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

[davidt1]

Anyone here uses Potplayer?

On my AMD laptop it plays tp files recorded from the Comet TV station with audio only and no video. The files play with both audio and video in VLC, MPV, and MPC-HT.

I can't figure out the problem. Thanks

Video codec not supported in Potplayer?

Thank you.

I don't know. TP files recorded from other stations play with sound and video.

[davidt1]

Why does the muxing page show the original video and the encoded video? The original video is not needed.

Apparently, x264 software encoding retains the embedded CC in ts files while hardware encoding such as x264_amf removes it. This is ffmpeg, not just Clever. Can someone explain the inconsistency? Thanks.

I tried with ffmpeg and then Clever. Same result.

[ProWo]

Why does the muxing page show the original video and the encoded video? The original video is not needed.

All source streams are always displayed in the multiplex window.
This guarantees maximum flexibility. Simply deselect or hide the streams you don't need.

Apparently, x264 software encoding retains the embedded CC in ts files while hardware encoding such as x264_amf removes it.

That's interesting.
I just tested it with an h264 (avc) stream with cc.
Encoded with x264 and x264_qsv, both with modified resolution too.
The cc are still included in all outputs.

[davidt1]

Why does the muxing page show the original video and the encoded video? The original video is not needed.

All source streams are always displayed in the multiplex window.
This guarantees maximum flexibility. Simply deselect or hide the streams you don't need.

Apparently, x264 software encoding retains the embedded CC in ts files while hardware encoding such as x264_amf removes it.

That's interesting.
I just tested it with an h264 (avc) stream with cc.
Encoded with x264 and x264_qsv, both with modified resolution too.
The cc are still included in all outputs.

It's just amd amf then. I am not a fan of how AMD handles encoding. Intel QSV looks better.

[Apolosky]

Hello @ProWo

I'm trying to extract audio segments from an AAC file using Cut, but I can't get it to work. When importing the audio into Clever, the AAC file shows a duration of 1:02:23.706, but if I mux it into an MKA container, it shows 1:03:56.099.

If I try to extract a segment—especially near the end of the file—using the raw file (without a container), for example: 00:55:00.000 - 1:03:00.000, I get an error message: "Endtime Error! / Selected Endtime > Duration."

What I need is to cut certain audio segments that I don't need and then concatenate the rest, without transcoding and being as precise as possible.

How should this type of AAC format be edited?"

[ProWo]

Hello @ProWo

I'm trying to extract audio segments from an AAC file using Cut, but I can't get it to work. When importing the audio into Clever, the AAC file shows a duration of 1:02:23.706, but if I mux it into an MKA container, it shows 1:03:56.099.

If I try to extract a segment—especially near the end of the file—using the raw file (without a container), for example: 00:55:00.000 - 1:03:00.000, I get an error message: "Endtime Error! / Selected Endtime > Duration."

What I need is to cut certain audio segments that I don't need and then concatenate the rest, without transcoding and being as precise as possible.

How should this type of AAC format be edited?"

Raw .aac files do not store duration information, therefore the duration is estimated and is inaccurate.
To avoid this, mux the raw aac stream to mkv and cut this new mkv.
Then you can join/merge this mkv junks.
If you need aac as output, extract it from the final mkv.

[Findu]

@ProWo
3.5.2.02 feedback

Hi ProWo. I hope you're doing well!

I noticed two things:

When you hover under the “HWA decoding” option, the tooltip appears again.

[Attachment 91528 - Click to enlarge]

When you try to cut between I (K) frames in "Frame accurate cut", the following message appears.

[Attachment 91529 - Click to enlarge]
This isn't a problem if the cut includes an I (K) frame.
You can try this with any video.

[ProWo]

Hi ProWo. I hope you're doing well!

Hi Findu, thanks for asking—everything's great.

I noticed two things:
When you hover under the “HWA decoding” option, the tooltip appears again.

Thx, will be removed.

When you try to cut between I (K) frames in "Frame accurate cut", the following message appears.
This isn't a problem if the cut includes an I (K) frame.
You can try this with any video.

Thanks, that's an interesting discovery.
I've never cut between keyframes before.

I took a look at it and made the changes.
You can test the solution. Check your PM.

[Findu]

Still testing 3.5.2.03...

I can now perform frame accurate cut between keyframes. Thank you for the change.

Funny thing: Such a cut file show each frame as a keyframe. But the file size is still small. So nothing to worry.

---

NEW: ffmpeg 8.1 offers D3D12 H.264/AV1 encoding.
I will test if it is running on my hardware.

[Findu]

Found a problem with Frame accurate cut, between keyframes.

This error occurs when the cut begins at a keyframe and the end of the cut is located before the next keyframe plus one frame.

System.IO.FileNotFoundException: Die Datei "C:\Users\Findu\Downloads\000_fasc_step2.mkv" konnte nicht gefunden werden.
Dateiname: "C:\Users\Findu\Downloads\000_fasc_step2.mkv"
bei System.IO.__Error.WinIOError(Int32 errorCode, String maybeFullPath)
bei System.IO.File.InternalMove(String sourceFileName, String destFileName, Boolean checkHost)
bei clever_FFmpeg_GUI.Form8.VB$StateMachine_297_Do_it. MoveNext()

There is no problem if the start of the cut is not a keyframe.

[Findu]

NEW: ffmpeg 8.1 offers D3D12 H.264/AV1 encoding.

h264_d3d12va and hevc_d3d12va running well on my Intel system.

A simple speed comparison:

0,59x ffmpeg -i input.mp4 -c:v libx264 output.mp4
3,32x ffmpeg -i input.mp4 -c:v h264_qsv output.mp4
4,05x ffmpeg -hwaccel qsv -hwaccel_output_format qsv -i input.mp4 -c:v h264_qsv output.mp4
4,33x ffmpeg -hwaccel d3d12va -hwaccel_output_format d3d12 -i input.mp4 -c:v h264_d3d12va output.mp4

0,34x ffmpeg -i input.mp4 -c:v libx265 output.mp4
4,78x ffmpeg -i input.mp4 -c:v hevc_qsv output.mp4
4,65x ffmpeg -hwaccel qsv -hwaccel_output_format qsv -i input.mp4 -c:v hevc_qsv output.mp4
4,36x ffmpeg -hwaccel d3d12va -hwaccel_output_format d3d12 -i input.mp4 -c:v hevc_d3d12va output.mp4

[ProWo]

Found a problem with Frame accurate cut, between keyframes.
This error occurs when the cut begins at a keyframe and the end of the cut is located before the next keyframe plus one frame.

End cut how many frames before the next kf?
What do you mean with plus one frame, at what position?

[ProWo]

NEW: ffmpeg 8.1 offers D3D12 H.264/AV1 encoding.
h264_d3d12va and hevc_d3d12va running well on my Intel system.

Thanks for testing.
The hevc_d3d12va is slower than hevc_qsv, so we can forget about that.
The h264_d3d12va seems promising, I might add it.

EDIT:
Tested: ffmpeg -hwaccel qsv -hwaccel_output_format qsv -i input.mp4 -c:v h264_qsv output.mp4
Got error: Driver does not support any RC mode compatible with selected options.
Latest Intel drivers for my I7 platform installed.
That doesn't seem to be fully developed yet.

[ProWo]

Found a problem with Frame accurate cut, between keyframes.
This error occurs when the cut begins at a keyframe and the end of the cut is located before the next keyframe plus one frame.

Reworked, check your PM.

[Findu]

@ProWo
3.5.2.03 feedback 2

No more errors at cut between keyframes. Thank you.

[Findu]

NEW: ffmpeg 8.1 offers D3D12 H.264/AV1 encoding.
h264_d3d12va and hevc_d3d12va running well on my Intel system.

Tested: ffmpeg -hwaccel qsv -hwaccel_output_format qsv -i input.mp4 -c:v h264_qsv output.mp4
Got error: Driver does not support any RC mode compatible with selected options.
Latest Intel drivers for my I7 platform installed.
That doesn't seem to be fully developed yet.

My Intel Core Ultra 7 155H has no problem with it running an old december 2023 or the latest ARC driver from march 2026.

[davidt1]

-c:v h264_d3d12va does not work on my AMD setup.

[ProWo]

The quality of the results is superb, but configuration options for AV!-QSV are very restricted, no global quality settings at all

The av1_qsv encoder apparently doesn't have a direct parameter for quality control (according to the FFmpeg help, see below).
It's possible that the general parameter `-global_quality` works; I can't test this myself because my hardware doesn't support av1_qsv.

But you can test this yourself.
Select the av1_qsv encoder in the encoding settings and then click “To batch.” Then close Clever.
In the destination folder, you’ll find a file named BT_open.txt
Open it with a text editor and replace:

-c:v av1_qsv

with

 -c:v av1_qsv -global_quality 25

Save the file, start Clever, click on Batch tasks, and execute.
If the encoding completes without issues, then the parameter is supported.

Post the result here in the thread; if it’s positive, I’ll adjust the encoder settings.

[Findu]

I can confirm that
-c:v av1_qsv -global_quality 25
is working on Intel Core Ultra 7 155H.
Also in combination with hardware accelerated decoding: -hwaccel qsv -hwaccel_output_format qsv

---

-c:v av1_d3d12va doesn't seem to work.

[ProWo]

I can confirm that
-c:v av1_qsv -global_quality 25
is working on Intel Core Ultra 7 155H.
Also in combination with hardware accelerated decoding: -hwaccel qsv -hwaccel_output_format qsv

Thx for testing, will implement the quality settings for av1_qsv.

[ProWo]

----------------------------------------------------------------------------------
March 26, 2026 release version 3.5.3.
----------------------------------------------------------------------------------

Added:
Reads fps from extracted video raw stream name now.
Mux mpeg1 and mpeg2 to mov.
Prevent form closing while executing batches.
3 new backgrund colors (browngrey).
Big logo on start screen.
Max gop size 120 for h264 and hevc encoders.
Quality choice to av1_qsv encoder settings.

Fixed:
Bug in 1st chapter creation time.
Bug with frame accurate cut between keyframes.
Bug with wrong format profile level values displayed.

Improved:
Interlaced encoding

Changed:
clevernorm/dynaudnorm coupling=off

----------------------------------------------------------------------------------

[jopavirando]

Beautiful home screen. Thank you.

[Findu]

@ProWo
3.5.3 feedback

1. When I select the encoder av1-qsv and h266 (vvc), the Pixel format get automatical set to yuv420p10le and not to "Like source" like for all other encoders. Can you please change this?

Because that cause problems using av1-qsv with HWA-Decoding qsv with example code
-vf fps=24000/1001,format=yuv420p10le

But it is working with code
-vf fps=24000/1001
or
-vf fps=24000/1001,format=qsv

2. Could you also add the supported pixel format qsv for encoder av1_qsv, mpeg2_qsv, vp9_qsv, h264_qsv, hevc_qsv?

3. How about adding an "Ignore errors" option to batch tasks so that the entire list is processed without interruption?


[Attachment 91705 - Click to enlarge]

4. How about this positions of the big logo? Bottom left or center.


[Attachment 91706 - Click to enlarge]


[Attachment 91707 - Click to enlarge]

[ProWo]

1. When I select the encoder av1-qsv and h266 (vvc), the Pixel format get automatical set to yuv420p10le and not to "Like source" like for all other encoders. Can you please change this?

I checked that again, and you're right—both encoders handle 8-bit encoding correctly.
So I'll make that change.

But it is working with code
-vf fps=24000/1001,format=qsv
2. Could you also add the supported pixel format qsv for encoder av1_qsv, mpeg2_qsv, vp9_qsv, h264_qsv, hevc_qsv?

I tested that with mpeg2_qsv, h264_qsv, hevc_qsv; format=qsv doesn't work here.

3. How about adding an "Ignore errors" option to batch tasks so that the entire list is processed without interruption?

I'll find another solution here.

4. How about this positions of the big logo? Bottom left or center.

I'll align the logo to the left of the top one; vertically, it will remain aligned with the Quit button.

[Findu]

I tested that with mpeg2_qsv, h264_qsv, hevc_qsv; format=qsv doesn't work here.

That's too bad for you.
Support for mpeg2_qsv has been removed for Intel ARC.
format=qsv works here for av1_qsv, vp9_qsv, h264_qsv, hevc_qsv.

[ProWo]

format=qsv works here for av1_qsv, vp9_qsv, h264_qsv, hevc_qsv.

Is there a performance advantage to using format=qsv over other pixel formats?

[Findu]

Sure!
Quick test with a 720p sampel:

17,4x av1_qsv like source
22,9x av1_qsv format=qsv

20,7x hevc_qsv like source
24,4x hevc_qsv format=qsv

20,8x h264_qsv like source
24,3x h264_qsv format=qsv

---

EDIT:

-c:v h264_qsv -g 120 -keyint_min 120 -global_quality 25

14,6x like source
17,4x format=yuv420p
16,8x format=nv12
13,2x format=yuv420p10le
19,0x -hwaccel qsv -hwaccel_output_format qsv like source
24,1x -hwaccel qsv -hwaccel_output_format qsv format=qsv

---

EDIT2:

By the way:
ffmpeg state for av1_qsv and vp9_qsv:
Please use the global_quality option and other options for a quality-based mode or the b option and other options for a bitrate-based mode if the default is not the desired choice.

It is working well for av1_qsv, but has no effect for vp9_qsv on my system.

[ProWo]

Sure!

Thx for testing.
Will add qsv pixelformat to all qsv encoders except mpeg2_qsv.

[mannequin80]

----------------------------------------------------------------------------------
March 26, 2026 release version 3.5.3.
Changed:
clevernorm/dynaudnorm coupling=off
----------------------------------------------------------------------------------

is there a specific reason you decided to disable the coupling? for "clevernorm" it shouldn't do anything since we're only amplifying one (center) channel (unless you've made further changes to the filter) and for the regular "dynaudnorm" disabling coupling destroys stereo image, panning effects and the overall intended mix of the original. everything becomes "loud" and unnatural without any sense of balance, especially multichannel movie audio. coupling=off has a very narrow usage scenario like a badly recorded stereo voice recording with a failing one channel, for example, and not much else.

[ProWo]

is there a specific reason you decided to disable the coupling? for "clevernorm" it shouldn't do anything since we're only amplifying one (center) channel (unless you've made further changes to the filter) and for the regular "dynaudnorm" disabling coupling destroys stereo image, panning effects and the overall intended mix of the original. everything becomes "loud" and unnatural without any sense of balance, especially multichannel movie audio. coupling=off has a very narrow usage scenario like a badly recorded stereo voice recording with a failing one channel, for example, and not much else.

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

But you're right, it doesn't make sense for multichannel audio.
I’ll revert the change, but leave it for stereo output.

As for Clevernorm, the situation is as follows:
Mono and stereo don’t have an FC channel, so the filter is applied to the individual channels.
For 3 or more channels, the filter is applied only to the FC channel.