feat(Voice): voice support for Windows and macOS

Refactored the entire voice subsystem, now with a generalized API.
Added ffplay-static to handle playback on Windows and macOS.
Added an option to specify additional ffmpeg args (filters etc.).
Linux requires installed ffmpeg, static does not include pulse support.
macOS compatibility not tested.

Partially resolves issue #10.

package.json

@@ -48,6 +48,7 @@
     "discord.js": "^12.5.1",
     "env-paths": "^2.2.0",
     "ffmpeg-static": "^4.2.7",
+    "ffplay-static": "^3.2.2",
     "i18n": "^0.13.2",
     "markdown-it": "^12.0.4",
     "node-notifier": "^8.0.1",

src/components/VoicePanel/VoicePanel.ts

@@ -9,7 +9,6 @@ import {
   QWidget,
 } from '@nodegui/nodegui';
 import { Input, Mixer } from 'audio-mixer';
-import { ChildProcessWithoutNullStreams } from 'child_process';
 import {
   Client,
   Constants,
@@ -28,7 +27,9 @@ import { ConfigManager } from '../../utilities/ConfigManager';
 import { createLogger } from '../../utilities/Console';
 import { Events as AppEvents } from '../../utilities/Events';
 import { __ } from '../../utilities/StringProvider';
-import { createPlaybackStream, createRecordStream } from '../../utilities/VoiceStreams';
+import { PlaybackStream } from '../../utilities/voice/PlaybackStream';
+import { RecordStream } from '../../utilities/voice/RecordStream';
+import { voiceProvider as vp } from '../../utilities/voice/VoiceProviderManager';
 import { DIconButton } from '../DIconButton/DIconButton';
 import { NoiseReductor } from './NoiseReductor';
 import { Silence } from './Silence';
@@ -58,13 +59,13 @@ export class VoicePanel extends QWidget {
 
   private currentPlaybackDevice?: string;
 
-  private playbackStream?: ChildProcessWithoutNullStreams;
+  private playbackStream?: PlaybackStream;
 
   private playbackVolumeTransformer?: VolumeTransformer;
 
   private currentRecordDevice?: string;
 
-  private recordStream?: ChildProcessWithoutNullStreams;
+  private recordStream?: RecordStream;
 
   private recordVolumeTransformer?: VolumeTransformer;
 
@@ -157,8 +158,8 @@ export class VoicePanel extends QWidget {
   private handleDisconnectButton() {
     this.statusLabel.setText("<font color='#f04747'>Disconnecting</font>");
     this.connection?.disconnect();
-    this.recordStream?.kill();
-    this.playbackStream?.kill();
+    this.recordStream?.end();
+    this.playbackStream?.end();
     this.mixer?.close();
     this.streams.clear();
     this.hide();
@@ -230,22 +231,22 @@ export class VoicePanel extends QWidget {
   }
 
   private initPlayback() {
-    if (!this.connection || !this.channel) {
+    if (!this.connection || !this.channel || !vp) {
       return;
     }
 
     const { channel } = this;
 
     this.mixer?.close();
-    this.playbackStream?.kill();
+    this.playbackStream?.end();
     this.streams.clear();
 
     pipeline(
       (this.mixer = new Mixer(MIXER_OPTIONS)), // Initialize the member streams mixer
       ((this.playbackVolumeTransformer = new VolumeTransformer({
         type: 's16le',
       })) as unknown) as Transform, // Change the volume
-      (this.playbackStream = createPlaybackStream()).stdin, // Output to a playback stream
+      (this.playbackStream = vp.createPlaybackStream()).stream, // Output to a playback stream
       (err) => err && debug("Couldn't finish playback pipeline.", err)
     );
 
@@ -255,14 +256,14 @@ export class VoicePanel extends QWidget {
   }
 
   private initRecord() {
-    if (!this.connection) {
+    if (!this.connection || !vp) {
       return;
     }
 
     this.connection.dispatcher.end();
 
     const recorder = pipeline(
-      (this.recordStream = createRecordStream()).stdout, // Open a recording stream
+      (this.recordStream = vp.createRecordStream()).stream, // Open a recording stream
       ((this.recordVolumeTransformer = new VolumeTransformer({
         type: 's16le',
       })) as unknown) as Transform,
@@ -277,7 +278,7 @@ export class VoicePanel extends QWidget {
   private async joinChannel(channel: VoiceChannel) {
     const { infoLabel, statusLabel, createConnection, initPlayback, initRecord } = this;
 
-    if (process.platform !== 'linux') {
+    if (!vp) {
       VoicePanel.openVoiceNotSupportedDialog(channel);
 
       return;

src/utilities/IConfig.ts

@@ -20,6 +20,8 @@ export type VoiceSettings = {
   inputVolume?: number;
   outputVolume?: number;
   inputSensitivity?: number;
+  playbackOptions?: string;
+  recordOptions?: string;
 };
 
 export type OverlaySettings = {

src/utilities/voice/DarwinVoiceProvider.ts

@@ -0,0 +1,35 @@
+/* eslint-disable global-require */
+import { ID, S16LE_OPTIONS } from './VoiceOptions';
+import { VoiceProvider } from './VoiceProvider';
+
+/**
+ * VoiceProvider for systems on macOS.
+ * THIS WAS NOT TESTED. I do not have a macOS system to test this
+ * implementation's behaviour, most likely it's similar to Windows.
+ * TODO: Input/output devices detection.
+ */
+export class DarwinVoiceProvider extends VoiceProvider {
+  public PLAYBACK_OPTIONS = [...S16LE_OPTIONS, ID.PLAYBACK_OPTIONS, '-'];
+
+  public RECORD_OPTIONS = [
+    '-f',
+    'avfoundation',
+    '-i',
+    `"none:${ID.RECORD_DEVICE}"`,
+    ID.RECORD_OPTIONS,
+    ...S16LE_OPTIONS,
+    '-',
+  ];
+
+  public readonly PLAYBACK_DEFAULT_DEVICE = '0';
+
+  public readonly RECORD_DEFAULT_DEVICE = '0';
+
+  createPlaybackStream = this.createFFplayPlaybackStream;
+
+  createRecordStream = this.createFFmpegRecordStream;
+
+  getInputDevices = () => ['0'];
+
+  getOutputDevices = () => ['0'];
+}

src/utilities/voice/FFProvider.ts

@@ -0,0 +1,14 @@
+import { platform } from 'os';
+import { join } from 'path';
+
+export function getFFmpeg() {
+  if (platform() === 'linux') {
+    return 'ffmpeg';
+  }
+
+  return join(__dirname, `ffmpeg${platform() === 'win32' ? '.exe' : ''}`);
+}
+
+export function getFFplay() {
+  return join(__dirname, `ffplay${platform() === 'win32' ? '.exe' : ''}`);
+}

src/utilities/voice/LinuxVoiceProvider.ts

@@ -0,0 +1,77 @@
+import { execSync } from 'child_process';
+import { createLogger } from '../Console';
+import { ID, S16LE_OPTIONS } from './VoiceOptions';
+import { VoiceProvider } from './VoiceProvider';
+
+const { error } = createLogger('LinuxVoiceProvider');
+
+/**
+ * VoiceProvider for systems on Linux with PulseAudio support.
+ */
+export class LinuxVoiceProvider extends VoiceProvider {
+  public PLAYBACK_OPTIONS = [
+    ...S16LE_OPTIONS,
+    ID.PLAYBACK_OPTIONS,
+    '-i',
+    '-',
+    '-f',
+    'pulse',
+    '-buffer_duration',
+    '5',
+    '-name',
+    ID.APP_NAME,
+    ID.PLAYBACK_DEVICE,
+  ];
+
+  public RECORD_OPTIONS = [
+    '-f',
+    'pulse',
+    '-name',
+    ID.APP_NAME,
+    '-i',
+    ID.RECORD_DEVICE,
+    ID.RECORD_OPTIONS,
+    ...S16LE_OPTIONS,
+    '-',
+  ];
+
+  public readonly PLAYBACK_DEFAULT_DEVICE = 'default';
+
+  public readonly RECORD_DEFAULT_DEVICE = 'default';
+
+  createPlaybackStream = this.createFFmpegPlaybackStream;
+
+  createRecordStream = this.createFFmpegRecordStream;
+
+  getInputDevices = () => {
+    try {
+      const sourcesOut = execSync('pactl list sources short').toString();
+      const sources = sourcesOut
+        .split('\n')
+        .map((value) => value.split('\t')[1])
+        .filter((value) => value);
+
+      return sources;
+    } catch (e) {
+      error(e);
+
+      return [this.RECORD_DEFAULT_DEVICE];
+    }
+  };
+
+  getOutputDevices = () => {
+    try {
+      const sinksOut = execSync('pactl list sinks short').toString();
+      const sinks = sinksOut
+        .split('\n')
+        .map((value) => value.split('\t')[1])
+        .filter((value) => value);
+
+      return sinks;
+    } catch (e) {
+      error(e);
+
+      return [this.PLAYBACK_DEFAULT_DEVICE];
+    }
+  };
+}

src/utilities/voice/PlaybackStream.ts

@@ -0,0 +1,7 @@
+import { Writable } from 'stream';
+
+export type PlaybackStream = {
+  stream: Writable;
+  device: string;
+  end(): void;
+};

src/utilities/voice/RecordStream.ts

@@ -0,0 +1,7 @@
+import { Readable } from 'stream';
+
+export type RecordStream = {
+  stream: Readable;
+  device: string;
+  end(): void;
+};