Convert text to voice on timeline text track subtitles to wav files with A.I.VOICE for Games

ConvertTimelineTextToVoiceWav.cs

using System.Collections.Generic;
using System.Globalization;
using System.IO;
using System.Linq;
using UnityEngine;
using CsvHelper;
using UnityEditor;
using UnityEditor.Timeline;
using UnityEngine.Timeline;

public class ConvertTimelineTextToVoiceWav : EditorWindow
{
    const string csvExportFilePath = "Assets/CSVTest.csv";
    const string voiceExportPath = "Assets/Voices/";
    const string timelineFilePath = "Assets/Timeline/TestTimeLine.playable";
    const string TEXT_SWITCHER_TRACK_NAME = "Text Switcher Track";
    const string AUDIO_TRACK_NAME = "Audio Track";
    
    [MenuItem("Window/Custom/ConvertTimelineTextToVoiceWav")]
    public static void ShowWindow()
    {
        // EditorWindowを作成
        // 型引数にはこのクラスを入れる
        ConvertTimelineTextToVoiceWav window = GetWindow<ConvertTimelineTextToVoiceWav>();
        window.titleContent = new GUIContent("CsvCreator");
    }
    
    private void OnGUI()
    {
        if (GUILayout.Button("Convert",GUILayout.Width( 200 ),GUILayout.Height( 100 )))
        {
            Extract();
        }

        if (GUILayout.Button("Attach",GUILayout.Width( 200 ),GUILayout.Height( 100 )))
        {
            Attach();
        }
    }

    void Extract()
    {
        TimelineAsset timelineAsset = AssetDatabase.LoadAssetAtPath<TimelineAsset>(timelineFilePath);
   
        var textDataList = new List<(string clipName, string text)>();
        if( timelineAsset != null )
        {
            IEnumerable<TrackAsset> tracks= timelineAsset.GetOutputTracks();
            
            var textSwitcherTrack = tracks.FirstOrDefault(track => track.name == TEXT_SWITCHER_TRACK_NAME);

            if (textSwitcherTrack != null)
            {
                foreach (TimelineClip timelineClip in textSwitcherTrack.GetClips())
                {
                    TextSwitcherClip tsc = (TextSwitcherClip)timelineClip.asset;
                    textDataList.Add((timelineClip.displayName, tsc.template.text));
                    Debug.Log(timelineClip.displayName + ", " + tsc.template.text);
                }
            }
        }

        var records = new List<AIVOICEData>();
        
        foreach (var textData in textDataList)
        {
            var record = new AIVOICEData()
            {
                ファイルパス = voiceExportPath + textData.clipName + ".aivoice",
                セリフ = textData.text,
                話者名 = "ユニティちゃん",
                音量 = 1.0f, 話速 = 1.0f, 高さ = 1.0f, 抑揚 = 0.5f, スタイル1 = 0.5f, スタイル2 = 0f, スタイル3 = 0f
            };
            
            records.Add(record);
        }

        using (var writer = new StreamWriter(csvExportFilePath))
        {
            using (var csv = new CsvWriter(writer, CultureInfo.InvariantCulture))
            {
                csv.WriteRecords(records);
            }
        }
        
        AssetDatabase.SaveAssets();
        AssetDatabase.Refresh();
    }

    private void Attach()
    {
        TimelineAsset timelineAsset = AssetDatabase.LoadAssetAtPath<TimelineAsset>(timelineFilePath);

        //テキストの名前とタイミング情報を取得
        var clipNameAndStartTime = new List<(string clipName, double startTime)>();
        if( timelineAsset != null )
        {
            IEnumerable<TrackAsset> tracks= timelineAsset.GetOutputTracks();
            
            var textSwitcherTrack = tracks.FirstOrDefault(track => track.name == TEXT_SWITCHER_TRACK_NAME);

            if (textSwitcherTrack != null)
            {
                foreach (TimelineClip timelineClip in textSwitcherTrack.GetClips())
                {
                    clipNameAndStartTime.Add((timelineClip.displayName, timelineClip.start));
                }
            }
            
            var audioTrack = tracks.FirstOrDefault(track => track.name == AUDIO_TRACK_NAME);
            
            DeleteAllClips(timelineAsset,audioTrack);
            
            foreach (var clipNameAndTime in clipNameAndStartTime)
            {
                var clip = audioTrack.CreateDefaultClip();
                var audioPlayableAsset = clip.asset as AudioPlayableAsset;
                
                var audioClip = AssetDatabase.LoadAssetAtPath<AudioClip>(voiceExportPath+ clipNameAndTime.clipName + ".wav");
                audioPlayableAsset.clip = audioClip;
                
                clip.start = clipNameAndTime.startTime;
                clip.duration = audioClip.length;
                clip.asset = audioPlayableAsset;
            }
        }

        TimelineEditor.Refresh(RefreshReason.ContentsAddedOrRemoved);
        
        AssetDatabase.SaveAssets();
        AssetDatabase.Refresh();
    }
    
    private void DeleteAllClips(TimelineAsset asset, TrackAsset track)
    {
        var clips = track.GetClips().ToArray();
        foreach (var c in clips)
        {
            asset.DeleteClip(c);
        }
    }
}

public class AIVOICEData
{
    public string ファイルパス { get; set; }
    public string セリフ { get; set; }
    public string 話者名 { get; set; }
    public float 音量 { get; set; }
    public float 話速 { get; set; }
    public float 高さ { get; set; }
    public float 抑揚 { get; set; }
    public float スタイル1 { get; set; }
    public float スタイル2 { get; set; }
    public float スタイル3 { get; set; }
}