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

[videoAI]

@ProWo

For ref ..
AC-3-Dynamic-Range-Control

2 things I want to ask.
Can you add the "AC3 Decoder DRC" option for EAC3 Input Files, too?

Will be done.

And maybe set the DRC default always to OFF ?
Thanks

No. because ffmpegs default is on.
You can set it off, if you like.
Dolby itself recommends turning it on.

[videoAI]

@ProWo
v3.4.5.11

.11 feedback .. Still crashes ..

@ProWo, please check index values before calling or using arrays.

@ProWo
v3.4.5.09

Deleted my old backup file to make a new one.

************** Exception Text **************
System.IndexOutOfRangeException: Index was outside the bounds of the array.
at clever_FFmpeg_GUI.CL_Settings.Backup_check()
at clever_FFmpeg_GUI.CL_Settings.VB$StateMachine_72_S tart_controls.MoveNext()
--- End of stack trace from previous location where exception was thrown ---
at System.Runtime.ExceptionServices.ExceptionDispatch Info.Throw()
************** Loaded Assemblies **************

[Findu]

@ ProWo

v3.4.5.11 Feedback

The new timings of the update/restore scripts and on the automatic deletion of updater.bat are working well so far. Many thanks.

---

Can you please rename video encoding av1-amd to av1_amf. It is like h264_amf and hevc_amf and follows the documentation.
I can't test av1_amf as my Ryzen 5 7530U does not support it.

[Findu]

@ ProWo

v3.4.5.12 Feedback

I have done some cycles of restore to 11 and update to 12. No problems found so far.
---
Does "Keep it" just close the window?

[Attachment 88654 - Click to enlarge]
---
Thank you for replacing the underscore from the video encoder names.
---
ffmpeg 8.0 now support -hwaccel amf
https://www.phoronix.com/news/FFmpeg-AMD-AMF-Decoder-FSR
I do not get warnings or errors using it.
So it can be add to the "Scan for errors" HWA-decoding drop down.
"-hwaccel auto" still use DXVA2 on my device ("-hwaccel amf" is often a bit faster).
---
I still test vpp_amf and sr_amf, also in combination with HWA decoding/encoding.
---
ffmpeg 8.0 also support:
- VP9 Vulkan hwaccel
- AV1 Vulkan encoder
- ProRes RAW decoder
- ProRes RAW Vulkan hwaccel
But my devices do not support Vulkan.

[ProWo]

Does "Keep it" just close the window?

Yep.

ffmpeg 8.0 now support -hwaccel amf
I do not get warnings or errors using it.
So it can be add to the "Scan for errors" HWA-decoding drop down.

Will be done.

[ProWo]

v3.4.5.11
.11 feedback .. Still crashes ..

Not helpful.
Please provide the exact steps that led to the crash.

[videoAI]

@ProWo
v3.4.5.12

Kindly click on my previous post .. it is all there ..
396490#post2782491


[Attachment 88658 - Click to enlarge]

v3.4.5.11
.11 feedback .. Still crashes ..

Not helpful.
Please provide the exact steps that led to the crash.

[davidt1]

Is this message necessary? I just want to get things done without having to click through pop-up messages.


[Attachment 88659 - Click to enlarge]

[ProWo]

Is this message necessary? I just want to get things done without having to click through pop-up messages.

Yes, this is necessary to avoid unwanted results.

[ProWo]

v3.4.5.12
.. Still crashes ..

The crash happens when you delete the backup?
You seem to be the only one this happens to.
All your crashes are strange.
Check if you have a “clever FFmpeg gui” folder in
C:\users\yourusername\appdata\local
; if so, delete it.

[davidt1]

Is this message necessary? I just want to get things done without having to click through pop-up messages.

Yes, this is necessary to avoid unwanted results.

Most software displays messages when you put the mouse over the button. It's more standard that way.

[videoAI]

@ProWo
v3.4.5.13

issue closed
-----------------------------------------------------------------
Edit#1 .13 does not crash if backup deleted at the disk level.
!!! Thanks .. for .. fixing !!!

It wasn't my environment after all ..
As always .. there is nothing wrong with my environment
------------------------------------------------------------------

To duplicate, exit clever, delete backup file, reopen clever, go to settings, ***CRASH***

v3.4.5.12
.. Still crashes ..

The crash happens when you delete the backup? [y]
You seem to be the only one this happens to. [don't think so, maybe I am the only one who deleted section #1 in settings]
All your crashes are strange. [thanks !!]
Check if you have a “clever FFmpeg gui” folder in
C:\users\yourusername\appdata\local
; if so, delete it. [I didn't, maybe I am the only one who deleted section #1 in settings]

[videoAI]

@ProWo
v3.4.5.13

v3.4.5.13 Feedback ..

New Issue: The video Encode Progress bar isn't working correctly [mp4],
it goes to100% in 2 secs, then stays there for the encoding duration. [15 min]

Q: under what scenario is the 'All Streams' checkbox grayed?
Multi Video; Multi Audio streams?


[Attachment 88661 - Click to enlarge]

[Findu]

@ ProWo

"Scan for errors" use (-init_hw_device "vulkan=vk:0"), which fall back to pure CPU usage on Core Ultra 7 155H.
That's why I just thought Vulkan wasn't supported.
But when I use (-hwaccel vulkan) it uses 95% of the GPU decoding engine.
How about your Intel system?

---

Video encoding (hwaccel:0 auto), testing on Intel:

6,79x 100 % ffmpeg -i input.mp4 -map 0:0 -c:v:0 av1_qsv -preset veryfast output.mp4
5,52x 81 % ffmpeg -vsync 0 -hwaccel:0 auto -i input.mp4 -map 0:0 -c:v:0 av1_qsv -preset veryfast output.mp4
8,02x 118 % ffmpeg -vsync 0 -hwaccel qsv -i input.mp4 -map 0:0 -c:v:0 av1_qsv -preset veryfast output.mp4

As you can see, (hwaccel:0 auto) isn't always the best choice. So would it make sense to use the HWA-decoding dropdown menu from "Scan for errors" also on the Video encoding page?

[ProWo]

v3.4.5.13 Feedback ..

New Issue: The video Encode Progress bar isn't working correctly [mp4],
it goes to100% in 2 secs, then stays there for the encoding duration. [15 min]

Are you playing with a multivideostream file again?
This happens, because the correct duration can not be found.

Q: under what scenario is the 'All Streams' checkbox grayed?
Multi Video; Multi Audio streams?

Multi videostreams.
Background: Streamcopy can not work with filter_complex.

[ProWo]

@ ProWo

"Scan for errors" use (-init_hw_device "vulkan=vk:0"), which fall back to pure CPU usage on Core Ultra 7 155H.
That's why I just thought Vulkan wasn't supported.
But when I use (-hwaccel vulkan) it uses 95% of the GPU decoding engine.
How about your Intel system?

Thx for testing. I'm on an old I7. No vulkan support.
Will change it for vulkan.

Video encoding (hwaccel:0 auto), testing on Intel:
6,79x 100 % ffmpeg -i input.mp4 -map 0:0 -c:v:0 av1_qsv -preset veryfast output.mp4
5,52x 81 % ffmpeg -vsync 0 -hwaccel:0 auto -i input.mp4 -map 0:0 -c:v:0 av1_qsv -preset veryfast output.mp4
8,02x 118 % ffmpeg -vsync 0 -hwaccel qsv -i input.mp4 -map 0:0 -c:v:0 av1_qsv -preset veryfast output.mp4

As you can see, (hwaccel:0 auto) isn't always the best choice. So would it make sense to use the HWA-decoding dropdown menu from "Scan for errors" also on the Video encoding page?

It seems that the folks at ffmpeg have improved their code.
Will see, what can be done.

EDIT: All done, test .14

[videoAI]

My replies ..

video Encode Progress bar isn't working correctly ...

Are you playing with a multi video stream file again? [NOOOOOOOO]
This happens, because the correct duration can not be found.

It is happening with some files .. and they are not multi-stream ..
will test a bit more and revisit the issue ..

Q: under what scenario is the 'All Streams' checkbox grayed?
Multi videostreams.
Background: Streamcopy can not work with filter_complex.

Thx

[videoAI]

I tested my AMD GPU on FFmpeg 8 ..

$input  = "C:\T\Test\Test.mp4"
$output = "C:\T\Test\out.mp4"

# • -hwaccel d3d11va – use AMD GPU for decoding (Direct3D 11)
# • -c:v h264_amf  – AMD AMF hardware encoder
# • -b:v 5M        – target bitrate
ffmpeg -hide_banner -y -benchmark `
       -hwaccel d3d11va `
       -i $input `
       -c:v h264_amf -b:v 5M `
       $output

utime= User mode CPU time spent in FFmpeg (seconds).
stime= System mode CPU time (seconds).
rtime= Real (wall clock) elapsed time for the operation (seconds).
maxrss= Peak resident set size – the maximum memory FFmpeg used (KiB).

[without h/w acceleration]
bench: utime=16.656s stime=5.438s rtime=8.402s
bench: maxrss=84864KiB

[with h/w acceleration]
bench: utime=11.953s stime=7.172s rtime=9.319s
bench: maxrss=118780KiB

Why hardware accelerated runs can appear slower:
------------------------------------------------
GPU warm up / driver overhead
The first call to `-hwaccel d3d11va` loads the AMD driver, creates a D3D11 device,
and copies frames to GPU memory. This one time cost is counted in `rtime`.

CPU2GPU data transfer
Each decoded frame must be copied from system RAM to GPU memory and the
encoded frame back to RAM (or directly to disk). If the video is short, the transfer
time can dominate the total wall clock time.

Bitrate / encoder settings
A higher bitrate (`-b:v 5M`) forces the encoder to work harder, sometimes more
so on the GPU than a simple CPU only libx264 preset.

Threading differences
The AMF encoder may use fewer CPU threads than libx264, so `utime` drops while
`stime` rises (driver work). The overall `rtime` can stay similar or increase.

I/O bottleneck
Writing the output file is still done by the CPU. If the disk is the limiting factor,
GPU speed won’t help.

Small test clip
Benchmarks are more reliable on longer clips. With a brief file, the fixed overhead
(GPU init, context creation) is a larger fraction of total time.

Interpreting the numbers:
-------------------------
- CPU time (`utime + stime`) dropped from ~22 s (CPU only) to ~19 s with GPU,
showing the GPU did some work.
- Wall clock time (`rtime`) increased from 8.4 s to 9.3 s because the fixed GPU
setup cost outweighed the speed gain for this particular clip.
- Memory usage (`maxrss`) grew (84 MiB → 118 MiB) due to driver buffers and
GPU related structures.

To see a real speed up:
-----------------------
1. Use a longer video (several minutes long) so the fixed overhead is amortized.
2. Match encoder settings – try a faster preset (`-preset fast` for libx264) vs.
the default AMF settings; adjust `-rc_mode` or `-quality` for AMF to a comparable workload.
3. Measure only the encoding stage – run `ffprobe -show_frames` to get frame
by frame timestamps, or pipe the decoded stream directly to the encoder
(`-f rawvideo -`) to eliminate extra file I/O.

My input file is 2.5 mb, my output file is 110 mb if I use no Bitrate settings,
and 44 mb if I use '-b:v 5M'.

Why a 2.5 mb source becomes a 110 mb file:
------------------------------------------
Default encoder bitrate
When you omit `-b:v` (or any rate‑control mode), the AMF encoder falls
back to a *very high* internal bitrate (often > 50 Mbps) to preserve quality.

Lossless‑ish mode
Some hardware encoders automatically switch to a lossless or near‑lossless
profile if no quality target is given, which inflates the file size dramatically.

Resolution / frame rate unchanged
The source’s resolution and frame‑rate are kept, so the encoder writes
many more bits per frame than the original compressed stream.

Container overhead
Switching from the original container (e.g., MP4 with variable bitrate) to a
new stream with a constant high bitrate adds extra metadata and index
tables, but the bulk of the growth is the video bitstream.

[ProWo]

I tested my AMD GPU on FFmpeg 8 ..

ffmpeg -hide_banner -y -benchmark `
       -hwaccel d3d11va `
       -i $input `
       -c:v h264_amf -b:v 5M `
       $output

Don't use d3d11va

Repeat the test with once with -hwaccel auto
then with -hwaccel amf

[Findu]

@ ProWo

v3.4.5.14 Feedback

Thanks for adding the HWD drop down to the Video encoding page. It works well. Also the changed vulkan entry.
I like that the HWD selection stay when changing the encoder and get set to none when reopen the mux page.
But why did you remove Cuda, OpenCL, the older Intel and the three DirectX entries?

---

I have done some cycles of restore to 13 and update to 14. No problems found so far.

[videoAI]

Thx for the tip, will try both ..

I tested my AMD GPU on FFmpeg 8 ..

Repeat the test with once with -hwaccel auto
then with -hwaccel amf

[lomero]

from latest v3.4.5 nightly missing update feature ... why ?

[ProWo]

from latest v3.4.5 nightly missing update feature ... why ?

You're on the last release (3.4.5) or on a nightly (3.4.5.x)?
Please specify the steps to your issue.

[ProWo]

But why did you remove Cuda, OpenCL, the older Intel and the three DirectX entries?

cuda, opencl and others will be removed, but the two directx (dxva2 and d3d11va) will still remain, because they are the only working with older CPU's
(as mine I7).
I'm working on a further change in videoencoding, to use HWA decoding with all encoders.

[lomero]

from latest v3.4.5 nightly missing update feature ... why ?

You're on the last release (3.4.5) or on a nightly (3.4.5.x)?
Please specify the steps to your issue.

nightly 345.14
before (344 version) from main page have update feature. now from setting&more (page 1 and 2) missing update feature ...
also: with new portable.config blurry characters (i talk about win11, need override high DPI scaling from Tab properties)


[Attachment 88691 - Click to enlarge]

[ProWo]

nightly 345.14
now from setting&more (page 1 and 2) missing update feature ...

Settings & more is right, but you'll see the update button only, if an update is available.

also: with new portable.config blurry characters

Do you have this file

clever FFmpeg GUI.exe.config

in your clever folder?
If not, download the 3.4.5 release and extract it from the zip.

[videoAI]

# -hwaccel amf or auto

ffmpeg -hide_banner -y -benchmark `
       -hwaccel auto `
       -i $input `
       -c:v h264_amf -b:v 5M `
       -stats_period 1 `
       -loglevel debug `
       -report `
       -progress "ffmpeg.log" `
       $output

[-hwaccel auto]
bench: utime=10.328s stime=5.281s rtime=7.048s
bench: maxrss=130548 KB

CPU time = utime + stime = 15.609 s
Memory used ≈ 127 mb
out 37 mb

[-hwaccel amf]
bench: utime=18.672s stime=17.156s rtime=44.432s
bench: maxrss=145340KiB

CPU time = utime + stime = 35.83s
Memory used ≈ 142 mb
out 37 mb

[-hwaccel auto]
# AMD GPU Device Name & ID [win11 (Powershell)]
wmic path win32_VideoController get Name,DeviceID
DeviceID Name
VideoController1 AMD Radeon(TM) Graphics

# AMD GPU HW Device List [ffmpeg]
ffmpeg -hide_banner -init_hw_device amf:list
Universal media converter

AI Analysis:
HW acceleration DXVA2 → D3D11 → AMF (`h264_amf` encoder) on device `dxva20`
DXVA2 decoder (`h264`) → filtergraph → AMF H.264 encoder
GPU accelerated decode (DXVA2) and encode (AMF) [!!]
CPU still contributed ~2 cores (user + system > real time)

FFmpeg debug mode:
[-hwaccel auto]
[AMF @ 000001faa1ee8380] AMFEncoderCoreH264: … Debug: GetProperty(EncoderGPU), -1
** Incorrect device selection
– you didn’t specify a GPU [for AMD you need -hwaccel amf or -device <id>].

# AMD: specify device explicitly
ffmpeg -hwaccel amf -init_hw_device amf:0
or -init_hw_device VideoController1
or -init_hw_device "Universal media converter"
All Didn't work, and generated errors ..
[** Supported hwaccels: cuda vaapi dxva2 qsv d3d11va d3d12va amf]

I don't think FFmpeg is recognizing my integrated GPU as a hardware device,
so H/W acceleration, for decode or encode, doesn't seem to kick in, the heavy
lifting is handled by the Amf s/w encoder. In the OS, my GPU utilization is at 29%
during the FFmpeg encode, which tells me the OS employs the integrated GPU
when there is a need to [video encoding].

Debug log attached ..

Repeat the test with -hwaccel auto
then with -hwaccel amf

[videoAI]

For a while now, I have been noticing that my win PC slows down just before an update
from MS is available.

Any of you guys have this issue?

Well .. according to AI, this is what's been happening ...

Windows can feel slower in the days leading up to a major update for several reasons:

1. Background preparation
Windows begins downloading update metadata, checking device compatibility,
and pre‑fetching parts of the update. Those network and disk operations compete
with your regular tasks.

2. Telemetry & diagnostics
Prior to a release Microsoft ramps up telemetry collection to gauge system health
and identify potential issues. The extra CPU and I/O load can reduce available resources.

3. Scheduled maintenance tasks
Windows often runs cleanup, defragmentation, and driver‑compatibility checks in the
background when an update is imminent, which can temporarily affect performance.

4. Resource reservation
The system may reserve RAM and disk space for the upcoming update, leaving slightly
less for active applications.

These activities are generally short‑lived and finish once the update is installed or the
preparation window closes. If the slowdown feels excessive, you can temporarily
pause automatic updates, run `disk cleanup`, or limit background bandwidth in
[Settings → Delivery Optimization].

[videoAI]

@ProWo
v3.4.5.14

Enhancement Request ..

Under video image properties; Could you please consider adding gamma (R,G,B).
Like -vf "eq=gamma_r=0.95:gamma_g=1.2:gamma_b=0.85".

It would help me adjust all 3 colors without having to edit the BT_open.txt file and
would give me preview as well.

Thank you .. and don't forget to unit test the changes ..

[lomero]

Settings & more is right, but you'll see the update button only, if an update is available

ah ok. clear now. on prev version update was always present

Do you have this file clever FFmpeg GUI.exe.config in your clever folder?

oh yes, you're right. i'm new for 345 version (a lot of changes from prev 344). now all solved. thanks