2 people found this helpful
I propose NOT using a high-bit rate H.264, but rather a lossless intermediate codec, like JPEG 2000. Furthermore, I propose using Adobe Media Encoder, then just replacing shots in PP.
Do you have a short comp that's already been rendered? You can set it up in AME and compare export times.
It might not be as nifty-cool as your method -- which sidesteps & automates what I propose above -- but c'mon! You're talking 150 comps! I for one do not have the confidence in Adobe software to perform your task flawlessly.
3 people found this helpful
I would suggest rendering to a production codec (DNxHD, Cineform, etc.). They're a lot less heavy on a computer vs. H.264
You can then take the resulting file back into AME to make the H.264.
That may be faster, but it might not. I would definitely have been doing render and replace along the way.
Alternatively, it might go faster if you render via After Effects into a production codec and then drop those clips in. It all kind of depends on the nature of your compositions.
You could also consider going with yesterday's update. There were a number of improvements in the AE to Premiere workflow including speed improvements.
Thank you, Dave! I am going to try your suggestion, I hadn't considered that option!
Szalam, Thank you for your help! I recently reran the render after cleaning up my system and if this doesn't complete, I am moving on to your method. Thank you again!
No problem. Looks like Dave and I had some similar thoughts!
Please let us know how things go one way or the other!
The video finished rendering- 49 hours total. I have some corrections to make, and I want to render it out with the Lagarith codec this time. Do you suspect that would take even longer or could that possibly go smoother than the h.264? Also, I am noticing on my trial uploads to youtube, every shot with any noise/grain added becomes distorted, almost as if the noise effect itself becomes mosiaced, do you know how I could combat that? In vlc player, I don't get the same artifacts I do as on youtube. Perhaps it's unavoidable with the YT compression, although I have seen other videos on YT with extensive noise that seems to go undistorted. Would vimeo fair better?
The most affected shot in the video is the opening title shot: YouTube
(Any encoding woes aside, nice work!)
Thank you very much, Kyle!
Yes, stuff on Vimeo tends to look better than on YouTube.
I don't have enough experience with the Lagarith codec to advise how it would behave.
1 person found this helpful
When you make changes you only have to re-render the changed frames. You then take your production master into something like Premiere Pro and do the edit. Then you render your final product through Premiere Pro using the Adobe Media Encoder - the only way you can render a Premiere Pro Sequence. I would never make a few changes in a render that took 49 hours to complete and then try and re-render the whole thing. I would never put a project that took 49 hours to render in the AME to render a multi-pass delivery product - it will take longer than 49 hours, be very difficult to fix if the render fails, and is basically a waste of time.
AE is for fixing or creating shots that you cannot fix or produce in a NLE. AE's best use to render production masters or DI's of those shots to be edited in a NLE like Premiere Pro. That sounds like it's more steps and less efficient but in the long run it's a much faster way to work on any project longer than a few seconds. Most of my AE comps are under 7 secoonds. Most or the films I work on are about 2 hours long. Many of my compositing renders from AE take as many as 2 or 3 minutes a frame to render, some render at several frames per second and the longest - by design - never take longer than 7 minutes a frame. My final edits in Premiere pro usually render at the rate of several frames per minute.
I hope these suggestions help. Managing rendering is almost as important as managing any other part of the production workflow if you are ever going to get anything done and ever have a chance at making a living making films.
Thank you for your generous response, Rick. I agree that efficiency is crucial not only in one's render process, but within their workflow all together. The few "changes" I alluded to making earlier was actually a pretty extensive VFX revision.
This project was interesting, in that it required an unorthodox workflow. I had found the structure and cut the piece prior to any of the environments having been built or any of the VFX that was done. I storyboarded the video AFTER organizing a kinetic sequence based on the musician's inherent performance, which during shooting was based on and influenced by an abstract narrative that has been evolving from pre into post production. I completed the narrative after cutting the performance, then moved onto concept art, then began building and animating the scenes, and then lastly compositing the elements together. What I ended up with, at the time, was the master stack of the video filled with unrendered dynamically linked compositions, which I couldn't playback in my nle because they hadn't been rendered, yet. This recent render was both a test as far as encoding and exporting settings in combination with streaming sites preform, as well as the first version of the project that I could see in real time in it's entirety- post the compositing phase.
This is also the first time I've explored a workflow like this and I suppose I wanted to experiment in a way with the process- it is definitely NOT the most efficient method as far as allocating processing power, time management, or for a project that is going to involve changes/revisions in your NLE down the road. I'm luckily not constrained by any of which for this project.
Your suggestions do help- I'll definitely be paying attention for more opportunities to maximize workflow efficiency in my next project. thank you again for your insight!
Just finished the final revisions. I'm going to be rendering out my master stack from Premiere in AME as a lossless PNG sequence, then I'll replace the dynamically linked compositions in Premiere with the image sequence and render that out with the HVEC codec, while up scaling to 1440p to force YouTube/Vimeo to process(Re-encode) the video at a higher bitrate than it would if I exported at 1080p.
It appears as though rendering out as a PNG sequence makes only a minor improvement in render speeds, at most perhaps it's 10% more efficient.. Although, if I were to make any changes now, I would only have to replace the PNG frames I've changed as opposed to re-exporting every frame again like I would have had to do if I were not using the image sequence workflow.
Thanks for the advice and suggestions!
I'll replace the dynamically linked compositions in Premiere with the image sequence and render that out with the HVEC codec, while up scaling to 1440p to force YouTube/Vimeo to process(Re-encode) the video at a higher bitrate than it would if I exported at 1080p.
I don't think you understand codecs, frame sizes or bit rates. You also don't seem to have a good understanding of YouTube and how they handle compression. 1440p is sometimes called quad HD - it's twice as big but if you scale up your comp 200% you'll loose a lot more resolution and quality than if you encode at the highest recommended bit rate for YouTube using muti pass encoding. Throw a 1440p file to YouTube and they will re-compress for 1080 playback and do a mediocre job of it. While you may get the higher bit rate with the 1440p file it will not look as good as sending YouTube a properly prepared HD file.
If you want to work in Quad HD (1440P) then you need to start there rather than scale up. If you insist in scaling up then you should not ever scale up when rendering, you should drop your completed HD render in a 1440p comp or sequence and then use the best scaling plug-in you can afford to resize the video. AE's content preserving upscale is pretty good. Premiere does a fair job, but for really good results look for other solutions from folks like Red Giant. Just dropping a 1080 file in the encoder and scaling it up by 200% will not produce anything close to professional results. There is no free lunch when it comes to encoding.
The original comps are 1440p.
Hello I've returned!
Feeling slightly triumphant considering my encoding woes have been defeated. I finished this project tonight. My final render time was around 1.2 hours with a two pass. Not a bad improvement. I appreciate everyone's input! The image sequence workflow has proven to be tremendously useful!
Somewhere above is the first draft's link on YouTube, going to link the final product here: Fusion Dance | zach johnson - YouTube