How to use TALQu
What you need to install in advance
・NET Framework 4.8
NET Framework 4.8 is installed as standard in the latest version of Windows 10, so no new installation is required.
If you have an older version installed, you may want to update it. NET versions have not been tested.
Microsoft Visual C++ 2015-2019 Redistributable (x64) - 14.24.28127
If you are using the GPU version, you will also need to add the following installation.
・CUDA
・cuDNN
There are CPU and GPU versions of TALQu, and you will have to select one of them when installing.
Unless you have a gaming PC with a fine graphics board, you can choose the CPU version.
Installation of TALQu
It is distributed on BOOTH. (Free DL, and you can donate)
https://haruqa.booth.pm/items/2755336
The following is a quote from the description of BOOTH. (As of July 17, 2021)
How to use (TALQuManager)
1.Launch
Extract the zip and start "TALQuManager.exe" in the TALQuManager folder inside the TALQu folder.
Do not change the folder structure, as the files will be extracted to a folder one above the TALQuManager folder.
2. download and setup
Select the CPU version or GPU version and click the「ダウンロードとセットアップ」(Download and Setup) button.
You will see several black screens, wait for a few minutes or tens of minutes, and then you will see the completion dialog.
3. launch TALQu
Go back to one folder above and check "ReadMe.txt".
How to use (TALQu)
1. start
Please start "TALQu.exe".
2. playback
Enter the sentence you want to be read out loud in "Text to read" and click the "Play" button.
It takes some time to compose and play long sentences at once.
It is recommended to save the wav file and play it with an external player.
3. stop
If you want to stop the playback in the middle, click the "Stop" button.
4. saving wav
Click "Save as wav" to save the synthesis result as wav.
If you want to stop the synthesis, click the "Stop" button.
If you want to save the file as it is, please use the "Save the voice with
adapted parameters." button.
If "Export text file" is checked, a text file with the name will be output in the same location as the saved wav.
Usage Tips
Correcting the pronunciation
Once played, the pronounced hiragana and alphabet will be output based on the guessed "sentence to be read".
You can change the "pronunciation hiragana" as you wish.
Enter the postposition "わ(wa)" for "は(ha)" and "え(e)" for "へ(he)" as you pronounce them.
You cannot use the Hyphen or long vowel symbol "-". Please enter the corresponding vowel.
You can also change the "Pronunciation Alphabet" as you wish.
Please note that it is using its own pronunciation alphabet, not the normal Hepburn alphabet.
Shortcut keys
Ctrl+E: Play
Ctrl+R: Play again
Ctrl+S: Save what you just heard
Ctrl+W: Save wav
Setting items to default values
You can restore the default values by double-clicking on the numeric parts of speech speed, sigma, denoiser strength, small pause length, large pause length, pitch height, and tone height.
Settings
Speech speed: 100 is the default, 50 is 1/2 speed, 200 is 2X speed.
Sigma: 0.666 is the default, lowering it may reduce the aperiodic sound. Increasing the value increases the aperiodic sound. Please note that extreme values will result in noise.
Denoiser: 0.1 is the default, lowering it may increase the noise. Raising it reduces noise.
Small pause length: Corresponds to the length of "、(,)".
Large pause length: Corresponds to the length of "。(.)”
Voice height: 1 is the default, 0.5 is one octave lower, 2 is one octave higher
Voice height: 1 the default, lower for male voice (older), higher for female voice (younger)
How to add a model
You can add a model by adding a whole folder in the Models folder in the TALQu folder.
After restarting TALQu, you can select a model from its name.
If the model does not appear in the list, please check if the model folder is placed correctly or check Config.csv in the model folder.
Beta features
By supporting Haruqa's Fanbox, you can use the TALQu editor with beta features.
(About 500 yen≒$5)
https://haruqa.fanbox.cc/
Reference
この記事が気に入ったらサポートをしてみませんか?