- 2023.1.21 This commitment has been fulfilled The Windows version has been added, the plug-in has been submitted for review, and the case has been updated
- Download it at the following address
- https://1drv.ms/u/c/ccd0ebab83bc10d3/EdMQvIOr69AggMz5BQAAAAABPDr9cZnnYvn_wKfY41XANg?e=gwnxxO
This is a good idea, I will record it in my development document and implement it as soon as possible in the next plan
- 2023.2.4 This commitment has been fulfilled,We provide the similarity algorithm, which can return the similarity with a keyword
This is in the next Plan, which is to support almost all languages (you may need to train your own language model, of course, to make the process as easy as possible).
Provide accurate offline voice-to-text services for VR,AR and Android platforms, such as oculus quest1/2/pro or pico3/4
The plug-in supports UE 4.26/4.27/5.0/5.1, and since the store only allows submission of the last three versions, you can compile the plug-in yourself if you need other versions
1.支持两种语音离线识别(英语和汉语)
- Support two kinds of offline speech recognition (English and Chinese) 2.简单,强大,创作者可以用几个蓝图函数轻松调用它
- Simple and powerful, creators can easily call it with several blueprint functions 3.不依赖任何其他服务,调用插件,打包android arm64就可以运行
- Do not rely on any other services, call plug-in, package android arm64 can run 4.为VR耳机开发,所以它超适合 oculus quest1/2/pro or pico3/4 我会提供一个VR一体机例子
- Developed for VR headsets, so it's super suitable for oculus quest1/2/pro or pico3/4 I'll give you an example of a VR all-in-one
How To Use:
- Build--Only Choose Support arm64 [aka arm64-v8a]
- Advanced APK Packaging--Add permissions to support Voice chat (RECORD_AUDIO)
- Param: Alternate parameter, can be random floating point number
- info = "1": Initialization model(Don't repeat calls,It is best to initialize the plugin in the level blueprint)
- info = "2": Initialization audio(Don't repeat calls,It is best to initialize the plugin in the level blueprint)
- info = "3": The interface is reserved. You are advised to use info=6
- info = "4": The interface is reserved. You are advised to use info=6
- info = "5": Reset the text of the model. After the call, the model returns the text to zero
- info = "6": Read speech and convert text in a thread in real time,Calling it again terminates the thread
- getJvoicePointer() gets an object that can bind real-time speech-to-text results to events
- Complete process, start real-time voice to text
2. the model returns the text to zero
3. Stop converting speech to text/Start converting speech to text
-
Call Jvoice WindowsInit to initialize
-
Invoke Jvoice OpenExe to start the voice stream server
-
Make sure the exe path does not contain any Spaces
-
Here you can download the Voice server exe package, unzip it, and then pass the absolute path of the exe to the function https://1drv.ms/u/c/ccd0ebab83bc10d3/EdMQvIOr69AggMz_BQAAAAABDZLAcV4t6P2x_UZiZwXIkw?e=3blYbw
-
To ensure that the latest audio is used every time we start, we execute Windows_Exit before WindowsInit,You can also skip Windows Sexit if you're sure you've logged out of the Windows server
- Call Jvoice_WindowExit to end the voice service when you don't need it or before you end the game
- Like Android, you can receive real-time data here
- Just fill in the keywords and the text similarity is returned
- Converts the Chinese in the string to Pinyin/把字符串中的中文转为拼音
- If you are a blueprint plug-in, you need to create a C++ class
If it is UE5.0 or below, create a new C++ class in the file TAB
- Visual studio opens after you create a new C++ class You can select debug directly, if there is a problem you can right-click the project and select rebuild solution
- If you are using the oculus branch of Unreal Engine, please send me an email with the Shopping Mall plugin purchase page and I will reply with a version that supports the oculus branch of Unreal Engine(Email: 1146094743@qq.com)
- The Unreal Engine of the oculus branch does not seem to support conditional compilation, so I have specially prepared a VR example of the oculus branch for you and our corresponding plugin
备注: 本项目的kotlin部分来自开源项目https://github.com/k2-fsa/sherpa-ncnn