hackmaster/Robovoice
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

v0.5: TCP STT client with PTT on/off, server endpoint in config + UI

Browse Source
This commit is contained in:
Mute
2026-08-11 09:03:50 +00:00
parent 08d6eeb46e
commit 4a58b94f35
11 changed files with 696 additions and 53 deletions
Download Patch File Download Diff File Expand all files Collapse all files
Robovoice.App/AppConfig.cs
+1
View File
@@ -12,6 +12,7 @@ public sealed class AppConfig
public int LengthScale { get; set; } = 100;
public int NoiseWScale { get; set; } = 800;
public bool MinimizeToTray { get; set; } = true;
public string ServerEndpoint { get; set; } = "127.0.0.1:5210";
public static string AppDataDir => Path.Combine(
Environment.GetFolderPath(Environment.SpecialFolder.LocalApplicationData),
Robovoice.App/MainForm.Designer.cs
Generated
+27 -2
View File
@@ -16,6 +16,9 @@ partial class MainForm
private Label lblFile = null!;
private Button btnBrowseFile = null!;
private Label lblFileName = null!;
private Label lblServer = null!;
private TextBox txtServer = null!;
private Button btnConnect = null!;
private Label lblLineStatus = null!;
private RichTextBox txtLog = null!;
private CheckBox chkMinimizeToTray = null!;
@@ -54,6 +57,9 @@ partial class MainForm
lblFile = new Label();
btnBrowseFile = new Button();
lblFileName = new Label();
lblServer = new Label();
txtServer = new TextBox();
btnConnect = new Button();
lblLineStatus = new Label();
txtLog = new RichTextBox();
chkMinimizeToTray = new CheckBox();
@@ -134,13 +140,29 @@ partial class MainForm
// lblFileName
lblFileName.Text = "(none)";
lblFileName.Location = new Point(155, 48);
lblFileName.Size = new Size(490, 23);
lblFileName.Size = new Size(280, 23);
lblFileName.TextAlign = ContentAlignment.MiddleLeft;
lblFileName.ForeColor = Color.Gray;
// lblServer
lblServer.Text = "Server:";
lblServer.Location = new Point(440, 48);
lblServer.Size = new Size(45, 23);
lblServer.TextAlign = ContentAlignment.MiddleLeft;
// txtServer
txtServer.Location = new Point(488, 45);
txtServer.Size = new Size(110, 23);
// btnConnect
btnConnect.Text = "Connect";
btnConnect.Location = new Point(603, 44);
btnConnect.Size = new Size(60, 25);
btnConnect.UseVisualStyleBackColor = true;
// lblLineStatus
lblLineStatus.Text = "";
lblLineStatus.Location = new Point(650, 48);
lblLineStatus.Location = new Point(645, 48);
lblLineStatus.Size = new Size(160, 23);
lblLineStatus.TextAlign = ContentAlignment.MiddleRight;
lblLineStatus.ForeColor = Color.DarkBlue;
@@ -254,6 +276,9 @@ partial class MainForm
Controls.Add(lblFile);
Controls.Add(btnBrowseFile);
Controls.Add(lblFileName);
Controls.Add(lblServer);
Controls.Add(txtServer);
Controls.Add(btnConnect);
Controls.Add(lblLineStatus);
Controls.Add(lblNoise);
Controls.Add(trkNoise);
Robovoice.App/MainForm.cs
+62 -23
View File
@@ -2,7 +2,7 @@ using NAudio.Wave;
using Robovoice.App;
using Robovoice.Core;
using Robovoice.Core.Voices;
using Robovoice.Stt.File;
using Robovoice.Stt.Tcp;
using Robovoice.Tts.LibPiper;
using System.Diagnostics;
@@ -16,7 +16,7 @@ internal sealed partial class MainForm : Form
private LibPiperTtsEngine? _tts;
private AudioOutput? _audioOutput;
private FileSttSource? _sttSource;
private TcpSttSource? _sttSource;
private Orchestrator? _orchestrator;
private PttHotkey? _pttHotkey;
private NotifyIcon? _trayIcon;
@@ -61,6 +61,7 @@ internal sealed partial class MainForm : Form
OnSliderScroll(null, EventArgs.Empty);
chkMinimizeToTray.Checked = _config.MinimizeToTray;
txtServer.Text = _config.ServerEndpoint;
btnBrowseFile.Click += OnBrowseFile;
btnTestVoice.Click += OnTestVoice;
@@ -70,6 +71,8 @@ internal sealed partial class MainForm : Form
txtPttKey.KeyDown += OnPttKeyDown;
cmbOutput.SelectedIndexChanged += OnOutputChanged;
cmbVoice.SelectedIndexChanged += OnVoiceChanged;
txtServer.Leave += OnServerChanged;
btnConnect.Click += OnConnect;
trkNoise.Scroll += OnSliderScroll;
trkSpeed.Scroll += OnSliderScroll;
@@ -239,7 +242,13 @@ internal sealed partial class MainForm : Form
lengthScale: trkSpeed.Value / 100.0f,
noiseWScale: trkNoiseW.Value / 1000.0f);
_audioOutput = new AudioOutput();
_sttSource = new FileSttSource();
if (_sttSource is null)
{
_sttSource = new TcpSttSource { ServerEndpoint = txtServer.Text, Log = Log };
Log($"STT endpoint: {_sttSource.ServerEndpoint} (press Connect)");
}
_orchestrator?.DisposeAsync().AsTask().Wait();
_orchestrator = new Orchestrator(_tts, _audioOutput, _sttSource, Log)
{
@@ -249,7 +258,7 @@ internal sealed partial class MainForm : Form
try
{
await _orchestrator.InitializeTtsAsync();
Log("Engine ready. Select a text file and press PTT key.");
Log("Engine ready. Press PTT to send to STT server.");
SetupHotkey();
}
catch (Exception ex)
@@ -271,22 +280,27 @@ internal sealed partial class MainForm : Form
private void OnPttPressed(object? sender, EventArgs e)
{
Log("PTT pressed");
try
{
_sttSource?.SendOn();
}
catch (Exception ex)
{
Log($"SendOn failed: {ex.Message}");
}
}
private void OnPttReleased(object? sender, EventArgs e)
{
Log("PTT released");
if (_orchestrator is null || _sttSource is null) return;
if (_sttSource.LineCount == 0)
try
{
Log("No text file loaded.");
return;
_sttSource?.SendOff();
}
catch (Exception ex)
{
Log($"SendOff failed: {ex.Message}");
}
UpdateLineStatus();
_orchestrator.TriggerNextLine();
UpdateLineStatus();
}
private void OnBrowseFile(object? sender, EventArgs e)
@@ -294,16 +308,16 @@ internal sealed partial class MainForm : Form
using var dlg = new OpenFileDialog
{
Filter = "Text files (*.txt)|*.txt|All files (*.*)|*.*",
Title = "Select a text file to read sequentially",
Title = "Select a text file to speak",
};
if (dlg.ShowDialog() == DialogResult.OK)
{
_sttSource?.LoadFile(dlg.FileName);
string text = File.ReadAllText(dlg.FileName);
lblFileName.Text = Path.GetFileName(dlg.FileName);
lblFileName.ForeColor = Color.Black;
UpdateLineStatus();
Log($"Loaded: {dlg.FileName} ({_sttSource?.LineCount} lines)");
Log($"Loaded: {dlg.FileName} ({text.Length} chars)");
_ = _orchestrator?.SynthesizeAsync(text);
}
}
@@ -411,15 +425,38 @@ internal sealed partial class MainForm : Form
SaveConfig();
}
private void UpdateLineStatus()
private void OnServerChanged(object? sender, EventArgs e)
{
if (_sttSource is null || _sttSource.LineCount == 0)
if (_sttSource is not null)
{
lblLineStatus.Text = "";
_sttSource.ServerEndpoint = txtServer.Text;
Log($"Server endpoint: {txtServer.Text}");
}
SaveConfig();
}
private async void OnConnect(object? sender, EventArgs e)
{
if (_sttSource is null)
{
Log("STT source not initialized.");
return;
}
int displayIdx = (_sttSource.CurrentIndex % _sttSource.LineCount) + 1;
lblLineStatus.Text = $"Line {displayIdx}/{_sttSource.LineCount}";
_sttSource.ServerEndpoint = txtServer.Text;
SaveConfig();
btnConnect.Enabled = false;
try
{
if (_sttSource.IsRunning)
await _sttSource.ReconnectAsync();
else
await _sttSource.StartAsync();
}
finally
{
btnConnect.Enabled = true;
}
}
private void SetupTray()
@@ -487,6 +524,7 @@ internal sealed partial class MainForm : Form
_config.LengthScale = trkSpeed.Value;
_config.NoiseWScale = trkNoiseW.Value;
_config.MinimizeToTray = chkMinimizeToTray.Checked;
_config.ServerEndpoint = txtServer.Text;
_config.Save();
}
@@ -508,7 +546,8 @@ internal sealed partial class MainForm : Form
{
_pttHotkey?.Dispose();
_trayIcon!.Visible = false;
_orchestrator?.DisposeAsync().AsTask().Wait();
_orchestrator?.DisposeAsync().AsTask().Wait(2000);
_sttSource?.DisposeAsync().AsTask().Wait(2000);
SaveConfig();
}
}
Robovoice.App/Orchestrator.cs
+2 -12
View File
@@ -1,5 +1,4 @@
using Robovoice.Core;
using Robovoice.Stt.File;
using Robovoice.Tts.LibPiper;
namespace Robovoice.App;
@@ -8,7 +7,7 @@ internal sealed class Orchestrator : IAsyncDisposable
{
private readonly LibPiperTtsEngine _tts;
private readonly AudioOutput _audioOutput;
private readonly FileSttSource _sttSource;
private readonly ISttSource _sttSource;
private readonly Action<string> _log;
private CancellationTokenSource? _currentCts;
private bool _disposed;
@@ -18,7 +17,7 @@ internal sealed class Orchestrator : IAsyncDisposable
public Orchestrator(
LibPiperTtsEngine tts,
AudioOutput audioOutput,
FileSttSource sttSource,
ISttSource sttSource,
Action<string> log)
{
_tts = tts;
@@ -97,15 +96,6 @@ internal sealed class Orchestrator : IAsyncDisposable
sw.Stop();
}
public void TriggerNextLine()
{
_sttSource.EmitNext();
}
public string? GetPendingLine() => _sttSource.GetPendingLine();
public int GetCurrentLineIndex() => _sttSource.CurrentIndex;
public int GetLineCount() => _sttSource.LineCount;
public async ValueTask DisposeAsync()
{
if (_disposed) return;
Robovoice.App/Robovoice.App.csproj
+1 -2
View File
@@ -3,8 +3,7 @@
<ItemGroup>
<ProjectReference Include="..\Robovoice.Core\Robovoice.Core.csproj" />
<ProjectReference Include="..\Robovoice.Tts.LibPiper\Robovoice.Tts.LibPiper.csproj" />
<ProjectReference Include="..\Robovoice.Stt.File\Robovoice.Stt.File.csproj" />
<ProjectReference Include="..\Robovoice.Stt.Udp\Robovoice.Stt.Udp.csproj" />
<ProjectReference Include="..\Robovoice.Stt.Tcp\Robovoice.Stt.Tcp.csproj" />
</ItemGroup>
<ItemGroup>