JavPlayer Manual Ver.2.

02b

Please do not sell, distribute or publish videos created using JavPlayer without
the permission of the copyright holder.

Overview
At first it was a video player, but now it is primarily a processing and recording application.
Recording requires complex processing and is time-consuming because it is done with external programs.
A watermark will be added to the created video, but JavPlayer has the ability to play it back while
removing it.

Requirements
If you just want to playback, a PC with Core-i3 (with built-in GPU) or higher is fine.
However, recording (including super-resolution) without a video card takes a tremendous amount of time.
I have taken into account that the basic functions can be used with a GPU of GTX750 or higher, but a
minimum of 4 GB of VRAM is required to support all external tools and optional settings in combination.
It can be used with GPUs that support DirectX12 such as Radeon, but some external tools cannot be used.
Operation has been checked on Windows10(64bit) and Windows11. *Windows7 is not supported.

Installation of JavPlayer
Extract zip to a location that does not contain double-byte characters in the full path and execute
JavPlayer.exe.
Due to changes in file structure, overwriting older versions may cause problems.
When updating, you should extract to a new folder and move the TG folder and JavPlayer.udj from the old
version to the new version.
If you have changed JavPlayer.ini, you will need to rewrite it again, but if no items have been added (or
deleted) in the new version, it is easier to copy.

Installation of Codec
If you cannot play H265(HEVC) video on Windows10, please install the HEVC codec.
If HEVC is selected for encoding, it is also required for recording.
HEVC Video extension
ms-windows-store://pdp/?ProductId=9n4wgh0z6vhq
*There may still be a download page for the free version, although it is now paid for.
Likewise, install AV1 if necessary.
AV1 Video extension
Installation of TecoGAN
TecoGAN (see below for details) must be installed to make recordings.
First, download TecoGAN Ver.1.11 built for Windows.
Since GoogleDrive has strict transfer volume limits, I recommend that you download from the download
page of Gumroad or the the JavPlayer introduction page ( http://x.gd/UJccc).
Next, extract the zip and copy the two folders (TGMAIN,TG-MODEL) to "JavPlayer/TG".
After installation, contents will look like this.

Uninstall
Delete the extracted folder.
If you plan to reinstall, please save & restore ”INSTALLATION_FOLDER/JavPlayer.udj".
If you delete JavPlayer.udj, user settings can not be able to restore.

Data Save
User data including settings panel contents, marker positions, etc. are automatically saved in a file named
JavPlayer.udj.
Any changes made to the video-specific settings will be saved when the next video is opened.
Even if there are no changes the video-specific settings, saved when the application is closed.
Terms
Cell The rectangles that make up the mosaic.
ProcessingSettings Parameters for mosaic reduction processing. Adjust to enhance the effect.
Marker Dividers placed on the seek bar to switch processing settings.
Swipe Operation to slide the mouse while pressing the left button.
MosaicDetection This process examines the position and roughness of the mosaic, and AI can be used to
improve accuracy.
HumanBodyDetection Increase the priority of the human body area to prevent mosaic roughness and
other factors from being determined based on the face and background.
Super-Resolution Blur-free enlargement process. Recently, AI has been used for this purpose.
TecoGAN This is the main video-super-resolution algorithm used by JavPlayer and I am distributing as a
tool for Windows.
SPAN A single-image-super-resolution algorithm that further sharpens the super-resolution results and is
built into JavPlayer.
NatCorder Asset for encoding built into JavPlayer, supporting H264 and HEVC.
FFmpeg A versatile tool for encoding, editing, etc. of videos on the command line.
H264 Standard video codec. It can encode/decode in a wide range of environments.
HEVC A new video codec. File size is about 2/3 of that of H264, but less versatile.
AV1 Next generation standard video codec. Not yet widely used.
ProRes Apple's lossless video codec and is widely used for intermediate data in editing work.
AAC Audio codecs for movies(mp4,mov) created by JavPlayer.
Equirectangular-format Recording format for VR movies, in which the captured circular image is
transformed into a square and recorded.
FishEye-format Recording format for VR movies, in which the captured circular image is recorded as they
are.
Basic Operation

① Record movies during playback or capture.

② Capture the screen of the DMM player etc in real time.
③ Open the file browser.
④ Load a video for comparison. Use the shortcut key (default: N) to switch videos.
⑤ Start audio lag correction.
⑥ Open the Key Assignment panel.
⑦ Open the Environment Settings panel.
⑧ Open the Operation Settings panel.
⑨ Open the Screen Settings panel.
⑩ Open the Color Correction panel.
⑪ Open the Processing Settings panel.
⑫ Pause / Resume *Same as double-click the screen
⑬ Move the handle and seek.*Horizontal swipe allowed
⑭ Move the handle and adjust volume. *Mouse wheel allowed
⑮ Switch to the debug log display screen. *Ctrl+MouseL+MouseR does the same.

Drag and drop the movie or image file to the window of this application and play it.
While the screen is switched, such as the file browser or RecSettings, ESC key can be used to return.
Hold the mouse L&R button for 1 second to hide the screen. *Cancel with pressing L&R.
Toolbar

It is displayed above the seek bar only when the 'toolbar' is ON in the environment settings panel.
It includes buttons related to screenshot, loop(record) range, play speed, marker editing, and bookmark.
You can assign keys to all functions.
Normally, buttons related to copying and pasting parameters are not displayed.
They are displayed instead of the bookmark buttons only while the markers are placed and the processing
setting panel is open.

Range specification

If you set the range with the toolbar buttons or the keyboard, only that part will be played repeatedly.
If only the start point is specified, the end of the movie will be the end point.
If only the end point is specified, the beginning of the movie will be the start point.
It returns to the start point at the moment past the end point, but before the start point it plays normally.
Loop is not executed during pause.
It is also used to specify the range to record, and the buttons on the right can be used to "Select Whole"
and "Unselect Range".

Bookmark

Selecting a bookmark (1-3) with the toolbar button or keyboard will place a bookmark marker at the
current playback point.
If the selected bookmark has already been set, it moves (seeks) to the marker position.
If you select a bookmark while holding down the Shift key, the corresponding marker is deleted.
You can use additional bookmarks (4 to 9) by assigning shortcut keys.
Audio lag correction

When button ⑤ is pressed, the toolbar changes like this.

In the case of "the sound is delayed by 0.5 second throughout the entire video." set as follows
Audio delay applying to start:-0.5, Audio delay applying to end:-0.5
In the case of "No lag at the beginning of the video, and delay 1sec at the end." set as follows.
Audio delay applying to start:0, Audio delay applying to end:-1
It will be immediately reflected in the video being played, so SAVE it after making fine adjustments.
The corrected video will have the file name "OriginalName_c*.mp4". '*' is sequential number.
Re-encode the audio as needed, but the video is copied from the original video so it doesn't degrade.
Because it is a function to modify the video created with JavPlayer, it does not support videos other than
mp4.
File Browser

Double-click the file to start playback

The folder of the played video will be added to the left folder link. *up to 8

Opening and closing with shortcut keys (default: F key) is convenient when using a file browser frequently.
Now it is windows standard style, but you can change it back by setting "UseOldFileBrowser=true" in
JavPlayer.ini.

Sequential Playback
Video switching (default: F7 key, F8 key) is a function to play videos in the same folder continuously.
Intended to be used while wearing goggles in VR mode, but it can also be used for still images (png, jpg).
You can also load the last file after rebooting.

If you drag and drop multiple videos together, they will be treated as a video group.
The target of video switching is limited to files in a group and can be used as a simple playlist.
Dropping a single file or loading from the file browser will ungroup it.
Key Assignment

You can assign up to 2 keys or buttons for each function.

The default is to use a combination of keyboard(101) and pad(Xbox360).
Please change according to your environment.

Seek (Forward or Rewind) will be executed continuously if you keep pressing.

If you specify the range (start or end) in the same position consecutively, it will be canceled.
Pressing the start and end specifing keys simultaneously cancels both.
"Cancel processing" is a function to confirm the original state (mosaic), which can be locked by pressing
the Shift key at the same time.
Screenshots will be stored in "JavPlayer/ScreenShot/*_ XXX.png".
Environment Settings

① The fps slider is usually set to the far left (VSync).

If VSync is disabled in the video card settings, select 30 or 60. *90 is for load measurement.
② Even if you turn off the toolbar, you can do all the operations with the keyboard shortcut.
③ Automatically expands to the size specified by LimitWindowSize. This is useful when playing videos of
different resolutions in succession.

Operation Settings

① Sets whether the title of the video is displayed in windowed and full-screen mode.
② Specify where on the screen messages such as warnings and recording progress should be displayed.
③ Select the mouse wheel function from Volume, Seek, Seek (reverses direction).
The seek interval can be set with the slider just below.
Hold down Shift and turn to skip (long-distance seek).
Screen Settings

① If the video contains a black border, press the measurement button and then press the apply button.
If you measure on a bright screen the correct value will be set usually.
If the capture target contains a non-black frame such as a status bar, you need to input manually.
② Select projection type.
Flat(2D) for normal videos, Equirect for normal VR videos, and FishEye for fisheye format VR videos.
③ Select the decoder to use. *The decoder in use is displayed.
The default MediaFoundation has a problem that some videos (avi) cannot be played and seek is slow.
DirectShow may have problems with high-resolution video.
Switch as necessary.
When recording in 1pass and 2pass, a separate decoder from these is used for high-speed frame-by-
frame transmission.
④ Select the aspect ratio of the screen.

Color Settings

① You can adjust the brightness of the screen when you move R,G,B while holding down the Shift key.
Processing Settings

Usually applies to whole, but you can use markers to switch between multiple settings.
The toggle switches on the top is black for ON and gray for OFF.
① SPUIT ICON Parts close to the specified color will be protected from processing.
It is effective in case of the lattice pattern of the white window is misjudged as mosaic.
② AI ICON Use AI for mosaic detection. *The playback load is higher.
It can reduce false detection of backgrounds and hard-to-distinguish mosaics.
③ BODY ICON Detects human bodies to assist in mosaic detection. *The playback load is higher.
Identify the human body and ensure that the mosaic contained therein is prioritized.
④ HD ICON Increases the resolution of the image for mosaic detection.
Effective for small mosaics, but weakens for rough mosaics and processing slower.
The button will not be displayed for low-resolution videos that fit in 640x384.
⑤ The selected picture quality is reflected in the automatic adjustment of sensitivity and accuracy.
If the mosaic is not processed and remains, try reducing the value.
Normally 0 (automatic adjustment OFF) is not used.
⑥ Setting a large value makes it easier to identify fine mosaics, but increase false positives.
It can be set only when the automatic adjustment is OFF.
⑦ Smaller values make it easier to identify blurred mosaics, but increase false positives.
It can be set only when the automatic adjustment is OFF.
⑧ Adjust the intensity of the paint filter.
Although it can reduce the mosaic without blurring, there is a problem that distortion occurs.
While recording in TG mode, this setting is ignored and fixed at maximum.
⑨ Set the cell width of mosaic.
Normally the auto is set, but manual is effective only in the section where automatic judgment fails.
In scenes without a mosaic, useless processing can be omitted by setting the left end (Off).
⑩ Set the cell height of mosaic.
The slider will appear only when the width is set manually.
⑪ Select the viewing angle for mosaic distortion correction of VR video. *Equirect<->FishEye
Select None for videos that play normally and the mosaic at the bottom edge appears square.
Correction is usually done at a viewing angle of 180 degrees, but if the mosaic distortion is not resolved,
try selecting another.
Only visible when Equirect or FishEye is selected as the projection type.
On the playback screen, you can only see the correction results on either the left or right side of the
screen. *Right-click to switch.

While the Processing Settings panel is open, simple super-resolution + blur processing is not
performed on the mosaic, and the mosaic analysis results are displayed.
Mosaics that are not included in the blue area are not detected and will not be processed in the video to
be created.
If the grid does not match the mosaic, blurring or blurring will result.
AI-based mosaic detection can be used to improve accuracy.
The red dot indicate the areas judged to be the most important, and if it is on the face or on the
background, good results will not be obtained.
AI-based human body detection can be used to give priority to the human body (excluding the face).

Adjusting parameters with the slider can reduce detection errors, but not significantly.
It is time-consuming and should be used when you want to "improve a scene that is about to break down
as much as possible".

*The process cancel key (default: M key) is useful for comparing mosaic and grid.
*If you hold down the Shift key while opening the Processing Settings panel, the mosaic analysis results
will not be displayed.
Processing Settings by section
Normally processing settings are applied to the whole movie, but you can also create sections by placing
markers.
・To reduce misidentification of the mosaic area, set the accuracy for each scene.
・Manually designate only the part where the mosaic roughness can not be measured correctly.
Setting the details as above will give you more effect.
*To see the result of automatic measurement about roughness of mosaic, turn on Disp infos in
environment settings.

Move the slider handle to the position you wish to divide and then press the Add Marker button on the
toolbar or the shortcut key ('A' by default) to create a section.

Move the slider handle to section 2 and change the parameter.

If you copy('C') in section1 and paste('V) to section3, only section2 will be set differently.
Recording Settings updated!

Delete working folder : If 'Before recording' is selected, the converted image will not be deleted even if
the recording is completed.
This is useful for finding the cause of troubles and re-recording from the middle.
Output video naming rule : You can add the recording mode (used model) to the video file name.
"^WM"(Can be changed in JavPlayer.ini) is added to the end of the file name to trigger the watermark
removal process.
It can also be defined by the user with "JavPlayer/rec_name_format.txt".
External program execution mode : If Console is selected, You can check the progress of tgmain.exe,
ffmpeg.exe, etc.
If the windows are in the way, minimize them, hide them, or limit to one.
Log file output level : If "Error and progress" is selected, the time required for each process of recording
will be output.
I recommend that you enable it because the effect on the recording speed is minimal.
Encoder and Codec : Normally FFmpeg/H264 is used for encoding.
NatCorder is an older encoder, and is present in case of problems with FFmpeg.
Even if you select an encoder that is not compatible with the recording method, it will automatically
switch to a close match, so there is no problem.
HEVC or AV1 will save size with higher picture quality, but will take longer to record.
Please note that encoding AV1 on RDNA3 (Radeon7xx0) will result in FHD videos 1920x1082.
Recording quality : The default (5) will still result in a higher bit rate than a typical mp4.
If FFmpeg is selected as the encoder, the right end is MAX (almost lossless), but the size of the video file
created is huge, so it is not suitable for saving.
Also, FFmpeg's quality setting is variable bit rate, so videos with little motion will have smaller file sizes.
Frame rate of the video for fast forwarding : In 1pass and 2pass, extraction and encoding are
performed frame by frame.
If your PC is not powerful enough, 60 may be better than 120.
If the cutout or seek fails, you may be able to solve the problem by selecting 30(leftmost).
Frame feeding is performed using the original video, which is slower but safer.
VR videos (usually 60 fps) always use the original video and this setting is ignored.
Skip processing automatically : Skips scenes with missing mosaics and avoids unnecessary processing.
It can reduce processing time and video degradation, but may ignore hard-to-see mosaics.
Can be used with manual skip (set 0 to mosaic cell size in section settings to turn off processing).

Destination folder for movies : The default is “JavPlayer/Movie”.

Set it if it is troublesome to move the created video.
It is necessary to create a folder in advance and specify it with a full path such as “D:/JavMovie”.
Processor to be used for tgmain : Select which processor will perform Super-Resolution by tgmain.exe.
If your PC is equipped with a video card, you will usually select GPU0.
Processor to be used for Super-Resolution : GPU ID of DirectML(tgmain) may differ from usual.
If you have one GPU, GPU0&GPU0 is fine, but if you have two GPUs, you may need to select GPU1&GPU0
to specify the same GPU.
Maximum parallel number of tgmain by CPU : The actual parallel number is limited by capacity of main
memory.
The main memory usage varies depending on the size of the image to be processed
Maximum parallel number of tgmain by GPU : The actual parallel number is limited by capacity of
VRAM and main memory.
The VRAM usage varies depending on the size of the image to be processed.
For Super-Resolution other than tgmain, the parallel number is controlled by batch files (SR-*, SRA-*)
regardless of this setting.
VRAM capacity of the GPU for Super-Resolution : Incorrect settings may result in insufficient VRAM or
inability to perform parallel processing.
Refer to the manual of the installed video card for the correct setting.
VRAM used by other than Super-Resolution tools : Reserve VRAM for use by other apps.
Larger values are safer if other work is to be done during recording, but the parallel number is reduced.
It is recommended to check the VRAM usage with GPU-Z before start recording.
TecoGAN installation folder : The default is “JavPlayer/TG”.
Set this if you do not want to write a large amount of data on the drive where JavPlayer is installed.
It is necessary to create a folder in advance and specify it with a full path such as “D:/JavTG”.
TGMAIN and TG-MODEL need to be moved to the specified folder.

From Ver. 1.10, you can run any batch file instead of tgmain.exe to perform Super-Resolution processing.
In this case, some of the above settings will not be reflected. *Depends on the contents of the batch file.
Parallel execution of processes is applied by settings in the batch file, but depending on the Super-
Resolution tool, parallelization may not be faster or VRAM usage may be too large to parallelize.

In addition, Additional Super-Resolution (SRA-*) processes the image again after enlarging it four times
in width and height (16 times in area).
You will often need to have at least 6GB of VRAM to parallelize.
Only those of the following items that are required in the current configuration will be displayed.
Processor to be used for encoding : Using a GPU does not significantly reduce video quality.
Depending on the performance of your GPU, it may be slower than the CPU, or it may not support
encoding.
image format (for Extraction and Encoding) : The file size of png is about x5 of jpg.
When the format of the images for extraction (raw*) is set to png, accuracy of the mosaic detection and
the video quality slightly increases.
When the format of the images for encoding (blend*) is set to png, the video quality slightly increases.
When encoding with FFmpeg in 1pass and 2pass, full screen extracted images are not output.
There is no degradation because the image is processed in memory, and the result is the same as png.
I recommend png&jpg, but choose png&png if quality is important and jpg&jpg if speed is important.
For normal recording (encoded with the new method), the result is the same as png&png.
Keyframe interval : The shorter the file size, the faster the seek or the more accurate the seek, making it
easier to handle, but at the cost of a larger video file size.
Number of frames for partial encoding : FFmpeg creates a video from still images, but since doing it all
at once requires a huge amount of free space, it is divided into small segments and encoded.
Once encoding is complete, the used still images are deleted or moved for storage.
Lesser numbers will address the lack of RAM disk space, but will reduce efficiency.
As for Safe, it is fixed at 1800 frames.
If the video is FHD and the number of parallels (see below) is 2 or less, it will fit in 4GB.
If the format of the images for encoding is jpg, 1GB will be sufficient.
Number of threads for generating images for encoding : Create images (png,jpg) before encoding in
a multi-threaded way to speed up the process.
The efficient thread number depends on the CPU and is also affected by other running applications.
If you are using a fast CPU, try a value greater than 4.
Number of processes for generating images for encoding : Parallelize JavPlayer by launching it
multiply in image generation mode for encoding.
Parallelize in folder units in the same way as Super-Resolution process.
Too many parallels may slow down the encoding of another process.
This setting is not reflected if 3xSR or SPAN* is used, as parallel processing is not possible.
Data transfer method to FFmpeg : Select Pipe (interprocess communication) or ImageFiles (export to
temporary folder).
Pipe (Async) is the default, but if you encounter problems such as “frames are swapped back and forth,”
select Pipe (Sync).
*The old method of encoding via image files is retained in case problems arise with Pipe.
Temporary folder for image files for encoding : Image files in this folder will be deleted in a short time.
Allocating RAM disk can reduce damage to SSD or HDD.
In Safe, the write destination switches to the work folder when there is insufficient space.

Storing image files used for encoding : When turned ON, a folder with the same name as the video is
created in the video destination and the images generated for encoding (synthesized super-resolution
results) are saved.
*Note that not only does it require a very large amount of free space, but it is not automatically deleted.
Recording

Only the area set in the range will be recorded, but if no range is set, the 5 seconds from the current
playback point will be recorded.
If you do not enter Width, the recording size will be the same as the original movie.

For the recording method (center list), usually select 1pass or 2pass.
1pass: The processing other than super-resolution is done in memory and is the fastest.
2pass: Almost same as 1pass, but mosaic detection rate is increased by dividing the extraction process.
The time required for extraction is twice that of 1pass, but it may not be worth it.
Safe: Used when 1&2pass cannot record properly.
Since all frames are written to the drive before various processes are performed, not only is this time
consuming, but it also requires drive space and endurance.

When using 1pass and 2pass, the original movie is clipped as "4FF_*.*" at 4x speed for fast frame-by-
frame forwarding.
It remains after recording for the next time, but is deleted when recording another video.
*If the original video is avi format, the extraction will be slower due to the inability to fast-forwarding.
In the rare case of wmv or mkv, seek of the clipped video may fail and the first step not proceed.
You can fix this problem by shifting the beginning of the recording range slightly or by changing the
recording method to "Safe".

If Safe is selected as recording method or if FFmpeg is specified for Encoder in the Recording Settings, all
still images are exported and then encoded with FFmpeg.
The encoding options for this can be changed in JavPlayer/ffmpeg_enc_opt.txt.

What model (right list) should be selected depends on the super-resolution tool used.
See TecoGAN section.

The default output destination is "JavPlayer/Movie/OriginalName _*.mp4".

During recording, the frame rate is changed regardless of usersetting.

When the recording mode is 'TG- *', you can suspend and resume.
Select resume point from the dialog that appears after pressing the resume button.
The items displayed in the dialog vary depending on the recording method and timing of suspend.
Even after restarting JavPlayer and loading another video, you can resume it by reloading the last video
used in TG mode.
Note that only one information for resumption is saved and will be overwritten.
To re-encode with other settings (size, bit rate, etc.) after recording is completed, it is necessary to switch
'Preferences-ADVANCED-TecoGAN-Delete work folders' to 'Before recording'.
If “RecSettings - FFmpeg - Data transfer method to FFmpeg” is set to Pipe and encoding is interrupted,
the video will be created only with the completed part and cannot be restarted from the middle.

Specify the Super-Resolution tool for the first time on the left and the second time on the right.
The default is tgmain&NONE, which uses tgmain.exe to perform the Super-Resolution process only once.
If you select *3xSR,*SPAN4X,*SPAN4Xhd in the list on the right, JavPlayer's internal processing will
perform the second super-resolution.
This processing is done in the encoding phase, not the super-resolution phase, and using SPAN4X in an
environment with low GPU performance will result in very long recording times.
Also, when external tools are installed, the corresponding batch files (SR-*.bat, SRA-*.bat) will be added
to the list.
SR-* is used instead of tgmain for the first super-resolution, and SRA-* is used for the second.
Unlike SPAN, SRA-* creates still images (png, jpg) in the super resolution phase.
For more information, please refer to the section of Batch file for Super-Resolution.

To compensate for the lack of detail, add noise or sharpening to mimic skin textures.
It may reduce the discomfort of the processed area.
If *ReMos or #LnSnTn is selected in the Additional Super Resolution list, it is disabled and the slider will
not appear.
Sequential recording
You can register multiple videos for recording in advance.
Press E(Entry) key to register the recording with the current settings .
Press Q(Queue) key to display the queue panel .
*The processing settings and color correction are applied to the video, not the entry.
*Each key can be re-assigned.
You can delete entries with the trash button in the queue panel.
Queue execution can be suspended, and where it resumes is determined automatically.
Select the action to be taken on completion from the drop-down list in the lower left.
You can also register a group of videos (see "Sequential Playback") at once by pressing Shift+E, but you
cannot specify a range.
VR Recording
By setting the projection type to Equirect or FishEye in Screen Settings②, you can split the screen into left
and right to process VR video.
Older videos with left and right screens arranged vertically are not supported.
If the horizontal resolution of the video to be created exceeds 4096, the codec must be HEVC or AV1.

Equirect is equirectangular format, which in simple terms means that the image is transformed into a
square.
There are two types of mosaic: one is deformed after adding mosaic, and the other is added after
deformation.
In the former case, the mosaic is gaudily distorted (curved) at the bottom edge of the screen, etc.
To compensate for this, you need to enable conversion in Processing Settings ⑪.
Usually None or 180deg (FoV), but some can be converted correctly with a slightly narrower angle.
You can check the converted status on the playback screen, but only one of the left or right side can be
displayed. *Right-click on the screen to switch.

FishEye is not deformed as a round image and has margins at each of the four corners of the left and
right screens.
The mosaic is rarely distorted and usually yields better results than the equirectangular format.

The recommended Encoder/Codec is FFmpeg/HEVC to reduce file size while maintaining the quality of
the video you create.
However, FFmpeg requires outputting still images of all the frames, which results in a huge amount of
writing to the drive.
It is recommended to place temporary folders on the RAM disk, since the writing and deleting of still
images will be repeated for at worst 1 TB.

Additional super-resolution with *SPAN or *SPANhd will require more VRAM than normal video.
Depending on the coarseness of the mosaic, 8 GB or more should be sufficient.

When “Recording method is Safe, Encoder is FFmpeg/H264, GPU (NVENC) is used”, an error will occur
if the width of the video to be created is larger than 2900 pixels.
The solution is as follows:
- Change the Encoder/Codec to FFmpeg/HEVC.
- Change the Processor used for encoding to CPU.
TecoGAN(tgmain)
It is a video Super-Resolution algorithm developed by MengyuChu, YouXie, LauraLeal-Taixe, and
NilsThuerey.
https://github.com/thunil/TecoGAN
It requires a CPU (SandyBridge, Bulldozer or later) that can use the AVX instruction set for execution.

If you use a GPU that supports DirectX12 (with at least 2GB of VRAM), Super-Resolution will be several to
several dozen times faster.
Switch 'ADVANCED-TecoGAN-Processor to use' to GPU0.

There is SuperResolutionTest button on the upper left corner of the startup screen.
With this, conversion speed of TecoGAN can be measured.
It is also useful for checking whether tgmain.exe can be executed normally.

TG-AF0, AF1, and AF2 in the model list of recording settings are Anti-Flicker models.
The effect increases with the number, but there is a disadvantage of blurring.
In addition, a noise filter is applied to the reduced image subject to Super-Resolution.
This noise filter is valid even when TecoGAN is not used (SR-* is used).
The strength of the noise filter is 0-9, and default is 3.
Normally, the default is fine, but you can adjust it by renaming the files (*.data-00000-of-00001,*.index)
in TG/TG-MODEL.
Rename TG-AF2.* to TG-AF2-5.* if you want to make the noise filter stronger.
To disable the noise filter for AF1, rename TG-AF1.* to TG-AF1-0.*.
If you want to use them together, copy and rename them.
Setting the strength to 1 makes the filter closer to the noise filter of Ver. 1.11 or earlier.
If you want the filter to be exactly the same as the old filter, specify something other than a number, such
as TG-AF2-X.
The noise filter is applied at the time of reduced image extraction.
Therefore, if you select a model and change the noise filter strength, the recording process starts over
from the beginning (extraction).
It can be resumed after changing from TG-AF1 to TG-AF2, but if you change to TG-AF1-2 or TG-STD, you
will start from the beginning.

In Ver. 1.10 or later, you can use other Super-Resolution tools instead of TecoGAN.
The use of TecoGAN in JavPlayer has become less meaningful now, because JavPlayerEZ has achieved
real-time super-resolution with TecoGAN.
If you want to improve the quality, try TecoGAN8X or BasicVSR++ on the download page.
Also refer to Batch file for Super-Resolution described below.
Additional Super-Resolution updated!
Super-resolution results from TecoGAN, etc. can be further super-resolved.
There are two ways to do this: using JavPlayer's built-in functions or using an external tool via a batch file.
The former is performed during the encoding phase, and the latter during the super-resolution phase.
The external tools (whose names are prefixed with “SRA-” in the list) are difficult to install because they
require a separate installation, and they write the results as image files, so they require free space and
durability of the drive.
On the other hand, built-in types (whose names are prefixed with “*” in the list) can be used easily.

＊3xSR
A very fast super-resolution algorithm that magnifies still images by a factor of 3.
The effect is weak, but it produces stable results with less noise.
It's like a "high-performance sharpening filter," so if you want to reduce blur, this is the one for you.

*SPAN4X
Use a high-quality super-resolution model that magnifies still images by a factor of 4.
It has the characteristics to smooth out jagged lines and remove noise, and can be expected to be very
effective.
Since detail tends to be lacking, I recommend using with Noise & Sharpen filter.
*In environments with less than 4GB of VRAM, errors may occur due to insufficient VRAM.

*SPAN4Xhd
HD (high detail) version of SPAN4X.
It has the property of adding details that are not present in the input image, and can be expected to have
a significant effect.
Blended with 3xSR super-resolution results created at the same time to reduce noise (excessive detail).
*In environments with less than 4GB of VRAM, errors may occur due to insufficient VRAM.

＊ReMosS,L
It is not super-resolution, but since the categories are the same, it is added to the list.
Apply a finer mosaic to the processed area than the original video.
With S, each rectangle is smaller and with L, each rectangle is larger.
#LnSnTn
This is another filter that does not do super-resolution, but clarification and noise addition.
The names indicate the function of the filter and the strength of each, and the breakdown is as follows
Line...Emphasize dark lines without making them thicker.
Surface...Sharpen details on flat surfaces.
Texture...Add skin patterns to reduce CG-like appearance.
If you rename #L3S3T3 to #L9S0T0 in the batch folder (TG/BATCH), the filter will only emphasize lines
to the maximum extent.
Batch file for Super-Resolution
You can use any batch file and executable instead of tgmain.exe (TecoGAN executable for Windows),
which is distributed as a standard external tool for super-resolution in JavPlayer.
Additional super-resolution can be also done with external tools in same way.
*I currently distribute only the super-resolution tool, as there is no additional-super-resolution tool that is
better than SPAN.

Advantate of internal additional-super-resolution

- Easy as no additional installation is required.
- No need to consider drive space or durability.
- Processing is fast.

Advantate of external additional-super-resolution

- Any executable file for command line can be used. *Need to write a batch file.
- Since the output image file remains, it can be resumed from the middle of the super-resolution phase.
- After the encodng, you can change codecs, etc. and encode again without additional-super-resolution.

If the latter is used when the recording method is Safe, the set value for Parallel number of image
blending is ignored and set to 1.
This is a specification to avoid VRAM shortage.

Supplemental Description
Images before super-resolution often have a width of 100 to 200 pixels, and the following is an example
of a 150 case.
Before super-resolution: 150 pixels
After first super-resolution (4x): 600 pixels
After additional super-resolution (4x): 2400 pixels
Since it exceeds the size of FHD, it is a little wasteful to reduce the size of the image when compositing.
Using TecoGAN8X, which is distributed separately, the result is 1200 after the first super-resolution and
the final result is 4800.
When using external tools, we are concerned about the life of the drive as it will be writing a large number
of still image files in excess of 4K.
Even if this is done by internal processing, it not only takes an enormous amount of time, but may also
result in a shortage of VRAM.
Please take into account that "more than 16 (4*4) times is wasteful" when setting about super-resolution.
Typical VR video is 4K (4096 wide), which is split and processed on the left and right sides, so the width is
2048, which is equivalent to FHD.
Debug Log

Press the Log Display button in Basic Operation (15) or press Ctrl+MouseL+MouseR to switch to this
screen.
Progress, warnings, errors, and other information can be viewed and exported.

① Outputs the entire log and details about the currently selected line (what is displayed at the bottom) as
a text file (*.txt).
② Outputs the currently visible screen as an image file (*.png).
③ Exit log display.
It is a slightly modified version of an asset called Log Viewer, and I don't know how to use it in detail, but
the scroll bar on the far right and the three buttons above are enough to use it.
The output files are saved in "JavPlayer/Log/".

If the recording process has stopped, for example, you may be able to find the cause by looking at the
end of the log.
I would also appreciate it if you could help me by tweeting with the saved images so that I can make
progress on the revisions.
Auxiliary Options updated!
Options that you may not need to change often are set in a text file (JavPlayer.ini).
The following is an explanation of some of the more obscure items.

AllowOverhang: Allow the window to protrude from the screen when opening a video.
Normally false, if a window exits out of the screen, it will correct its position before starting up.
If the position correction is disturbing, such as in a multi-display environment, set to true.
If the window is not visible, use Shift+Escape to move it to the center of the screen.

WatermarkedFile: Videos that contain the specified string in the file name will be targeted for watermark
removal.
This string will be added to the name of the video you create, and the watermark will not be visible when
played back in JavPlayer.
Only watermarks added by JavPlayer Ver.2.00 or later can be removed.

WatermarkedFolder: Videos in folders containing the specified string will be targeted for watermark
removal.
Subfolders are covered as well.
Multiple strings can be set by separating them with commas.
If WatermarkedFile is left blank, the videos will not be distinguishable by their file names, so please
move them to a folder with the name specified here.

MosaicDetectionAI: Mosaic detection by AI will be enabled by default.

Default is OFF, but if your PC is equipped with a GTX1060 or higher GPU, you can turn it ON all the time.
There will be few cases where OFF will be better than ON.

HumanDetection: Human body detection will be enabled by default.

I do not know how effective it is because it is not accurate and not well tested.
It is recommended to turn ON only when recording videos with mosaic processing on the face.

SpanVramLimit: If less than 100 is specified, *SPAN (additional super-resolution by internal processing)
is performed without using the full of estimated free VRAM space.
Specifying a smaller value will increase the number of divisions in the tiling process and reduce efficiency.

UseOldASR: The specification is to exclude areas without mosaic and perform additional super-resolution
for internal processing, but the target will be changed to the entire screen.
So far I have not been able to identify any “problem with some part not getting additional super-
resolution”, so there is no need to be true.
Trouble Shooting

TROUBLE : Encoding stops with the message “FFmpeg terminated with an error”.
CAUSE : ffmpeg.exe fails to encode.
SOLUTION : Change “RecSettings - FFmpeg - Processor to be used for encoding” to CPU.
: Change ”RecSettings – FFmpeg – Data transfer method to FFmpeg” to ImageFiles.

TROUBLE : An error occurs in encoding and processing does not proceed.

TROUBLE : The addition of audio fails and a silent video is generated.
CAUSE : FFmpeg cannot be written to the movie folder.
SOLUTION : Change the access permission settings for the folder.
: Set 'ADVANCE-Common-FolderInputField' to change the save destination.
: Move the JavPlayer folder to another location (drive).

TROUBLE : When the recording method is 1pass or 2pass, cutting out video for frame feeding fails.
CAUSE : Unknown *The video that should have been cut out does not exist.
SOLUTION : Change "RecSettings – General - Frame Rate of the video for Fast Forwarding" to 30.
: Change the recording method to Safe.

TROUBLE : When the recording method is 1pass or 2pass, seek fails and recording does not start.
CAUSE : The video cut out for frame feed is not compatible with the decoder used.
SOLUTION : Slightly shift the start of the recording range.
: Change "RecSettings – General - Frame Rate of the video for Fast Forwarding" to 30.
: Change the recording method to Safe.

TROUBLE : Super-resolution fails and an error message appears on the screen in red text.
CAUSE1 : Security software prevents creation/deletion of work folders.
SOLUTION1 : Turn off security software or add JavPlayer to the exclusion list.
CAUSE2 : For some reason, the Super Resolution Tool executable file cannot be accessed.
SOLUTION2 : Move the JavPlayer folder to another location (drive).

TROUBLE : Recording process will not proceed.

CAUSE: Unexpected (message not prepared) error.
SOLUTION : Display the debug log (Ctrl+MouseL+MouseR) and look around the end.
: Look for 'Error' or 'Exception' and deal with it if you find the cause.
: If it is impossible to deal with the problem, report it on Twitter, etc.
TROUBLE : The mosaic is not detected and not processed.
CAUSE : The quality of the target video is poor.
SOLUTION : Get high quality video.
: Turn on the high resolution switch and AI switch on the processing settings panel.
: Move the Picture Quality slider in the Processing Settings panel to the center or left.

TROUBLE : The screen turns white during Super-Resolution process using GPU.
CAUSE : VRAM is insufficient.
SOLUTION : Set the memory capacity and the number of parallels of 'ADVANCE-TecoGAN' appropriately.
: Increase the value of 'ADVANCED-TecoGAN-VRAM capacity used by other than TecoGAN'.

TROUBLE : Rewinding or fast-forwarding continues even though no operation has been performed.
CAUSE : Input from pads etc.
SOLUTION : Calibrate the input device.
: Erase the pad input setting (right column) by key assignment.

TROUBLE : Failed to Super-Resolution and the error message "AttributeError: module 'pandas' has no
attribute 'compat'" appears in the command prompt.
CAUSE : Blocked by ImDisk (RAM disk).
SOLUTION : Create RAM disks with other tools.
: Leave "Advanced Recording Settings - FFmpeg - Temporary Folder" blank.

TROUBLE : After the error occurs frequently in Super-Resolution process, "The number of files in input*
and output* do not match" is displayed.
CAUSE : VRAM is insufficient.
SOLUTION : Check if 'ADVANCED-TecoGAN-VRAM capacity of the selected GPU' is set correctly.
: Increase ' ADVANCED-TecoGAN-VRAM capacity used by other than TecoGAN'.
: Decrease 'ADVANCED-TecoGAN-Number of parallel...'.
*If Number of parallel is 1, it works normally even if the two VRAM settings are wrong.

The Twitter account @Javski2 is used for announcements and answering questions.