Windows computer and Intel iMac that I realised this issue was isolated to the M1 Pro chip.

It's really weird - the hardware encoded M1 Pro file is substantially worse quality, and it's very noticeable when compared to the same project I exported on my Windows desktop and my Intel iMac using hardware encoding.
 
Sorry I should have added I shoot in log with Canon EOS R - but the file is still a MP4 with H264 when I look at the metadata. I'm not that clued up about codecs in truth!

I hear what you're saying but why would the files be same with software/hardware encoding on my Windows desktop or my Intel iMac compared with significantly different files for software/hardware encoding on the M1 Pro? It's only when I tested on Windows computer and Intel iMac that I realised this issue was isolated to the M1 Pro chip.

It's really weird - the hardware encoded M1 Pro file is substantially worse quality, and it's very noticeable when compared to the same project I exported on my Windows desktop and my Intel iMac using hardware encoding.

Good you're shooting log but would be even better if the camera supported the MOV format. One of my clients was sending me mp4 files and they were a pain to grade on macOS and Windows and I hated the exports. Then I recommended them a videographer who shoots log MOV files and everything was smooth since then.

I don't know if it is Adobe or Apple's encoder that is responsible for this file size estimation discrepancy. I would imagine Adobe needs to update the app to recognise which hardware encoder is being used and then make a correct estimation. In any case if they don't fix this, you can just increase the bit rate in the export window and get the file size you want.
 
Good you're shooting log but would be even better if the camera supported the MOV format. One of my clients was sending me mp4 files and they were a pain to grade on macOS and Windows and I hated the exports. Then I recommended them a videographer who shoots log MOV files and everything was smooth since then.

I don't know if it is Adobe or Apple's encoder that is responsible for this file size estimation discrepancy. I would imagine Adobe needs to update the app to recognise which hardware encoder is being used and then make a correct estimation. In any case if they don't fix this, you can just increase the bit rate in the export window and get the file size you want.
It's not that I'm bothered so much about the file size estimate, it's more that the quality is so poor with the hardware exported version on the M1 Pro. My initial complaint to Adobe has been escalated to a 'senior engineer' who is going to call me tomorrow so I will see what they say.

Thanks again for running that test export!
 
I would hope that files exported whether software or hardware would produce the same exact file. The method being used should not change the file; The settings should.

I would want this fixed because whether the cpu or gpu is doing the work shouldn’t result in different files.
 
  • Like
Reactions: birmexer
I would hope that files exported whether software or hardware would produce the same exact file. The method being used should not change the file; The settings should.

I would want this fixed because whether the cpu or gpu is doing the work shouldn’t result in different files.
Agree; CPU or GPU doing the work should make no difference to the file. Will see what Adobe say tonight - they will probably just fob me off with "use software encoding then" again!

Apple were not interested at all and just said it's Adobe who need to look into and investigate as it's an issue with their software - not sure how accurate that statement is though. It could be the way the chip handles the data or anything causing this issue.
 
  • Like
Reactions: gomakoto
@gomakoto @metapunk2077fail

OK, I noticed another fundamental issue today - which I think is related to the GPU again.

I've got a 25fps sequence and added footage shot at 50fps for slow motion. When I add a warp stabiliser to this footage on a 25fps it becomes heavily pixelated (looking very similar to how it exports when using hardware encoding). This doesn't happen if I put this 50fps footage on a 50fps sequence and then add a warp stabiliser.

Also, this works fine with no pixelation on my Intel Mac and Windows PC so again, it looks related to the M1 Pro chip.
 
  • Like
Reactions: gomakoto
@gomakoto @metapunk2077fail

OK, I noticed another fundamental issue today - which I think is related to the GPU again.

I've got a 25fps sequence and added footage shot at 50fps for slow motion. When I add a warp stabiliser to this footage on a 25fps it becomes heavily pixelated (looking very similar to how it exports when using hardware encoding). This doesn't happen if I put this 50fps footage on a 50fps sequence and then add a warp stabiliser.

Also, this works fine with no pixelation on my Intel Mac and Windows PC so again, it looks related to the M1 Pro chip.

Is the footage h.264?

If you can find a similar file with the same type of content, same frame rate and codec online then I can download it and run the same test. None of the footage I have has those same specs.
 
Is the footage h.264?

If you can find a similar file with the same type of content, same frame rate and codec online then I can download it and run the same test. None of the footage I have has those same specs.
Ah OK, nice one. I'll upload some to OneDrive/Google Drive and put the link here.
 
  • Like
Reactions: metapunk2077fail
Is the footage h.264?

If you can find a similar file with the same type of content, same frame rate and codec online then I can download it and run the same test. None of the footage I have has those same specs.
I'm just uploading some now but I've just figured out that it doesn't happen on a clip without lumetri colour adjustments... if you add the warp stabiliser first and then add lumetri colour it doesn't produce this issue.

Also, if you change the order of the effects in the effects list so warp stabiliser is first (above lumetri colour) it doesn't do this and the footage looks fine.
 
  • Like
Reactions: gomakoto
I'm just uploading some now but I've just figured out that it doesn't happen on a clip without lumetri colour adjustments... if you add the warp stabiliser first and then add lumetri colour it doesn't produce this issue.

Also, if you change the order of the effects in the effects list so warp stabiliser is first (above lumetri colour) it doesn't do this and the footage looks fine.
That's got to be an Adobe bug. They have a forum for reporting bugs but not a good in-app feedback system. Ever heavy app should have a 'Report Bug' item in the app somewhere.
 
That's got to be an Adobe bug. They have a forum for reporting bugs but not a good in-app feedback system. Ever heavy app should have a 'Report Bug' item in the app somewhere.
I've reported it on their platform just now. They're supposed to be calling me at 6pm to discuss this hardware encoding issue so I'll mention to them then as well.
 
I've reported it on their platform just now. They're supposed to be calling me at 6pm to discuss this hardware encoding issue so I'll mention to them then as well.

I confirm your results. If I do warp before colour there is no pixelation. If I colour first and then warp, the image quality is obviously pixelated. I used the YouTube 1080p export preset with max quality and 20mbps bit rate.
 
My M1 hasn’t arrived yet (Nov 12). However if it was here I’d be happy to assist with bug reporting / testing.
 
I confirm your results. If I do warp before colour there is no pixelation. If I colour first and then warp, the image quality is obviously pixelated. I used the YouTube 1080p export preset with max quality and 20mbps bit rate.

This is the thread on Adobe's bug platform thing if you want to vote it up...

Thanks for taking a look for me!
 
I confirm your results. If I do warp before colour there is no pixelation. If I colour first and then warp, the image quality is obviously pixelated. I used the YouTube 1080p export preset with max quality and 20mbps bit rate.
Adobe confirmed they know about both issues with M1 chip range and said they are working to release a fix at the end of this month for both.... we'll see!
 
I'm having similar problems with smaller files than anticipated and artifacts. My head spun a bit reading this chain as I'm not all that technical, but am I distilling correctly that we basically have to use software encoding and everything will be fine? Until Adobe fixes it?

Thanks for these posts btw...it's very helpful.
 
I'm having similar problems with smaller files than anticipated and artifacts. My head spun a bit reading this chain as I'm not all that technical, but am I distilling correctly that we basically have to use software encoding and everything will be fine? Until Adobe fixes it?

Thanks for these posts btw...it's very helpful.
Yes, just use software encoding for the time being and Adobe should release a fix for it.

Interestingly though I swapped my M1 Pro for a M1 Max and it doesn't seem to do it anymore on M1 Max! Weird.
 
Yes, just use software encoding for the time being and Adobe should release a fix for it.

Interestingly though I swapped my M1 Pro for a M1 Max and it doesn't seem to do it anymore on M1 Max! Weird.
Thanks! For whatever it’s worth, I’m using a M1 Max and having the issue. Videos export with pixelation (and they never did on my old computer). Trying software encoding now. Fingers crossed!
 
  • Like
Reactions: birmexer
I am so glad I found this thread. I thought I was loosing my mind! I have a 14" M1 Max, 64GB. I have done two projects so far in Premiere Pro 2022. I accidentally stumbled across setting Premiere to use software for encoding and it worked. I also had to set Media Encoder to use software too. I was having all the above listed issues. Media Encoder would start then just freeze up and after 10-15 minutes maybe start encoding - or not. Changed it to software and it took off finishing in minutes. Video looked great. Most of my videos are in the one hour range.

Hopefully Adobe will get this fixed.
 
  • Like
Reactions: birmexer
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.