Log in

View Full Version : a question about video's resolution


Pages : 1 [2]

hello_hello
29th August 2011, 23:26
Likewise I didn't arbitrarily come up my CRF choices without testing. Admittedly I've got to deal with "happy" file sizes, but if by your own admission a CRF encode and a 2 pass encode of the same file size will be of the same quality, then by not picking the same set file size each time, CRF encoding saves me bits on one video I can spend on another, whether the difference is perceivable or placebo (depending on which argument you're using there at any given time), and encoding is much faster.

If I wanted to, the time I saved encoding using the CRF method instead of 2 passes could be spent using better quality settings instead. I don't mind "happy" file sizes.

lever1119
30th August 2011, 04:36
@lever1119: There are no rules regarding resolutions and bitrates. When I think about encoding for personal purposes, I don't care how big something is, just that the quality is fairly consistent and it's smaller than the source. Usually, CRF values between 18 and 22 accomplish this goal for me very nicely.

If you're concerned with having all your content have identical file sizes, then by all means, use 2 pass VBR. My suggested bitrates are probably appropriate for most content.

Thanks a lot. I know what you mean.
Just wanna be sure. 1000K bitrates, fullscreen watching, is it good that I set up 720* resolution for all my videoes?

Blue_MiSfit
30th August 2011, 05:44
lever1119: I'd say it's a terrible waste to downscale all the way to 720x480, especially if you're coming from a nice 1080p source. I'd maybe do 1280x720 at ~4mbps. This will be very close to transparent for most sources.

If you're hard set on 720x480, then I'd go with at least 1.5mbps.

Derek

nurbs
30th August 2011, 06:16
I do 720p with CRF 21 and it looks pretty good even on 1080p screens. With my settings I get 1.5 to 3 Mbps on most sources, so if you use 4 you should practically always get a transparent result.

hello_hello
30th August 2011, 08:18
I use CRF 20 for 720p encodes. The files are larger then the OP wants (they vary between around 3 and 5gig) but the quality is great. Still looks really good on my 50" TV.

Is there any way to run a quality test before running two passes these days (compression test, that sort of thing)?

When it comes to x264 encoding I generally decide on a resolution and CRF value and let the file size be whatever it needs to be, but as the OP wants to keep the file sizes down.....

Back in the Xvid days using AutoGK I'd try to aim for a certain quality and adjust the resolution down if necessary to keep the file sizes under control. To me dropping the resolution to achieve a file size seems preferable to dropping the quality.

Unfortunately most GUIs don't report the quality that way, but if I was trying to achieve the same file sizes as the OP I'd probably be tempted to still use a CRF value I was happy with and take a guess at the required resolution. Once the encoding has been running for 5 or 10 minutes you generally have a pretty good idea if the file size will fall into an acceptable range. If it looks like it'll be too large I'd stop the encoding, reduce the resolution and start again. It might take a few goes but it'd still probably be quicker than running two full passes, and because you're using CRF encoding you'll know what sort of quality to expect.

What are people's thoughts on tackling it that way?

lever1119
30th August 2011, 08:39
lever1119: I'd say it's a terrible waste to downscale all the way to 720x480, especially if you're coming from a nice 1080p source. I'd maybe do 1280x720 at ~4mbps. This will be very close to transparent for most sources.

If you're hard set on 720x480, then I'd go with at least 1.5mbps.

Derek

Could you make sure for me at last that 720* with 1000k bps is better than 1280* with 1000k bps for most videos on fullscreen watching, please?

(Any technical reasons? Just experiences?)

Target Practice
30th August 2011, 09:01
Could you make sure for me at last that 720* with 1000k bps is better than 1280* with 1000k bps for most videos on fullscreen watching, please?

(Any technical reasons? Just experiences?)OP, run some tests yourself. see my earlier post if you'd like to see how I ran one.

nm
30th August 2011, 09:27
Could you make sure for me at last that 720* with 1000k bps is better than 1280* with 1000k bps for most videos on fullscreen watching, please?

If you must have a bitrate of 1 Mbps, then 480p or something in between might be a better choice than 720p for most videos. I'd go for 960x540.

But as Derek said, 480p is quite a low resolution, so you'll lose a lot of detail from the 1080p sources. Whether that matters to you or not, depends on the screen size and viewing distance.


+1 to using CRF for your encoding task.

lever1119
30th August 2011, 13:47
OP, run some tests yourself. see my earlier post if you'd like to see how I ran one.

If I do tests by myself, I should run a lot, maybe hundreds of footages, 'cause movies got different types and details. For example, this footage fits 720*, but that one fits 1024*. I need an average resolution for all the movies. So it's unlikely to get the result on this way.

lever1119
30th August 2011, 13:48
If you must have a bitrate of 1 Mbps, then 480p or something in between might be a better choice than 720p for most videos. I'd go for 960x540.

But as Derek said, 480p is quite a low resolution, so you'll lose a lot of detail from the 1080p sources. Whether that matters to you or not, depends on the screen size and viewing distance.


+1 to using CRF for your encoding task.

Did you see the rule that Derek said?

nm
30th August 2011, 14:14
If I do tests by myself, I should run a lot, maybe hundreds of footages, 'cause movies got different types and details. For example, this footage fits 720*, but that one fits 1024*. I need an average resolution for all the movies. So it's unlikely to get the result on this way.

But it's a good way to get a result that pleases you instead of whoever is giving the advice.

Did you see the rule that Derek said?

Yes, and what about that? I prefer to keep a slightly higher resolution even if there aren't that many bits to spend. Just denoise the grain out if you decide to go for 960x540 @ 1 Mbps.

However, I'm starting to wonder if you have missed some of the more valuable advice given in this thread. Did you see this later comment:


@lever1119: There are no rules regarding resolutions and bitrates.

hello_hello
30th August 2011, 14:49
If I do tests by myself, I should run a lot, maybe hundreds of footages, 'cause movies got different types and details. For example, this footage fits 720*, but that one fits 1024*. I need an average resolution for all the movies. So it's unlikely to get the result on this way.

Hundreds of footages is probably somewhat of an exaggeration, but I think you'll find that's how most of the regulars here decided on the settings/resolution/bitrates they use.... through having run their own tests.

I'd still be starting with the minimum resolution you'll be happy with and running CRF encoding. By the time MeGUI's been running for five or ten minutes you'll have a pretty good idea what the file size will be. If it's too low, increase the resolution or decrease the CRF value and start again. Even if it takes a few tries till you're happy it'll still be quicker than running two full passes and at least you'll know what the final quality will be.

There's not really any "one size fits all" resolution or bitrate. That's why most of us use CRF single pass encoding.
The 700MB to 1.5GB encodes on the net which you referred to all vary dramatically in quality.

Target Practice
30th August 2011, 14:52
If I do tests by myself, I should run a lot, maybe hundreds of footages, 'cause movies got different types and details. For example, this footage fits 720*, but that one fits 1024*. I need an average resolution for all the movies. So it's unlikely to get the result on this way.that's another reason to test yourself, as I (and everyone else that posted) probably doesn't have every title in your collection.

if I did, and I found the resolution that got best results to my eye, it would be only my eye's opinion. its yours that counts.

lever1119
30th August 2011, 15:32
If you have to try everything by yourself, you guys don't need to come here for help.
You guys have no idea about it. I need professional advice.

lever1119
30th August 2011, 16:36
I think there should be a rule for resolutions and bitrates. Or next time I give 2M, 3M....bitrates for videoes, what should I do? Test again and again?

nurbs
30th August 2011, 17:29
Wouldn't it be great if it was so simple? Unfortunatly there isn't just on variable, the resolution, there are many. The video quality at a given bitrate depends on the video compression standard, the encoder, the settings you use as well as the framerate, resolution and properties (complexity) of your source video. Then there is the fact that the relation between the quality and bitrate for framerate and resolution isn't linear, e.g. twice the pixels require less than twice the bitrate for similar quality.

I didn't see you mention why you need the files to be between 700 MB and 1.5 GB, because that limitation makes things much harder for you.

If that's not a hard limit but just a range you want the majority of your encodes to fall in you can still use CRF encoding and not worry about bitrates too much. You just need to do some test encodes on parts of a source at different resolutions (say 960*x and 720*x or whatever) and with different CRF values and then find the smallest encode that still looks acceptable to you. Then confirm that by testing some other sources and you are set. You don't even need to encode the whole movies, just parts with SelectRangeEvery(). You can do that in a couple of hours.

lever1119
30th August 2011, 23:42
Wouldn't it be great if it was so simple? Unfortunatly there isn't just on variable, the resolution, there are many. The video quality at a given bitrate depends on the video compression standard, the encoder, the settings you use as well as the framerate, resolution and properties (complexity) of your source video. Then there is the fact that the relation between the quality and bitrate for framerate and resolution isn't linear, e.g. twice the pixels require less than twice the bitrate for similar quality.


Actually I know CRF very much. Even if I use CRF, I have to choose a resolution for all my videos. Resolutions on CRF value is just like resolutions on bitrates. You understand?

hello_hello
31st August 2011, 05:00
Actually I know CRF very much. Even if I use CRF, I have to choose a resolution for all my videos. Resolutions on CRF value is just like resolutions on bitrates. You understand?

Yes. You want the "magic" answer. There isn't one.
The suggested bitrates in post #12 are probably about as close as you'll get.