Mainconcept AVC/AAC Single Pass vs Two Pass Render

john_dennis wrote on 10/19/2017, 12:12 AM

Most people don't want to spend time rendering a video after the time it takes to edit their masterpieces, but I often stare at the two pass check box dreaming of fast hardware that produces instantaneous renders and wonder "Why not check the box?"

Rather than watch Empire with my wife, I rendered two versions of a recent project with the Mainconcept AVC/AAC Internet template customized like this:

The resulting Single Pass render produced this Mediainfo result:

The time to render Single Pass was 10:20 (mm:ss) for a 07:23 (mm:ff) timeline.

The Two Pass render produced the following Mediainfo result:

The time to render Two Pass was 20:14 (mm:ss) for a 07::23 (mm:ff) timeline.

This video graphically describes the bit rate distribution between the single pass and two pass renders:

Here are my observations:

1) The two pass renders match the target bit rate expectation more closely than single pass renders.

2) The bit rates appear more variable using two pass. Conceivably, the bits are allocated to more closely match the detail and/or motion requirements of the source. That's probably giving the encoder the benefit of the doubt since I haven't proved the premise scientifically.

3) The two pass render produces a file that is marginally smaller.

If you're perplexed about what this all means, look at my profile. I'm curious.

All renders done in Vegas Pro 13-453.

Comments

Musicvid wrote on 10/19/2017, 10:56 AM

What were render times for the two?

john_dennis wrote on 10/19/2017, 11:26 AM

I didn’t notice, but I’ll recreate later. I had already written off time when I checked the Two Pass box.

NormanPCN wrote on 10/19/2017, 1:55 PM

Here are my observations:

1) The two pass renders match the target bit rate expectation more closely than single pass renders.

Agreed. Mainconcept rate control is pretty slack, especially in single pass. They make assumptions about a certain amount of savings and if you don't meet that then they overshoot. Sometimes by a lot. How much they overshoot is also related to how much above the average rate the max rate is.

2) The bit rates appear more variable using two pass. Conceivably, the bits are allocated to more closely match the detail and/or motion requirements of the source. That's probably giving the encoder the benefit of the doubt since I haven't proved the premise scientifically.

Yep and it typically should be more variable. The first pass the encoder simply looks at every frame and makes a determination of complexity and bitrate requirements. On the second pass they can use this information. In single pass mode they only know what happened in the past and cannot know what comes. The two pass mode gives visibility of all future frames.

3) The two pass render produces a file that is marginally smaller.

Similar to #1. The rate control is closer to the target. Less overshoot. Actually with two pass they (MC) should hit the target with a bullseye. No excuses. x264 can hit the target in ABR mode in a single pass. But we all know Mainconcept is no x264. In more ways than final quality.

john_dennis wrote on 10/19/2017, 8:39 PM

I updated the original post to include the time to render for each case.

Thanks @NormanPCN for your observations. Sometimes I think of myself more as an archivist than a videographer. If your time horizon is decades the time to render becomes less significant. Then, there is always the argument that if the render completes while I'm on my daily walk, "did it really take that long?"

Former user wrote on 10/20/2017, 2:35 AM

2-pass is best solution for getting the highest quality at lowest bitrate. That is where a circumstance requires a bitrate lower than optimal. Maybe 6mbit at 1080p30. for end delivery via a website (The encode the website viewer watches). I would think 2-pass is your only choice for best quality video. It's still going to depend on the video. If it's 10mins of frantic constant movement and explosions and ocean waves you will have a poor quality video, but if only 5mins is like that, and the rest is nice stable shots using gimbals and tripods and some seated interviews it could be a good encode.

If you had very bad internet upload speeds would also be a use for 2-pass, but it seems in real use situations it's always about compromise and reducing average bitrate to the lowest required for a situation. If you encode for highest quality without compromise 2-pass just doubles your render time without any real benefit.

Musicvid wrote on 10/20/2017, 10:30 AM

I would think 2-pass is your only choice for best quality video.

CQ/CRF encoding is much faster and better quality metrics than 2-pass ABR. That goes double at reduced bitrates.

A lot of people would be using if it were natively available in Vegas. The downside is that it is source-driven, so file size and bitrate targets cannot be predicted.

john_dennis wrote on 10/20/2017, 10:45 AM

Thank you for your observations bob-h. After my first cup of coffee this morning and reading the whole thread again, I realized that my personal bias is not “the highest quality at the lowest bit rate.” When I typed 28 mbps into the template, I had a strong expectation that the resulting average bit rate would be 28 mbps and the upper limit would be 35 mbps. The single pass render failed to produce the result that I asked of it. To NormanPCN’s point, predictable rate control seems to only be available in Mainconcept with two pass renders.

Rich Parry wrote on 11/2/2017, 11:58 PM

This interesting thread concentrates on 1 vs. 2 pass rendering. It got me to thinking, is there general agreement if the Sony or MainConcept provides better quality video assuming both use single pass rendering since Sony doesn't provide that option.

Rich

john_dennis wrote on 11/3/2017, 12:38 PM

Some years ago, I used the Sony template for Blu-ray because it was faster than the Mainconcept given my limited hardware. These days, for Blu-ray I've been using Mainconcept Two-Pass at or about the same bit rates the camera captured it. I'm probably too blind to even make a quality judgment.

Anecdotal Sharing

On one of my recent Blu-rays, the out-of-the-box Mainconcept render template in Vegas 13-453 produced a video file that would not pass through DVD Architect after rendering. It was title 10 of 12 in the DVD Architect project and it took me a while to identify which file was at fault. The failure was repeatable through multiple renders using the same template. I was unable to get it to pass until I lowered the 40 Mbps upper limit a few Mbps. The bit rate distribution showed peaks above the 50 Mbps limit of the Blu-ray hardware standard. That's what caused me to look at bit rate variations. I've only had that one event, but it was a big time waster. One has to wonder how I've done something 9 times without a problem and now on render number 10 I have a showstopper?

My SWAG

There are some source files and/or edits that can slip through a single pass Mainconcept render and produce troublesome or illegal bit rates that cause downstream failures. Two pass renders appear to have the rate control advantage even if there is no quality improvement.