You're correct. Following the "downmix to stereo" DSP with the matrix mixer makes no sense, because it'd be stereo before it hits the matrix mixer and the audio would be passed through untouched. I'm pretty sure the "downmix to stereo" DSP includes the LFE channel, which my matrix mixer settings don't, and I kind of remember it doesn't reduce the volume of the surround channels by 3dB as my matrix mixer settings do, and unless it's changed since the last time I used it, the "Downmix to stereo" DSP just downmixes. It doesn't have a normalise option as the matrix mixer does so there's a fair chance of peaks above 0dB when the channels are combined. Try downmixing with nothing else other than the "Downmix to stereo" DSP, and then again with nothing but my matrix mixer settings ("normalise" checked). I think you'll find there's a fair volume difference. How the vlevel DSP handles peaks above 0dB, I don't know as I've never used it. I might try it later, but it seems odd that you need to boost the volume by +4dB after downmixing without reducing the volume to prevent clipping (normalising as per the matrix mixer), followed by a DSP that "compresses" by increasing the volume of the quiet parts.
Which AAC encoder are you using? And why constant bitrate? They all accept a 32 bit float input but FDKAAC converts to 16 bit integer internally before encoding so peaks above 0dB will be clipped. QAAC encodes peaks above 0dB. The others accept 32 bit float input but I can't remember what they do with it.
I've no idea what the fading in and out problem is. My best guesses would be you have the "don't reset DSPs between tracks" option checked under Processing in the converter setup, or vlevel isn't very clever at handling the initial part or end part of the audio, although I'm not sure either would cause the problem you described. I might try it later.
There's an option in the encoder configuration labelled "do not convert in multiple threads" When it's checked, you can tell fb2k to convert multiple files, but it'll only convert them one at a time. See if that makes a difference.
If RockSteady was compressing too much, you could always reduce the compression. Aside from applying a -6dB limiter, the WinAmp bridge itself should have no effect on compression, although if you're downmixing with the "downmix to stereo" DSP, maybe the WinAmp bridge was limiting the peaks that would otherwise have exceeded -6dB, and possibly exceeded 0dB, and so you were effectively limiting before compressing with RockSteady.
+ Reply to Thread
Results 61 to 90 of 121
Last edited by hello_hello; 3rd Nov 2018 at 01:15.
I gave vlevel a spin. With the strength set to 0.2 it doesn't compress all that much. The defaults seem to be roughly on a par with the DAN.
Mind you, now I think I'm experiencing a weird fade-in problem with the DAN. Which version of fb2k are you using? For me it only seems to be effecting the DAN (so far). When I get motivated I'll try 1.4.1 beta2. I think that's the version I was using when I uploaded the samples in post #50, and those samples don't sound the same (no fade-in) I'm using 1.4.1 beta5 at the moment.
After converting to wave and then adjusting the volume to -23 LUFS while encoding to AAC.
(Source file: "Bride Of Chucky (1998) AC-3 6 Channel.wav" from post #49)
vlevel Strength 0.2 -23.04 LUFS, Peak -2.02 dB
vlevel Strength 0.8 -23.04 LUFS, Peak -6.81 dB
Edit: Well now I'm confused. Surely I couldn't have done something silly and uploaded all the wrong encodes for the DAN samples, but attached is a sample of what I keep getting today, and I've tried fb2k 1.4, 1.4.1 beta2 and 1.4.1 beta 5. I'm sure I was using beta2 when I uploaded the earlier samples. Is the fade in for the DAN sample here the sort of thing you're experiencing?
Last edited by hello_hello; 31st Oct 2018 at 14:34.
if possible would you be able to upload a video as this is all confusing to me and i don't think i'm doing things correctly, and i apreciate all your time posting replies and helping out
when i add all my extracted files and put them onto foobar i scan all items as album the apply the tags then when i convert i have the downmix to stereo followed by the winamp dsp and i have attached some pictures so you can get an idea what i'm doing or if i'm doing it right or wrong
[Attachment 47072 - Click to enlarge]
[Attachment 47073 - Click to enlarge]
[Attachment 47074 - Click to enlarge]
[Attachment 47075 - Click to enlarge]
[Attachment 47076 - Click to enlarge]
Last edited by circulationds; 1st Nov 2018 at 06:20.
Fade ins/outs happens at start or end of audio with dynaudnorm filter. Use b=1 (alternate boundary mode) if you do not want fades.
You can also use DynamicAudioNormalizer standalone application, it have same option, -b switch.
I know where i have been going wrong i missed out the part where you detail to With RG Info" preamp to -6dB, movies are perfect now dialogue is at same level as loud action scenes, sorry for all the messages, but also like to thank you for your time and patients, i have just converted ant man the dialogue was really low on that and it sounds amazing now, again thanks for your time and direction of usage on foobar, have converted 10 movie files and they all sound the same very happy with the outcome, i can't thank you enough, i'm staying away from dynaudnorm and loudnorm from now on
Last edited by circulationds; 1st Nov 2018 at 10:40.
I've never used the -b switch with the DAN. I still think it's odd the same sample didn't appear to need it originally, then a couple of days later it did. Maybe it was user error, although I've no idea how I could've gone wrong. I'll try again with the -b switch later.
thoughts regarding your screenshots.....
You should use Track Gain (and a Track Gain scan) rather than Album Gain. Album Gain is designed for scanning a group of files as an Album, so when you apply ReplayGain each track is adjusted by the same amount. For an Album, their relative volumes remain the same. To adjust each track to the same volume, use Track Gain (although Album Gain will probably do the same thing as long as you only scan a single file at a time).
Normally you do want to reset DSPs between tracks, so uncheck that option (it was possibly contributing to the fade-in, fade-out problem).
Don't forget you can't adjust the volume to a particular ReplayGain volume while converting and compressing at the same time. The ReplayGain option adjusts the volume before it's compressed. If you want to compress and adjust the output files to the same volume, it needs to be a 2 step process.
Downmix, compress and output to a lossless format such as a wave file. Scan the output wave files with TrackGain. If any have peaks of 1.00000 (100%) or greater they could be clipped, but assuming they don't.... save the ReplayGain data to the wave files, then convert them to the final format while adjusting the volume with the ReplayGain option. Chances are, after compressing they'll all be a quite similar volume and the ReplayGain adjustment won't be necessary, assuming you're only doing it to make them the same volume. If it's unrelated audio, such as different movies, I wouldn't bother with the ReplayGain step.
I'd use my matrix mixer settings for downmixing rather than the fb2k dsp.
For QAAC, use this custom command line instead of the fb2k preset:
--ignorelength -s --no-optimize --no-delay -V 91 -o %d -
Same thing as the fb2k preset you're using, except with the --no-delay option added. Why.....?
All lossy encoders add padding (silence) to the beginning and end of the audio (something like 50ms for AAC). The amount of padding is saved when the file is encoded, but that info can be lost when muxing (although MKVToolNixGUI does account for it when muxing AAC, if the info exists). The --no-delay option tells QAAC to remove the initial "padding". For soundtrack audio it means the audio/video sync can't be effected. Don't use --no-delay for encoding CD tracks (for various reasons), but it's a good idea for soundtrack audio.
RockSteady only accepts a 16 bit input. If you select 32 bit in the WinAmp Bridge DSP, as per your screenshot, you'll either have no audio or it'll be passed through uncompressed (I can't remember which). RockSteady doesn't support multichannel audio either. It's probably also a good idea to enable the RockSteady "Use Smart Limit" option.
16 bit means 16 bit integer, much like a 16 bit wave file. Anything above 0dB gets clipped.
24 bit means 24 bit integer, much like a 24 bit wave file. Anything above 0dB gets clipped
32 bit means 32 bit floating point. Floating point can store values above and below 0dB. No clipping.
For lossy source files, fb2k decodes to 32 bit float. All fb2k DSPs I'm aware of process the audio as 32 bit float, so they won't cause the peaks to be clipped even if they end up above 0dB.
Whenever 32 bit float is converted to 16 or 24 bit integer, peaks above 0dB will be clipped. RockSteady only accepts a 16 bit input, which is why the Winamp DSP bridge includes a peak limiter. I think most Winamp 2.0 plugins are 16 bit.
That's the ReplayGain option for adjusting MP3 and AAC audio losslessly, as opposed to adjusting the volume when converting. It works with MP3 and AAC in an MP4/M4A or MKV container, even if it also contains video, or with standard MP3s.
Scan the file with TrackGain, save the ReplayGain info, then right click and select "ReplayGain/Apply Track Gain to File Content". The volume has to be adjusted in increments of 1.5dB, so it'll be adjusted losslessly to the nearest increment.
That's a way to avoid intermediate wave files after compressing if you also want to adjust the volume with ReplayGain. You could downmix/compress and convert to AAC, then scan the AAC files, save the ReplayGain info and adjust their volumes losslessly.
The TrackGain results show the individual volumes, relative to ReplayGain's target volume. That's 89dB for ReplayGain (long story, but 89dB is actually a sound pressure level, which is why it doesn't seem to make much sense), so in more understandable terms the ReplayGain volume is -18 LUFS.
LUFS is the term invented for R128 scanning because the audio is kind-of scanned in sections (much like the DAN or RockSteady window size) and apparently an additional level of confusion was required. It stands for "loudness units relative to full scale". LU can be considered the same thing as dB. So a volume of -18 LUFS is the same as -18 dB.
So when the fb2k TrackGain scan shows +5.44 dB it means the volume needs to be increased by 5.44dB to achieve -18 LUFS, which means the volume is -23.44 LUFS. Mental, isn't it?
The peak values are dBTP (dB True Peak). Just the loudest peak for a group of files scanned as an album, or the loudest peak for an individual file when it's Track Peak. Zero dB is still "maximum", so +10 dBTP is a fair amount above maximum. QAAC will encode the peaks if they weren't clipped on the way to the encoder, but you really want peaks of 0dB or lower.
If you're interested, attached are instructions in picture form for adding some new fb2k Playlist columns. Hopefully they'll be fairly self explanatory. I've called them "Volume", "Peak" and "Track Gain". They display the Track Gain volume in dB (same as LUFS) instead of ReplayGain's 89dB (as it's somewhat meaningless), the Track Peak in dB, and ReplayGain's Track Gain (the adjustment required to achieve the ReplayGain target volume). You'll need the latest 1.4.1 beta of fb2k and not the stable 1.4 version for the "Peak" column, as version 1.4 doesn't understand the "replay gain peak dB" field. It's more convenient being able to save the ReplayGain volumes and see them in a playlist. They only display Track Gain (not Album Gain).
Text for the "Volume" column (credit goes to Case):
Screenshot 2: In the configuration window, add the new Volume, Peak and Track Gain columns.
Screenshot 3: Right click on a playlist column header and select each newly added column to enable it.
Screenshot 4: Shows the new columns displaying the volume in dB, the Peak in dB, and the TrackGain adjustment required to achieve the Replaygain target volume. I had to scroll right to see them, but you can resize and re-order the columns etc. If there's no Track Gain info, the columns will be empty.
Last edited by hello_hello; 15th Nov 2018 at 23:30.
Today I rediscovered the ability to duplicate the playlist view in fb2k. I have a bunch of tabs bottom/right, so I added another for a duplicate playlist view. It's handy because you can have the "playlist tabs" configured as you'd normally prefer when viewing a playlist, while the duplicate playlist view can have different columns and therefore display different info. If there's more than one playlist, it'll display the contents of whichever Playlist Tab is selected. I'm using it as a way to easily see the results of the ReplayGain scan after I've saved the info to tags.
I thought I'd mention it in case you want to experiment with layout editing at some stage.
Last edited by hello_hello; 4th Nov 2018 at 08:51.
Hey hello_hello was away to see family in scotland for a few days, thanks for posting all the above information in regards to my steps, i have managed to find a good setting that seems to be doing the trick. been using the QAAC as you guided along with the parametres you posted as well, i did come accross a couple of issues with the matrix mixer though? for some reason it was not lowering the volume on some video's. the video's where 6 channel which i'm not entirely sure, i have used the default downmix to 2 channel stereo, and reduced the track gain by -6 db so that each file is roughly 83db after the first converting process, then i rescan them converted files and use the rocksteady plug in see picture of my new settings for the rocksteady dsp i used my headphones to do this and just tweaked it so that there was no clipping that i could hear through my headphones,
[Attachment 47121 - Click to enlarge]
I know you said something a long the lines about the default downmix to 2 channel, left one of the channels, but i can't tell if there is anything wrong? if this works out the way it should i think the files will be near 84 to 85db, i'm not noticing any clipping yet, i did with 47 ronin and a night at the roxbury, thats when i used the little tweak settings for the rocksteady dsp and it seems to be working ok, if there is any clipping i can't hear it. i like what you have done with the duplicate layout! that is so handy maybe when you get time you can show me how to do that, after i have done my video's i will be moving on to my music files and getting them all sorted so this layout editing really is of interest to me as i did use mp3 gain a lot but like i said in previous posts i did not know what clipping meant when i was working on mp3 files using the mp3gain, luckily i don't have 100's of clipped music files, and the ones i did use the gain on i still have the cd's for them, so i can rip them again on pc and use foobar this time
Last edited by circulationds; 8th Nov 2018 at 14:10.
For a 5.1ch downmix, make sure you have both the "back" and "side" channels included in the downmix, as per this screenshot as the "surround" channels could be in the side or back channels.
To briefly explain.... formats such as AC3 and DTS have surround channels, but there's no surround channels in the wave file layout. The wave file channel order, which is the same as the matrix mixer channel order, is here. To make it worse, different lossy codecs encode multichannel audio using different channel orders.
These days it's nothing to worry about though. When lossy audio is decoded it's decoded to match the wave file layout, and when a lossy encoder is fed multichannel audio in the wave file channel order, it gets remapped by the encoder and encoded in the correct order. For 5.1ch audio though, sometimes the "surround" channels are decoded to the wave file "back" channels and sometimes to the wave file "side" channels, which is why you need to include both when downmixing, so you don't need to worry about where the surround channels are being decoded to each time.
The matrix mixer "normalise" option is clever, in that it decreases the volume when combining channels so there's no chance of clipping, but it's also dependant on whether the specified channels actually exist in the source. The upshot is, if you have it configured to downmix 5.1ch to stereo, it'll reduce the volume a fair bit for a 5.1ch source, but even when it's configured to downmix to stereo, it won't reduce the volume at all if the source is stereo because there's no channels being combined.
For the layout editing....
If it's not shown by default, right click on the menu bar and check "layout editing mode toggle". The layout editing button should appear on the toolbar. Click on it to switch to layout editing mode. In layout editing mode, right clicking performs layout editing instead of the usual functions.
The fb2k GUI is a combination of "UI elements". You can resize and replace them etc. There's a tabbed UI element and a splitter element included. The tabbed UI element lets you add multiple UI elements to the same "space" and switch between them using tabs. The splitter UI element lets you split a "space" into sections and add a UI element to each. When you right click on an existing UI element, the existing element name is show at the top of the menu.
My layout is fairly basic in that it mostly uses the built-in fb2k elements. Columns UI is quite popular and replaces the default fb2k interface completely and makes it easy to apply skins and view artwork (apparently, I've never used it).
Below are a few screenshots of my setup in layout editing mode to give you an idea. The right side is split vertically and the main element in the bottom half is the "Tabs" element. The second screenshot shows that under the "Track Volume" tab (I renamed it) is the "Playlist view" element. The Equaliser tab is split into two elements with the Equaliser element on the left and the Peak Meter element on the right. It's pretty versatile and everyone has their own preferred setup. Play around and see how you go.
Last edited by hello_hello; 20th Nov 2018 at 04:32.
User tried to use alternate boundary mode for dynaudnorm and failed.
I do still find this 'fade in/out' issue quite intriguing. Me thinking out-loud > Is it just something that's a 'one off', a rare event? Something a little more common? Maybe the current dynaudnorm default settings need a re-think? Is it just a case of a single parameter that needs changing? Could it actually be a fault/bug? With the correct sample it may be worth raising the issue with the dynaudnorm author?
video.baba i'm not all clued up in this factor i can only explain to the best i can, but when i was experimenting with the box4 some movies where giving me hassle and i could not get the results i was wanting until i done the converting with pazera and then repacked the lower volume audio's back in with the settings as stated, some movies where perfect but some it didnt work as good.
At the end of the day i think it's all down to poor downmixed audio, but a good way to start is the volume of the actual source file, maybe changing some of the dan settings within the box4 program, either way we can't get perfection all the time and it's maybe something that will be better over time, with parameters and stuff maybe change the default window frame down a bit and the gaussian filter is too slow to respond on 31. But even using the dynaudnorm with foobar i was getting the same issue with the bride of chucky movie and a few others so it has nothing to do with the box4 program it's the dynaudnorm itself.
Maybe you can try and implament the rocksteady plug-in that's done a good job so far but i'm not 100% clued up like you guys on here but will definately give you my thoughts the best way i can, all i know is from hello_hello is the rocksteady dsp doesn't handle 5.1 audio files so has to be downmixed to stereo before the efects can work. there is also a matrix mixer with settings that hello_hello created to downmix the audio and normalise at the same time. I have managed to normalise all my audio with foobar so far but i'm keeping the normalised audio and default audio in my movie files as i did when i was using box4 program. i would however like to learn how to use dynaudnorm on it's own one day as i do enjoy a challenge but it's a bit frustrating that i don't have the full knowledge yet for doing this
Last edited by circulationds; 12th Nov 2018 at 02:52.
Last edited by circulationds; 13th Nov 2018 at 14:36.
Thanks for your very kind words, appreciate it very much.
For the files that foobar2000 won't decode.... the common types it won't decode are EAC3, AC3 and DTS and there's third party DSPs for decoding AC3 and DTS, which I assume you have. For EAC3 you have to use the FFmpeg Decoder Wrapper. That'll let fb2k decode anything ffmpeg can decode. You have to know the type of audio you're wanting to decode to set it up, but for EAC3 it'd look something like the screenshot below. I use the same version of ffmpeg for encoding and decoding, so I just point it to ffmpeg in the Encoders folder, but you need to put ffprobe.exe there as well. See the 1st screenshot below.
-loglevel 0 -drc_scale 0
-drc_scale 0 prevents ffmpeg from applying any compression information that may be included with eac3/ac3 files, as it's decoding.
If the files you're having problems with aren't eac3 you'll have to use MediaInfo to determine the audio type.
By the way, I was adding the volume info to a different part of the GUI so I fiddled with the way my Playlist columns display a bit. The text copy I made as a backup is attached. It might give you some ideas if you decide to play around some more.
The "Title" column displays "file name" in brackets when there's no Title info in tags and the file name is being displayed instead (sometimes it's handy to be able to easily tell the difference). A playing track is highlighted (orange for me). You can choose the colours under "Default User Interface/Colors and Fonts". The "Codec" column displays the number of channels and the codec, with "pcm" for fixed point PCM and "float" for floating point PCM.
Oh, and you inspired me to add the peak mater to the right of all the tabs so now it's visible no matter which tab I select.
PS In case you haven't seen it yet, fb2k 1.4.1 stable has been released.
Screenshot 1 : Configuring the ffmpeg decoder wrapper for decoding eac3.
Screenshot 2: My latest Playlist columns configuration.
Edit 1: Replaced the attached "fb2k Playlist Columns" text file with a new version with simpler alignment for columns.
Edit 2: And replaced it again to fix a silly mistake.
Last edited by hello_hello; 16th Nov 2018 at 00:44.
Thanks for uploading the column text file, i will be adding them, i will download the new stable version of foobar 1.4.1. also thanks for letting me know about the decoding of EAC3, AC3, DTS files, they were the ones that foobar was not able to decode. It was only about 8 files so i just decoded them with pazerra, wasn't that bothered as it was only a few. I have to say that this foobar program is awsome and i love what you can add and view within files so you know what the audio files are and what they contain within them files, i do have the fmpeg Decoder Wrapper installed that was one of the first things i downloaded as advised from you at the beginning.
I have one more movie folder to do and the movies are complete, gonna mess round on foobar after that and see what else i can learn, at the beginning i was using it just to track scan then apply gain, i didnt even encode with it lol. But thanks to you i have learned a lot about peaks and clipping and what codecs to use and a better knowledge on audio editing
Thanks for helping me i should have my plex running tonight and can't wait to watch some movies on it. It won't be bride of chucky anyway, heard the audio of that too much lol
Last edited by circulationds; 16th Nov 2018 at 03:38.
After fiddling with what it displays, here's the setup I'm using for the "Now Playing" tab in my foobar2000 configuration if you want to try it. It's attached as a text file. You need the Text Display element.
If you add the Text Display element to the GUI somewhere, right click inside it, select "settings" and paste the text from the text file into it, it'll display some useful information regarding the file that's either playing or selected. Under the right click menu there's a "prefer playing track" option. When it's checked, it'll always display the info for any currently playing track, in preference to the selected track.
Depending on the font, you might have to manually adjust the spacing to get it to line up nicely (using tabs doesn't seem to work). I'm using Microsoft Sans Serif, size 9 (under the Layout & Style tab in the Text Display Settings). Anyway.... something to play with if you want to try it.