[brazilianaire]

Hi Sam and Grace,

It’s interesting the choking and buffering, because this happens constantly with my videos on my system and browser, but not with Grace’s.

All of my videos choke when I view them on my site, once in a while they stream well, but not usually and I assumed it is because of going through your pipe, and so was quietly waiting for your standalone solution.

I have a liquidweb VPN which delivers everything else quickly. I primarily use Firefox but also test with Chrome and Edge.

Best, Jamie

[brazilianaire]

Hi Grace,

They all look similar, and you know what’s crazy is I was born in Milwaukee Wisconsin which he starts the show with! Hahahaaa

The 480 is more than adequate and I didn’t see much difference among the three watching both on a laptop screen and an HDMI feed to a 1080 LCD TV. If cost were an issue then 480p and 720p would be enough resolutions to serve all but UHD tv sets.

[brazilianaire]

Will check them out.

The LTO6 should be able to serve as a working drive, since version 6 is quite fast and the tapes have great capacity.

Best, Jamie

[brazilianaire]

Yes, Blackmagic and other tools are designed to use standard resolutions, so maybe open the SD file into a 1080p timeline and fill to fit, but then only export via media encoder at 720p…see how that works.

If you are investing then buy an LTO6 drive and you can store and access at really great costs and have the safety of institutional archival security

[brazilianaire]

Hi Grace and Sam,

Drifting audio sync is strange…Does it happen with your mezzanine-level transcodes?
This is a reason to stick with the original movie’s frame rate at all times and not separate your audio during export to mp4.

When you say you are uploading your audio masters to bubble what do you mean? Aren’t they muxed to the video already? If not, they should export as part of the mp4 (H264) from Media Encoder to avoid video sync issues with mp3s from WAVs.

I understand storage space is a cost concern, but believe it is also wise to to prepare all elements with final resolution, frame rate and audio sampling as early as possible in your pipeline and not leave anything for Elastic Transcoder to do except compress. No conversion or correction…
Delete the mp4 master later if cost is an issue.

I export my mp4s at much higher bit rates and have much better looking stream compressions as a result.
Typically 12mbps with a 18mbps peak video data rate, and sometimes 24mbps with 36mbps peak.
(I was told peak should be 50% more than average for best results) and AAC audio should emphasize Bit rate.

Try max keyframe distance of 1 for better frames.

If 44.1K 320kbps stereo is your goal, then make that happen out of Media Encoder.
Do everything to your master once, using media encoder, which is an excellent tool with perfect pixels, so ET only compresses at cleanly divisible resolutions. I don’t use 240p at all.

If your master is SD-29.97i, test making your export timeline the final desired res and rate of 720-30p (29.97p) so you can use your preview renders on export and accelerate exports.
Always max render quality checked in Media Encoder.

Would love to see your results!

Best, Jamie

[brazilianaire]

Hello Sam and Vintage,

I just had a very enlightening conversation with a key audio partner.

He says despite 48K being a better quality signal than 44.1 (44.1 kills high-end and subtle frequencies), today most cellular phones, earbuds, consumer speakers, mobile devices and many consumer sets like laptops and TVs are “pathing” all audio signals at 44.1K without professional buffers, so that a 48K signal will not be transcoded, but simply “sent”, and can result in a lousy final sound if the original audio is 48K.

His advice is that a “properly mastered” 44.1K sample at 320kbps stereo, then encoded for streaming is an excellent compromise across all devices.

This quality audio at a live show would not be acceptable, but for cross-platform use almost no one would notice.

In addition, he would avoid AACs and MP3s at less than 320kbps stereo whenever possible, because the damage to signal quality is intense at lower kbps, and a much more significant quality factor than 48k versus 44.1.

Sam, do you know if the transcoder can enable audio dithering?
Dithering solves the signal gaps created by audio compression algorithms.

I hope this helps!

Best,
Jamie

[brazilianaire]

Hi Sam,

Love what you are preparing here!

For me I would test with 48K over 44.1 even at your per channel bit rate of 128 (256kbps stereo), but I would love to see if going to 160kbps (320kbps stereo) makes that much difference.

Soundcloud really steps on the original audio with the compressions they use and my musicians are not happy with their results on Soundcloud. My goal is to offer higher quality that still streams cleanly.

VIDEO: When I encode in the studio I leave my max keyframe distance to 1 and there appears to be no increase in file size and my h264s are gorgeous.
That is also because I push my video bit rates up more for a much cleaner file, so when served from Cloudfront I can let the resolution switcher handle the stream bottlenecks over simply having more compressed frames at all resolutions.

I believe once you launch a standalone pipeline that does not flow through S3Bubble throughput should be amazing for everyone!

Thank you for everything!

Best, Jamie

[brazilianaire]

Hi Grace,

I don’t want to reply for Sam or S3 Bubble.
That said, 44.1K should save some storage space over 48K but does impact sound quality for very high quality audio with great dynamic range.

Both 44.1 and 48 can be set when compressing to mp3/mp4 file format, and your kbps becomes the primary quality driver thereafter. The best “standard” compressed audio is 48K 320kbps Stereo (160kbps per mono audio channel)

I would do some tests, since if much of the audio comes from video originally mastered for traditional TV broadcast (48K 16Bit Stereo), the dynamic range should be only around audio 6 decibels (or not a very high quality audio for a modern audio sound system), whereas audio from a motion picture soundtrack could have a range of 15+ decibels and much of the details could be lost with a 44.1K conversion.

Returning to video formats, the potential issue with encoding at your proposed resolution is that the HD flat screen tvs of today and moreover the UHD screens coming on the market will need to upres your sub-HD pixels and those pixels were already compromised at time of compression so the resulting tv upres will probably be really tragic and could affect your library value over time.

If you do some math on how many transcoded pixels will have to be divided and reassembled from your 846×486 to get to 1080p or UHD 2160p and consider how the pixel-division-math is cleaner with HD resolution displayed on an HDTV set (which is not always variable resolution like a computer screen) you may find a higher quality image from native HD masters before transcoding.

Lastly, have you tested mp4 conversion using the native file frame rate family of 30p instead of 23.976?

Best,
Jamie

[brazilianaire]

Hi Grace,
I am following your thread and have worked extensively on motion picture telecined to Standard Def in the past and still have my own elements on SD.

You may consider aiming for a 1280x720p with pillarbox…ideally when you digitize from tape or workprint you digitize directly to 1280×720 and not 720×486…or at minimum resample the non square D1 pixels to square pixel creating a 720x540p…getting you much closer to an HD resolution of 1280x720p.

The footage is already vintage so it should survive upresing to 720p especially if your concept is consistent across your library.

As for your audio, I am curious as well as I prefer 48K even if it is an MP3.

Best,
James P. O’Malley

[brazilianaire]

Thank you Sam!

Best, James

[brazilianaire]

I am curious as well.

James

[Author]

Posts