clever FFmpeg-GUI, small FFmpeg UI, fully portable, no installation.

[mannequin80]

I changed that because I noticed that in many movies with stereo audio, the two channels aren't balanced.

yes, for stereo audio it makes more sense as long as stereo and multichannel are properly detected and separately processed.

[mannequin80]

btw, i noticed that latest ffmpeg builds (at least starting with 8.x?) are having issues writing the finished file to the disk if dynaudnorm applied only to one channel and there's splitting and joining involved. something to do with how newer builds are handling the buffering? not sure. are you aware of this "issue" and what might be a possible workaround for that?

[ProWo]

btw, i noticed that latest ffmpeg builds (at least starting with 8.x?) are having issues writing the finished file to the disk if dynaudnorm applied only to one channel and there's splitting and joining involved. something to do with how newer builds are handling the buffering? not sure. are you aware of this "issue" and what might be a possible workaround for that?

To apply dynaudnorm only to one channel, you can use the h switch, instead of splitting/joining.
To apply to the FC channel only, append

:h=FC

after the other switches.

[tb52]

Hi. It's been a while. I hope you're doing well.

A new issue. I just updated both clever and ffmpeg... When I try to cut a mkv file [I don't see any option other than keyframe] as I have done many times, now when I go to preview the latest cut, the video playback goes to the end of the file and not to the cut. I tried it on 3 different files with similar results. The cut file is at the cut, but the preview is as I said. The indexing seems to be taking much longer too.

Windows 11 system

[ProWo]

A new issue. I just updated both clever and ffmpeg... When I try to cut a mkv file [I don't see any option other than keyframe] as I have done many times, now when I go to preview the latest cut, the video playback goes to the end of the file and not to the cut. I tried it on 3 different files with similar results. The cut file is at the cut, but the preview is as I said. The indexing seems to be taking much longer too.

Thx for the hint, you are right, the preview is broken.
I've fixed this now, please update.

Btw: The indexing has'nt changed. The key accurate cut is available for avc (h264) and hevc (h265) videos only.

[Findu]

@ProWo
3.5.3.02 feedback

Thanks for adding pixel format qsv. Depending on the source and codec, it can accelerate encoding up to 25% on my Core Ultra.

Nice to collect problematic batch tasks as yellow tasks after finish batch processing. Thanks.


1. I discoverd a strange behaviour. When you change the pixel format for any encoder, WHA decoding always change to qsv and does not remain unchainged.
That also happens when you switch to some none qsv encoders like DNxHR, av1-libaom, h264-amf, hevc-amf, hevc-nvenc, h266 (vvc). Some change to WHA decoding qsv when you switch direction up, some down!


2. I still can't use qsv HWA decoding with normal filters like resize or rotation etc. As it is expected to use the qsv filter vpp_qsv.

-vf scale=800:450
need to be replaced with
-vf vpp_qsv=w=800,vpp_qsv=h=450

Therefore I retested qsv filter with ffmpeg 8.1
For example scale does not longer slow down the encoding speed for the ARC1 series, compared to the normal filter and ffmpeg 8.0.

Performance test for scaling from 1080p h264 to 800x450 h264:

6,59x
ffmpeg -i Input1.mp4 -map 0:0 -c:v h264 -global_quality 25 Output.mp4

17,9x
ffmpeg -i Input1.mp4 -map 0:0 -c:v h264_qsv -global_quality 25 Output.mp4

19,8x
ffmpeg -hwaccel qsv -hwaccel_output_format qsv -i Input1.mp4 -map 0:0 -c:v h264_qsv -global_quality 25 Output.mp4

---

22,8x
ffmpeg -i Input1.mp4 -vf scale=800:450 -map 0:0 -c:v h264 -global_quality 25 Output.mp4

26,0x
ffmpeg -i Input1.mp4 -vf scale=800:450 -map 0:0 -c:v h264_qsv -global_quality 25 Output.mp4

36,4x
ffmpeg -hwaccel qsv -hwaccel_output_format qsv -i Input1.mp4 -vf vpp_qsv=w=800,vpp_qsv=h=450 -map 0:0 -c:v h264_qsv -global_quality 25 Output.mp4

Here we see a nice increase in performance!

[tb52]

A new issue. I just updated both clever and ffmpeg... When I try to cut a mkv file [I don't see any option other than keyframe] as I have done many times, now when I go to preview the latest cut, the video playback goes to the end of the file and not to the cut. I tried it on 3 different files with similar results. The cut file is at the cut, but the preview is as I said. The indexing seems to be taking much longer too.

Thx for the hint, you are right, the preview is broken.
I've fixed this now, please update.

Updated and working beautifully again. Thanks

[ProWo]

1. I discoverd a strange behaviour. When you change the pixel format for any encoder, WHA decoding always change to qsv and does not remain unchainged.
That also happens when you switch to some none qsv encoders like DNxHR, av1-libaom, h264-amf, hevc-amf, hevc-nvenc, h266 (vvc). Some change to WHA decoding qsv when you switch direction up, some down!

Thx, will be fixed.

-vf scale=800:450
need to be replaced with
-vf vpp_qsv=w=800,vpp_qsv=h=450

vpp_qsv=w=,vpp_qsv=h= does'nt work here with me.

[mannequin80]

To apply dynaudnorm only to one channel, you can use the h switch, instead of splitting/joining.
To apply to the FC channel only, append

:h=FC

after the other switches.

hmm, i think "volume" filter doesn't support the "h=" switch though so you'd have to apply volume to all channels in this case. what if you only want to apply it to center channel?

[ProWo]

hmm, i think "volume" filter doesn't support the "h=" switch though so you'd have to apply volume to all channels in this case. what if you only want to apply it to center channel?

No, the h switch is for dynaudnorm only.

For the volume filter you must apply channelsplit/merge as before.
Example for increasing left channel only with volume=6dB from a video with stereo audiostream index 1 and exporting this audiostream only:

ffmpeg -i input_audio_2ch -filter_complex [0:1]channelsplit=channel_layout=stereo[l][r],[l]volume=6dB[vl],[vl][r]amerge=inputs=2[aout] -map [aout] -c:a aac -b:a 192k -ar 48000 -vn -sn -dn output.aac

[davidt1]

When I click on "Update", Clever closes instead of updating.

[ProWo]

When I click on "Update", Clever closes instead of updating.

Which version are you using?
Post the contents of your “clever FFmpeg-GUI” folder.

[mannequin80]

No, the h switch is for dynaudnorm only.

sent you PM.

[davidt1]

When I click on "Update", Clever closes instead of updating.

Which version are you using?
Post the contents of your “clever FFmpeg-GUI” folder.

It's working now. Just updated to v3.5.3.02. Didn't work before.

[Findu]

@ProWo

Info: Updating to versions 3.5.3.01 and 3.5.3.02 has changed the previously selected app color.

What if you could set the RGB color values for the background and borders yourself?

---

Like Intel qsv, I did some testing for AMD amf.
The syntax is the same for qsv and amf.
So I tested
-hwaccel amf -hwaccel_output_format amf
-vf format=amf
-vf vpp_amf=w=800,vpp_amf=h=450

My AMD Ryzen 5 7530U with Radeon Graphics (Vega 7 2017-2019) does not realy accelerate from the amf filters.
But Navi 1x, 2x, 3x, 4x might accelerate.
So how about adding the pixel format amf for encoders av1_amf, h264_amf, hevc_amf?

[ProWo]

Like Intel qsv, I did some testing for AMD amf.
The syntax is the same for qsv and amf.
So I tested
-hwaccel amf -hwaccel_output_format amf
-vf format=amf
-vf vpp_amf=w=800,vpp_amf=h=450
So how about adding the pixel format amf for encoders av1_amf, h264_amf, hevc_amf?

Thx for testing, will add it.
I'll also add the scale replacement (vpp_) for QSV/AMF with related HWA decoding.