Check Point

2026-09-12 09:07:09 -04:00
parent e9040d110a
commit a4073b9020
63 changed files with 2459 additions and 632 deletions
@@ -0,0 +1,34 @@
+// Model configuration fields are checked against the installed QVAC 0.19 SDK.
+export const TTS_PRESETS = Object.freeze({
+  'supertonic-en': { model: 'TTS_EN_SUPERTONIC_Q8_0', engine: 'supertonic', sampleRate: 44100 },
+  supertonic3: { model: 'TTS_MULTILINGUAL_SUPERTONIC3_Q8_0', engine: 'supertonic', sampleRate: 44100 },
+  chatterbox: { model: 'TTS_T3_TURBO_EN_CHATTERBOX_Q8_0', engine: 'chatterbox', sampleRate: 24000 },
+  parler: { model: 'TTS_MINI_V1_EN_PARLER_TTS_Q8_0', engine: 'parler', sampleRate: 44100 },
+});
+
+export function ttsConfiguration(settings) {
+  const preset = TTS_PRESETS[settings.ttsPreset];
+  if (!preset) throw new Error('Unknown speech model preset');
+  const config = { ttsEngine: preset.engine, useGPU: settings.ttsUseGpu };
+  if (preset.engine === 'supertonic') {
+    Object.assign(config, { language: settings.ttsPreset === 'supertonic3' ? settings.ttsLanguage : 'en', voice: settings.voiceId, ttsSpeed: settings.ttsSpeed, ttsNumInferenceSteps: settings.ttsSteps });
+  } else {
+    Object.assign(config, { threads: settings.ttsThreads, seed: settings.ttsSeed });
+    if (preset.engine === 'chatterbox') {
+      Object.assign(config, { language: 'en', s3genModelSrc: 'TTS_S3GEN_EN_CHATTERBOX', cfmSteps: settings.ttsCfmSteps });
+      if (settings.ttsReferenceAudio) config.referenceAudioSrc = settings.ttsReferenceAudio;
+    } else {
+      if (!settings.ttsDescription) throw new Error('Describe the Parler voice before applying settings');
+      Object.assign(config, { description: settings.ttsDescription, temperature: settings.ttsTemperature });
+    }
+  }
+  return { model: settings.ttsModel || preset.model, sampleRate: preset.sampleRate, config };
+}
+
+export function scaleSpeech(samples, volume = 100) {
+  if (volume === 100) return samples;
+  const output = new Int16Array(samples.length);
+  const gain = Math.max(0, Math.min(100, volume)) / 100;
+  for (let i = 0; i < samples.length; i++) output[i] = Math.round(samples[i] * gain);
+  return output;
+}