ben.carrasco/speech-utils
1
0
Fork 0
mirror of https://github.com/jambonz/speech-utils.git synced 2026-01-25 02:08:26 +00:00
Code Issues Packages Projects Releases Wiki Activity

wip

Browse Source
This commit is contained in:
Hoan HL
2026-01-12 17:18:11 +07:00
parent 460ca70ea7
commit ded60cb7aa
2 changed files with 7 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
lib/synth-audio.js
View File

@@ -417,7 +417,7 @@ const synthGoogle = async(logger, {
}) => { }) => {
const client = new ttsGoogle.TextToSpeechClient(credentials); const client = new ttsGoogle.TextToSpeechClient(credentials);
const isGemini = credentials.use_gemini_tts; const isGemini = !!model;
const isVoiceCloning = typeof voice === 'object' && voice.voice_cloning_key; const isVoiceCloning = typeof voice === 'object' && voice.voice_cloning_key;
// Build input based on voice type // Build input based on voice type

13
test/synth.js
View File

@@ -179,7 +179,7 @@ test('Google Gemini TTS synth tests', async(t) => {
const creds = JSON.parse(str); const creds = JSON.parse(str);
const geminiModel = process.env.GCP_GEMINI_TTS_MODEL || 'gemini-2.5-flash-tts'; const geminiModel = process.env.GCP_GEMINI_TTS_MODEL || 'gemini-2.5-flash-tts';
// Test basic Gemini TTS synthesis // Test Gemini TTS with model and instructions (both required for Gemini)
let result = await synthAudio(stats, { let result = await synthAudio(stats, {
vendor: 'google', vendor: 'google',
credentials: { credentials: {
@@ -187,17 +187,17 @@ test('Google Gemini TTS synth tests', async(t) => {
client_email: creds.client_email, client_email: creds.client_email,
private_key: creds.private_key, private_key: creds.private_key,
}, },
use_gemini_tts: true
}, },
language: 'en-US', language: 'en-US',
voice: 'Kore', voice: 'Kore',
model: geminiModel, model: geminiModel,
text: 'Hello, this is a test of Google Gemini text to speech.', text: 'Hello, this is a test of Google Gemini text to speech.',
instructions: 'Speak clearly and naturally.',
}); });
t.ok(!result.servedFromCache, `successfully synthesized Google Gemini TTS audio to ${result.filePath}`); t.ok(!result.servedFromCache, `successfully synthesized Google Gemini TTS audio to ${result.filePath}`);
t.ok(result.filePath.endsWith('.r24'), 'Gemini TTS audio file has correct extension'); t.ok(result.filePath.endsWith('.wav'), 'Gemini TTS audio file has correct extension');
// Test Gemini TTS with instructions (prompt) // Test Gemini TTS with different voice and instructions
result = await synthAudio(stats, { result = await synthAudio(stats, {
vendor: 'google', vendor: 'google',
credentials: { credentials: {
@@ -205,7 +205,6 @@ test('Google Gemini TTS synth tests', async(t) => {
client_email: creds.client_email, client_email: creds.client_email,
private_key: creds.private_key, private_key: creds.private_key,
}, },
use_gemini_tts: true
}, },
language: 'en-US', language: 'en-US',
voice: 'Charon', voice: 'Charon',
@@ -223,12 +222,12 @@ test('Google Gemini TTS synth tests', async(t) => {
client_email: creds.client_email, client_email: creds.client_email,
private_key: creds.private_key, private_key: creds.private_key,
}, },
use_gemini_tts: true
}, },
language: 'en-US', language: 'en-US',
voice: 'Kore', voice: 'Kore',
model: geminiModel, model: geminiModel,
text: 'Hello, this is a test of Google Gemini text to speech.', text: 'Hello, this is a test of Google Gemini text to speech.',
instructions: 'Speak clearly and naturally.',
}); });
t.ok(result.servedFromCache, `successfully retrieved Gemini TTS audio from cache ${result.filePath}`); t.ok(result.servedFromCache, `successfully retrieved Gemini TTS audio from cache ${result.filePath}`);
@@ -240,12 +239,12 @@ test('Google Gemini TTS synth tests', async(t) => {
client_email: creds.client_email, client_email: creds.client_email,
private_key: creds.private_key, private_key: creds.private_key,
}, },
use_gemini_tts: true
}, },
language: 'en-US', language: 'en-US',
voice: 'Leda', voice: 'Leda',
model: geminiModel, model: geminiModel,
text: '<speak>This SSML should be stripped for Gemini TTS.</speak>', text: '<speak>This SSML should be stripped for Gemini TTS.</speak>',
instructions: 'Speak naturally.',
disableTtsCache: true disableTtsCache: true
}); });
t.ok(!result.servedFromCache, `successfully synthesized Gemini TTS with SSML stripped to ${result.filePath}`); t.ok(!result.servedFromCache, `successfully synthesized Gemini TTS with SSML stripped to ${result.filePath}`);