Awesome

Next Gen Dialogue User Manual

Read this document in Chinese: 中文文档

Content

• Features
• RoadMap
• Supported version
• Install
• Quick Start
  • Create a Dialogue Tree
  • AI Bake Dialogue
• Nodes
• Modules
  • General Modules
  • AIGC Modules
  • Experimental Modules
    • Localization Extension
    • VITS Speech Extension
• Experimental Function Introduction
  • One-click Translation
  • Bake Voice
• Resolvers
  • How to Switch Resolver
• Create Dialogue by Code

Features

Next Gen Dialogue plugin is a Unity dialogue plugin combined with large language model design, won the Unity AI Plugin Excellence Award from Unity China. It combines the traditional dialogue design method with AI technique. Currently this package is an experimental attempt and hopes you enjoy it.

It has the following features:

1. Visual dialogue editor
2. Modular dialogue function
3. Support AIGC baking dialogue in Editor
4. Debug during runtime

Demo project: https://github.com/AkiKurisu/Next-Gen-Dialogue-Example-Project

Demo video: https://www.bilibili.com/video/BV1hg4y1U7FG

Supported version

• Unity 2021.3 or Later

Install

Using git URL to download package by Unity PackageManager https://github.com/AkiKurisu/Next-Gen-Dialogue.git

The experimental features of Next Gen Dialogue are placed in the Modules folder and will not be enabled without installing the corresponding dependencies. You can view the dependencies in the README.md document under its folder.

To use core functions, you need to install Newtonsoft Json in PackageManager.

Quick Start

If you are using this plugin for the first time, it is recommended to play the following example scenes first:

1.Normal Usage.unity this scene contains the use of NextGenDialogueTree and NextGenDialogueTreeSO;

2. Editor Bake Dialogue.unitythis scene contains the sample of baking conversation conversation in the use of AI dialogue Baker in Editor;

3.Build Dialogue by Code.unity this scene contains the use of Code to generate dialogue.

4.Bake Novel.unity An example of using ChatGPT to infinitely generate dialogue trees.

Create a Dialogue Tree

NextGenDialogueTree and NextGenDialogueTreeSO are used to store dialogue data. In order to facilitate understanding, it is collectively referred to as dialogue tree. The following process is to create a dialogue tree that contains only a single dialogue and a single option:

1. Mount NextGenDialogueTree on any gameObject

2. Click Edit DialogueTree to enter the editor

3. Create the Container/dialogue node, the node is the dialogue container used in the game

4. Connect the Parent port of the dialogue Node to the root node. You can have multiple dialogue in one dialogueTree, but only those connected to the root node will be used.

5. Create the Container/Piece node and create our first dialogue fragment

6. Right -click Piece node Add Module add Content Module, you can fill in the contents of the conversation in Content

7. Create a Container/Option node and create a dialogue option corresponding to the PIECE node

8. Right-click Piece node Add Option, connect Piece with Option

9. <b style = "color:#ee819e"> Very important: </b> At least one Piece node needs to be added to the Dialogue as the first piece of the dialogue.You can right -click dialogue's Add Piece to connect with the connection or quoting its PieceID. You can also right -click dialogue's Collect All Pieces to add all the piece in Graph to the dialogue and adjust the priority of the Piece

   • For priority, please refer to 《General Module-Condition Module》
   <Img src = "Docs/Images/CreateDialogue.png">

10. Click on the upper left of the editor's Save to save dialogue

11. Click Play to enter PlayMode

12. Click on NextGenDialogueTree Play dialogue to play conversation

13. Click Debug DialogueTree to enter the debug mode

• <span style = "color:#f8d09d"> Tips: </span> The currently played dialogue piece will be displayed in green

AI Bake Dialogue

You can use AI dialogue Baker to bake the dialogue content generated by AI in advance when designing the dialogue tree, so as to improve the workflow efficiency without affecting your design framework.

1. The basic dialogue tree design is consistent with the process of 《Create a Dialogue Tree》
2. Add AI Bake Module for the fragments or options that need to be baked, and remove the module for nodes that do not need to be baked
3. Select the type of LLM you are baking with
4. <b>Select in turn</b> the nodes that AI dialogue Baker needs to recognize, the order of recognition is based on the order selected by the mouse, and finally select the nodes that need to be baked
5. If the selection is successful, you can see the preview input content at the bottom of the editor
6. Click the Bake Dialogue button on the AI Bake Module and wait for the AI response
7. After the language model responds, a Content Module will be automatically added to the node to store the baked dialogue content
8. You can continuously generate conversations based on your needs

AI Generate Novel

Different from talking directly to AI in baking dialogue, novel mode allows AI to play the role of copywriter and planner to write dialogue, so it can control options and fragments more precisely. Please refer to the example: 4.Bake Novel.unity

Nodes

NGD use node based visual editor framework, most of the features are presented through nodes.

The construction dialogue are divided into the following parts in NGD:

In addition, in order to add interest to the dialogue such as adding events and executing actions, you can use the following types of nodes in the behavior tree framework in NGD:

Modules

In addition to the above nodes, a more flexible concept is used in NGD, that is, Module. You can use Module to change the output form of the dialogue, such as Google translation, localization, add callbacks, or be executed as a markup.

General Modules

The following are built-in general modules:

AIGC Modules

The following are the built-in AIGC modules:

Experimental Modules

The following are experimental modules, you need to install the corresponding Package or configure the corresponding environment before use:

Localization Extension

Based on the UnityEngine.Localization plugin to support the localization of dialogue

VITS Speech Extension

For VITS local deployment, please refer to this repository: VITS Simple API

If you want to use the VITS module, please use it with VITSAIReResolver. For the use of the Resolver, please refer to the following 《Resolver》

Editor Function Introduction

One-click Translation

Add Editor/EditorTranslateModule in the Dialogue node, set the source language (sourceLanguageCode) and target language (targetLanguageCode) of the translation, right-click and select Translate All Contents to perform all Piece and Option with ContentModule translate.

For nodes other than ContentModule, if the TranslateEntryAttribute is added to the field, you can right-click a single node to translate it.

namespace Kurisu.NGDT.Behavior
{
    public class SetString : Action
    {
      //Notify field can be translated
      //* Only work for SharedString and string
      [SerializeField, Multiline, TranslateEntry]
      private SharedString value;
    }
}

Bake Voice

Before use, you need to install the corresponding dependencies of Modules/VITS and open the local VITS server (refer to Modules/VITS/README.md). Add AIGC/VITSModule to the node where speech needs to be generated, right-click and select Bake Audio

If you are satisfied with the generated audio, click Download to save it locally to complete the baking, otherwise the audio file will not be retained after exiting the editor.

It is no longer necessary to start the VITS server at runtime after baking is complete.

• If the AudioClip field is empty, the run generation mode is enabled by default. If there is no connection, the conversation may not proceed. If you only need to use the baking function, please keep the AudioClip field not empty at all times.

Resolvers

Resolver is used to detect the Module in the Container at runtime and execute a series of preset logic such as injecting dependencies and executing behaviors, the difference between NGD's built-in Resolver is as follows:

How to Switch Resolver

1. In-scene Global Resolver You can mount the VITSSetup script on any GameObject to enable AIResolver in the scene

2. Dialogue specified Resolver

   You can add VITSResolverModule to the dialogue node to specify the resolver used by the dialogue, and you can also click the Setting button in the upper right corner of the module and select which Resolvers to be replaced in Advanced Settings

Create Dialogue by Code

NGD is divided into two parts, DialogueSystem (NGDS) and DialogueTree (NGDT). The former defines the data structure of the dialogue, which is interpreted by Resolver after receiving the data. The latter provides a visual editing solution and inherits the interface from the former. So you can also use scripts to write dialogues, examples are as follows:

using UnityEngine;
public class CodeDialogueBuilder : MonoBehaviour
{
    private DialogueBuilder builder;
    private IEnumerator Start()
    {
        yield return new WaitForEndOfFrame();
        PlayDialogue();
    }
    private void PlayDialogue()
    {
        var dialogueSystem = IOCContainer.Resolve<IDialogueSystem>();
        builder = new DialogueBuilder();
        //First Piece
        var piece = DialoguePiece.GetPooled();
        piece.Content = "This is the first dialogue piece";
        piece.PieceID = "01";
        piece.AddOption(new DialogueOption()
        {
            Content = "Jump to Next",
            TargetID = "02"
        });
        generator.AddPiece(piece);
        //Second Piece
        piece = DialoguePiece.GetPooled();
        piece.Content = "This is the second dialogue piece";
        piece.PieceID = "02";
        var callBackOption = DialogueOption.GetPooled();
        //Add CallBack Module
        callBackOption.AddModule(new CallBackModule(() => Debug.Log("Hello World !")));
        callBackOption.Content = "Log";
        piece.AddOption(callBackOption);
        generator.AddPiece(piece);
        dialogueSystem.StartDialogue(generator);
    }
}