I am using Adobe Premiere CS3 and have a 10 minute+ video that I need to compress for the web. Our development team has our own flash player, but the compression options I have in Premiere are creating a file that is 75 Mbs or so. I know that this can be made significantly smaller without a huge reduction in quality. I need to know if there are settings in Premiere that I am not using that can reduce the size, or if there is a better compressor than Sorenson Spark or On2 VP6 for Flash.
Currently my settings are high:
Bitrate: 768 kbps
Bitrate: 192 kbps
On the low all the settings are the same but:
Bitrate: 265 kbps
Any information on this topic would be much appreciated. Also, open source would be preferred, or a guaranteed compressor at a low cost. Thank you.
+ Reply to Thread
Results 1 to 30 of 49
See why the difference is "Pure"
Your audio bitrate is probably a bit high. 128 should be plenty for flash. Also, Sorensen isn't very good looking at low bitrates.
Run some noise reduction on your source file and encode 2 pass. That should reduce the output file.
H264 is now permissible for flash, as long as you have flash 9 installed.
Check out the examples on this site. If you can run command line executables, you'll be able to compress the material quite nicely. There's a drop down box on the left side of the screen for more examples.
I agree with Soopa -
h.264/aac in .mp4 clearly beats VP6, which clearly beats h.263 (at any given bitrate)
You can use very low bitrates for audio with Nero AAC HEPS, even 16kbps sounds about the same as MP3 at 64kbps
Many free open source front ends (GUIs) to choose from, Ripbot264, MeGUI, Staxrip, Xvid4PSP, AutoMKV...many others
I rendered the video and it looks great, and it rendered the audio and video separately. I need to know how i can compile them together because our own video player (dev in house) cannot sync them properly alone. Thank you.
The video looks amazing though, thanks a ton.See why the difference is "Pure"
The example video gave me some great quality settings, but unfortunately those settings in Premieres Media Encoder produce (Format: H264) two files, a .m4v (video) and .aac (audio). If I go to Quicktime Format and choose the h264, it produces a video fine, but the size is too large again (even at 13 seconds, it was at 8 mbs). Is there another encoder that will give me the options to produce a .mov or .flv using H264 encoding, or am I going about this the wrong way.
The player in that example URL is great, but it can only play .mov and .flv as it states, but the examples themselves are .mp4. I totally confused this.
Basically i need to know how I can produce H264 formatted video (because of compression and quality), while still producing something that either our inhouse player can play, or the player in the above URL from Soopa, can play. Thanks, this is time critical as well.See why the difference is "Pure"
You can use YAMB which is the GUI front end for mp4box to mux together your elementary video & audio streams. If they are out of sync, you can specify a media delay in the options
Adobe suggests you use .mp4, not flv for these types of files.
That player is JW media player, and will accept h.264/aac in .mp4 containers. Adobe Flash 9 update 3 is required.
You can find more info here
It looks like Adobe Media Encoder has a .mp4 multiplexer for the h264 format. The player in the above URL can play it no problem, so that may work. I rendered a 13 second portion of the video and it is coming up as 4 mbs. I'll post what the full video is in just a second.See why the difference is "Pure"
Well when I take these settings:
766*575 (for our playable area)
VBR 2 Pass
Target Bitrate: 2537 kbps (in Premiere its 2.537 Mbps)
Max Bitrate: 5075 kbps (5.075 Mbps)
Bitrate: 92 kbps
With that I get a file size of 198 MBs, which as we all know is way too big to post. That example page has for instance the Indiana Jones trailer at those bitrates and still ends up with a 1.71 Mb file. If I lower my bitrates to a 10th of that, then the file drops to 26 Mbs, more managable. Let me know if I'm reading that example incorrectly, or if they are doing something else that I didnt state in my settings. ThanksSee why the difference is "Pure"
Filesize = running time x bitrate. The only factor you can change is bitrate in your example, unless you want to cut bits out of your video
You could probably use even lower bitrate for audio depending on what it is and quality of the source (speech audio vs. concert music etc.) I bet you are using AAC LC, not AAC HC or even HEPS.
Your could lower the video bitrate for that frame size, depending on the quality of your source and what it is (e.g. anime, vs. detailed closeups, vs. action sequences - they have different bitrate requirements), and what end quality you are going after. Quality is completely subjective. What looks good to me might look like crap to you. I would try dropping the bitrate on the video for a sample test to see if its ok for you (i.e. test a small 30 sec clip to save time). You can use lower bitrates if you reduce the frame size dimensions. It is highly dependent on souce quality, so we couldn't advise on specific settings without seeing a sample source clip.
If you decimate frames 29.97=>23.976 you could save some more bitrate as well
I understand that you are under some sort of deadline because this is work related, so this might not be advice you can use. (you might not have enough time to play around with these programs)
Personally, I would export my edited piece as uncompressed or use a lossless codec then use one of the open source encoders that uses x264 because they have significant quality enhancements in quality and speed over retail AVC offerings, which means significantly smaller file size than something you'd get out of Premiere. The settings are highly customizable, and fortunately there are presets for many of the front end programs which makes life easier
766*575 (for our playable area)
That's a really inefficient dimension to convert a file to. I'd make sure to resize to any dimensions divisible by 4, preferably 8.
768x576 would be better.
At that size, you should get some nice quality by encoding to 2000kbs max - 2 pass vbr
Here's a file encoded at 1800kbs 848x480 dimensions. Looks pretty good.
If I'm not mistaken you want best quality at LOW bitrates ~300-400kbps , that is in a "flash streamable" format?
As a general rule, the best very low bitrate encodes are achieved with the highest quality settings (very CPU intensive), and you can also do post-processing/filtering to achieve slightly better results.
Also you should use mod16 resolutions, for maximum compressibility.
Another trick is to use zones for encoding: basically you encode non complex, non critical scenes at lower bitrates (higher quantizer) - like film credits, simple scenes - and use higher bitrates (lower quantizer) for more complex scenes. This way you can reduce the overall bitrate and thus filesize.
I'm going to recommend you ditch the Adobe h264 Encoder, because of the poorer quality (it's based on Mainconcept's engine if I'm not mistaken). You cannot configure very many settings, and as such you have to compensate with a higher bitrate (therefore ending up with a larger filesize). There have been significant advances in the open source x.264 CLI version (such as adaptive quantization, psy-rdo), that other h264 versions lack.
The following examples are encoded from the publicly available 1080p Apple Movie Trailer "Wanted." to 640x368, with video bitrates ~365kbps, audio bitrate 16kbps.
Notice the difference in audio quality from stereo 16kbps AAC HEPS compared to MP3 (which you can only encode mono with low sample rates at that bitrate) - this translates to bitrate savings and smaller end filesize than if you had used the regular "flv" h263 Sorensen Spark with MP3 audio. The screencaps are taken at a P frame where the scene changes, you get pixellation at these points with most codecs that do not have internal deblocking. If you use mediainfo (view=>text) you can see what settings were used on the x.264 videos. The "tweaked" video also used 1 avisynth filter, LimitedSharpenFaster()
1) "Regular" FLV 365kbps video
2) x264 Unoptimized 365kbps video - Standard settings AVC Main Level 5.1 - no enhancements like AQ and psy-rdo or tweaks. This is comparable to what you'd get out of Nero AVC or Mainconcept AVC
3) x264 Tweaked 365kbps video
DOWNLOADS (having problems uploading directly to site...)
Each video download is about 13 seconds and <1MB
640x368, ~365kbps video bitrate + 16kbps audio bitrate
02 x264 unoptimized
03 x264 tweaked
I appreciate the information guys. I'm not savvy with command line programs, but I found another program , meGUI, that was suggested to this end. So I'll render this out at uncompressed and see how this works. I'll come back with my results and hopefully a page to visit. Thanks again.See why the difference is "Pure"
Hi Pureweb, I hate using command lines too. I use MeGUI also for all x264 encoding. It's a bit difficult to configure at first (specifying program paths and such), but once you get it working it's great. I suggest you use configure the settings to use the development servers(options=>settings=>extra config=>configure servers=>select development in the drop down menu). There are presets for x264 settings that you can use that simplify the process as well.
Okay, the render time for an uncompressed was ridiculous and Premiere crashed at the end (after 8 hours). Is there a way I can use the other computers in the office as a satellite renderers for Premiere. This will continue to be a problem if I can't solve this. Thank you.See why the difference is "Pure"
I haven't used Premiere for years, so I hope someone else can chime in. I'm guessing something else is causing the problem - I don't think it should take that long on a quad - especially for a 10 min video. Exporting as uncompressed should take less time than encoding it to AVC as you did earlier. It might be a bad setting or some incorrect option selected...
You might be able to export using a lossless codec (such as lagarith or huffyuv - although I heard there were quality problems with huffyuv and Premiere specifically)
As for network rendering over a LAN , I would love to learn about that and how to set it up too.
Since you are on a timeline, you might have to settle toying with Premiere's/Mainconcept's AVC settings, but it's still miles ahead of Sorensen or VP6.
I ended up using the ME in Adobe because I didnt want to risk an overnighter that just fails. I think I'm going to try to create it again now that I have viable copy for the web. The ME encoder definitely lack settings and the intelligence to compress it as well as what I've seen using x264. Thanks for the advice, I will be back Monday with a report on using the method you all told me, until then...See why the difference is "Pure"
See why the difference is "Pure"
Can you post the full error message or logfile?
Did create an .avs file to "frameserve" into MeGUI? You can use the Avisynth script creator in the tools if you want
MeGUI is actually not the easiest to learn at first or setup the file paths, etc...
There are a few guides to help if you push the "guides" button when you look under MeGUI in videohelp. Here is one to get you started http://www.digital-digest.com/articles/MeGUI_H.264_Conversion_Guide_page1.html
The easiest x264 front end to use would probably be xvid4psp (don't let the name fool you, it works very well). It contains most of the files when you do the install, so less fiddling around trying to configure it. Other easy to use GUIs would be ripbot264, or staxrip
I forgot to install an AVIsynth and now its working, I am trying to find setting that will produce a good qaulity (off of HQ-Slow preset). This is creating pretty good quality at only 1.03 for 23 seconds of video, but there is some artifacting at the last bit where the logo comes up.See why the difference is "Pure"
It would be hard to say what's causing it without seeing it, but if this is for work/client then you probably can't post it.
I can tell you the default MeGUI profiles have deblocking set sharp (-1,-1) instead of the default (0,0). This can cause ringing sometimes, more common when you are doing low bitrate encodes. One of the authors (Sharktooth) prefers sharp, but I usually set it back to (0,0) (Hit config, Look in the deblocking section under strength and threshold)
I can also tell you are using the older presets from the regular branch of MeGUI, because "HQ-Slow" no longer exists. This should be fine, but if you wanted to test out the newer versions you have to configure the update servers to the development branch (see a few posts up). IIRC, the older profiles had (-2,-1) set as default, which might explain the artifacting
Originally Posted by lordsmurf
I'm sure Adobe Encoder (Mainconcept h264) works fine and is definitely a better choice than Sorensen or VP6; but I think this thread is about getting the best quality results (at low bitrates). This is not necessarily about "tweaking options." There is a substantial quality difference between x264 and other AVC encoders. Even using lower bitrate and lower quality settings with x264 often yields better results than other AVC encoders. Why?
Mainconcept's h264 inplementation is inferior to x264. It's quality has charateristics shared by most "brands" of early h264 implementations: inloop deblocking. This is the single biggest factor that makes it and all AVC encoders so great for low bitrates compared to other codecs such as DivX, XviD, MPEG2, VP6, Sorensen (h.263). However the major criticism is that it often has a "smoothed over" or "plastic doll" look; images lack detail unless you use throw more bitrate at it. The same can be said about Ateme, Nero AVC, even Real Video RV10 (which is based on h264). This is where x264 development diverges and why it's superior.
x264 has very active development, and is updated probably a few times a week. While most of the patches are minor, some major psy enhancements recently have brought up quality - especially with respect to detail and grain retention. This is why I recommended using the development servers for MeGUI's autoupdate. The regular branch of MeGUI - as well as many other x264 GUI's - are often using older versions of x264 that do not feature the recent improvements.
The following examples are encoded from:
1&2) 1080p Apple Movie Trailer "300" => 720x304 ~750kbps. Note "300" is supposed to be a "grainy source" so I've included the original resized for comparison. It's not "noise" that you are seeing with x264 - it is grain/detail preservation and supposed to look like that. Look at the differences in detail with the wall texture and the "monster's" chest
3) 1080p Apple Movie Trailer => 720x304 ~750kbps. An example of a "Western". Please compare the details in the face, beard and hat
4) Big Buck Bunny as a CGI example 1080p => 720x400 @~500kbps. Note the "smoothed over" look for Mainconcept vs. detail retention in the fur for x264
-Mainconcept settings were tweaked and used 2-pass "High" @L4.1 profile (as opposed to the lower quality "Main" profile that the OP used earlier).
-The x264 encodes are using r938 and are encoded using the standard 2-pass HQ (High@L3.1) profile. The only thing changed was deblocking from (-1,-1) to (0,0) (same as Mainconcept's). NO tweaks or filters were used for x264, they were encoded "as is," with default settings.
Please save the images to your desktop and "flip back and forth" with Windows Picture viewer. Examine the detail preservation at equivalent bitrate with "unoptimized" settings. The difference isn't with only 1 or 2 frames, or with only 1 genre of video. When every frame is missing information and "smoothed over," the viewing experience is inferior. While Mainconcept h264 is still way ahead of Sorensen or XviD, the quality is inferior to x264. Don't take my word for it - please do some testing for yourself, then add adjust some settings to get even better encodes. Note x264 has more settings that you could tweak compared to Mainconcept's h264 implementation, should you choose to optimize your encodes. Why pay $ for worse results when a free open source alterative is clearly better?
My understanding is that x264 does not integrate with Premiere, because x264 is freeware licensed under the GNU public license.
You have to export as uncompressed or use a lossless codec, then encode with x264 (again I suggest using MeGUI with the development servers, because it always has the most recent revisions)
Unfortunately there is no freeware equivalent NLE to Vegas or Premiere
Nice comparison study, poisondeathray. It's quite obvious that X264 is retaining the sharpness where Mainconcept isn't.
Originally Posted by poisondeathray
I'm also facing the same problem of publishing video at portfolio site with the good quality/size for on-line view...
Another question is if there is a key in x264 to move atom to the beginning of the file so it could be streamed before downloading?
p/s/ I'm using http://megui.org/auto/ development server
I have tried Unrestricted 2 pas HQ 750kbs with Nero AAC from .mp4 PAL 25fps file and have a audio/video unsinc.
Does anyone have any clue?
For h264 flash, I think it is more a function of the software and server settings for streaming immediately (e.g. for the JW player the default for "autostart" is "false")
Unrestricted 2passHQ would be close, but I adjusted a few other settings, e.g. psy-trellis (which is off by default), increased the ME range, analyze all and a few other things.
It depends on your source and what your goals are. e.g. you can probably get 10-20% better compression if you turn off psy-rdo/psy-trellis, and have higher deblock alpha & beta values.
For sync issues, you have to provide more information: it is it a constant sync issue or a progressively worsening issue? (the former can be easily fixed by entering a delay)
What is your original source material? Was it decrypted properly?
Regarding fast start, it's not a properties of the flash player or flash itself. Have a look at this link:
there is a software which should move atom (information about video) from the end of the file to the beginning. Otherwise streaming will not start unless whole file is loaded.
Could you give values you have used for psy-trellis and ME range? Just want to experiment a bit.
What do you mean by "analyze all"?
Regarding sync, it's progressively worsening one. Input file is produced by nero h264 codec grabbing from dvd.
Plays fine in wmp classic.
How do i know if it was decrypted properly? Video seems to be fine, but i hear no audio during playing it in preview.
and video is rather slide-show
I have just finished encoding from debugmode frameserver using same audio codec and it's perfectly ok.
Seems that out of sync problems is due to pre-encoding by nero...
I think the moov atom only concerns QT, not adobe flash (i.e. h264/aac in .mp4). I know if you toggle autostart on in the JW player, it will play immediately (you can test it yourself). You can play around with the settings and test it with this setup wizard
Here are some examples that stream immediately (use the drop down menu for more)
--partitions all --merange 32 --psy-rd 1.0:1.0
When you experiment you will find different settings work better for different genres of source material, and there is a tradeoff if you want to preserve more detail or want better compression/smaller size (tradeoff). e.g. increasing the psy settings will enhance details, but will require more bitrate
Note that x264 is constantly being improved; for example, when these were done, it was using the old b-frame decision; things have changed recently with a new b-frame decision (--b-adapt 2). Subpixel refinement has also changed, there is a "9" setting now.
Regarding sync issues, you need to use AnyDVD or DVDFab HD Decrypter if it is from a recent DVD. If you have a VOB source, then did you use DGIndex to index it?
I don't understand what you mean when you say using "nero h264 codec grabbing from dvd" ? Did you use nero recode? for audio & video? I'm very confused about your process or what software you are using?