Copy link to clipboard
Copied
Ran into trouble using proxies that were generated without the same number of audio channels as the original source. Without matching audio track assigns PPro will not allow it to be a workable proxy.
Worked out a number of ways to accommodate this and a series of consequent challenges, but in retrospect, would it no make more sense for PPro's proxy logic to allow it to ignore audio altogether. Sound is not generally the playback bottleneck necessitating proxies, so would it not make more sense, in Proxy mode, to use picture-only from the Proxy, sound only from the Original?
I can imagine, for technical reason, this is not possible: perhaps playing sound from the original media requires the same throughput as playing sound + picture. But who knows.
Copy link to clipboard
Copied
Copy link to clipboard
Copied
Copy link to clipboard
Copied
Using Cineform proxies, the sound should always match the original, thus obviating any issues.
It's only with the H.264 presets that things go weird.
Copy link to clipboard
Copied
Thanks Jim. Much appreciated. Will start testing and follow up.
Copy link to clipboard
Copied
Jim_Simon wrote
Using Cineform proxies, the sound should always match the original, thus obviating any issues.
Hi Jim,
By Cineform, you mean GoPro Cineform?
Did indeed protect the audio track assigns and worked as a Proxy.
Jim_Simon wrote
It's only with the H.264 presets that things go weird.
Helpful again. Thanks.
Since yesterday got some interesting progress. Switching over to ffmpeg (instead of AME) for proxies I was able to generate an MP4 file (wrapper) with H264 video and AAC audio, all tracks mapped properly from source. It works perfectly as a PPro Proxy.
The benefits of the ffmpeg/MP4 approach are pretty dramatic
Starting with an original UHD "4k" file with 4 channels of audio at 27gb
Codec | File Size | Savings |
---|---|---|
Original File: MXF wrapper VIdeo: H.264 Audio: PCM (4 channels) | 27gb | original file |
Cineform at 1270x720 Quality 5 (highest) | 5.2gb | 5x |
Cineform at 1270x720 Quality 1 (lowest) | 1.9gb | 14x |
Cineform at 640x360 Quality 1 (lowest) (terrible) | 887mb | 31x |
ffmpeg-generated MP4 file Video: H.264 Audio: AAC (4 channels) | 373mb | 73x !! |
The quality of the ffmpeg-generated MP4 is excellent, and can be dialed up or down depending on file size / quality tradeoff
The channel mapping is automatic, so 2-channel sources and 4-channel sources map appropriate to the final.
The down-side of using FFMPEG is the effort involved in leaving the Adobe ecosystem / working in Command Line.
The Proxies don't connect by themselves, so you're either manually doing each on individually or running some ExtendScript to connect in batch.
But you also have more custom control over naming conventions and target folder than you do with AME.
So the good news is that it is possible create a Premiere-Compatible Proxy file with excellent quality at small file sizes using H.264 compression and more than 2 channels of sound.
But so far I don't how to achieve this with AME
Copy link to clipboard
Copied
I think you've got the wrong viewpoint here. The point of proxies is to make editing easier. In the early days of NLE's, hard drives were the limiting factor, so smaller files made sense. Today, the CPU is the limiting factor, where the small file size (compression) is actually the problem.
I think the best option is to just use the included Cineform presets. Hard drives are cheap.
Copy link to clipboard
Copied
Hard drives are cheap indeed. So are production companies. (accompanied by priority-scrambling spreadsheets)
But your point is well-taken.
Copy link to clipboard
Copied
Been testing using mp4 / h.264 proxies instead of the PPro presets (e.g. Cineform, ProRes, etc)
There's no performance practical performance hit as a result of the h.264s, and you get the 73x drive savings.
There's a slight color shift but minimal, and it's easy enough to toggle back to high res for fine tuning color correction.
So far between the drive-hungry PPro presets, and the compressed h.264 / mp4, the latter wins.