mirror of
https://github.com/jambonz/speech-utils.git
synced 2026-01-25 02:08:26 +00:00
Merge pull request #135 from jambonz/feat/gemini_tts
support gemini tts
This commit is contained in:
Dave Horton
@@ -204,7 +204,10 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
|
||||
const startAt = process.hrtime();
|
||||
switch (vendor) {
|
||||
case 'google':
|
||||
audioData = await synthGoogle(logger, {credentials, stats, language, voice, gender, key, text});
|
||||
audioData = await synthGoogle(logger, {
|
||||
credentials, stats, language, voice, gender, key, text, model, options, instructions,
|
||||
renderForCaching, disableTtsStreaming, disableTtsCache
|
||||
});
|
||||
break;
|
||||
case 'aws':
|
||||
case 'polly':
|
||||
@@ -409,72 +412,124 @@ const synthPolly = async(createHash, retrieveHash, logger,
|
||||
}
|
||||
};
|
||||
|
||||
const synthGoogle = async(logger, {credentials, stats, language, voice, gender, text}) => {
|
||||
const client = new ttsGoogle.TextToSpeechClient(credentials);
|
||||
// If google custom voice cloning is used.
|
||||
// At this time 31 Oct 2024, google node sdk has not support voice cloning yet.
|
||||
if (typeof voice === 'object' && voice.voice_cloning_key) {
|
||||
try {
|
||||
const accessToken = await client.auth.getAccessToken();
|
||||
const projectId = await client.getProjectId();
|
||||
|
||||
const post = bent('https://texttospeech.googleapis.com', 'POST', 'json', {
|
||||
'Authorization': `Bearer ${accessToken}`,
|
||||
'x-goog-user-project': projectId,
|
||||
'Content-Type': 'application/json; charset=utf-8'
|
||||
});
|
||||
const synthGoogle = async(logger, {
|
||||
credentials, stats, language, voice, gender, key, text, model, options, instructions,
|
||||
renderForCaching, disableTtsStreaming, disableTtsCache
|
||||
}) => {
|
||||
const isGemini = !!model;
|
||||
const isVoiceCloning = typeof voice === 'object' && voice.voice_cloning_key;
|
||||
// HD voices have pattern like en-US-Chirp3-HD-Charon
|
||||
const isHDVoice = typeof voice === 'string' && voice.includes('-HD-');
|
||||
|
||||
const payload = {
|
||||
input: {
|
||||
text
|
||||
},
|
||||
voice: {
|
||||
language_code: language,
|
||||
voice_clone: {
|
||||
voice_cloning_key: voice.voice_cloning_key
|
||||
}
|
||||
},
|
||||
audioConfig: {
|
||||
// Cloning voice at this time still in v1 beta version, and it support LINEAR16 in Wav format, 24.000Hz
|
||||
audioEncoding: 'LINEAR16',
|
||||
sample_rate_hertz: 24000
|
||||
}
|
||||
};
|
||||
|
||||
const wav = await post('/v1beta1/text:synthesize', payload);
|
||||
return {
|
||||
audioContent: Buffer.from(wav.audioContent, 'base64'),
|
||||
extension: 'wav',
|
||||
sampleRate: 24000
|
||||
};
|
||||
} catch (err) {
|
||||
logger.info({err: await err.text()}, 'synthGoogle returned error');
|
||||
throw err;
|
||||
// Streaming support for Google TTS (Gemini, HD voices, and standard voices)
|
||||
// Voice cloning does not support streaming
|
||||
if (!isVoiceCloning && !JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
|
||||
// Strip SSML tags for Gemini TTS (it doesn't support SSML)
|
||||
let inputText = text;
|
||||
if (isGemini && text.startsWith('<speak>')) {
|
||||
inputText = text.replace(/<[^>]*>/g, '').trim();
|
||||
logger.info('synthGoogle: Gemini TTS does not support SSML, stripped tags from input');
|
||||
}
|
||||
|
||||
let params = '{';
|
||||
params += `credentials=${Buffer.from(JSON.stringify(credentials.credentials)).toString('base64')}`;
|
||||
params += `,playback_id=${key}`;
|
||||
params += ',vendor=google';
|
||||
params += `,voice=${voice}`;
|
||||
params += `,language_code=${language || 'en-US'}`;
|
||||
params += `,write_cache_file=${disableTtsCache ? 0 : 1}`;
|
||||
const useLiveApi = options?.useLiveApi ?? isHDVoice;
|
||||
const useGeminiTts = options?.useGeminiTts ?? isGemini;
|
||||
params += `,use_live_api=${useLiveApi ? 1 : 0}`;
|
||||
params += `,use_gemini_tts=${useGeminiTts ? 1 : 0}`;
|
||||
if (model) params += `,model_name=${model}`;
|
||||
if (gender) params += `,gender=${gender}`;
|
||||
// comma is used to separate parameters in freeswitch tts module
|
||||
const prompt = options?.prompt || instructions;
|
||||
if (prompt) params += `,prompt=${prompt.replace(/\n/g, ' ').replace(/,/g, ';')}`;
|
||||
params += '}';
|
||||
|
||||
return {
|
||||
filePath: `say:${params}${(isGemini ? inputText : text).replace(/\n/g, ' ')}`,
|
||||
servedFromCache: false,
|
||||
rtt: 0
|
||||
};
|
||||
}
|
||||
|
||||
const opts = {
|
||||
voice: {
|
||||
...(typeof voice === 'string' && {name: voice}),
|
||||
...(typeof voice === 'object' && {customVoice: voice}),
|
||||
const client = new ttsGoogle.TextToSpeechClient(credentials);
|
||||
|
||||
// Build input based on voice type
|
||||
let input;
|
||||
if (isGemini) {
|
||||
// Gemini TTS does not support SSML - strip tags if present
|
||||
let inputText = text;
|
||||
if (text.startsWith('<speak>')) {
|
||||
inputText = text.replace(/<[^>]*>/g, '').trim();
|
||||
logger.info('synthGoogle: Gemini TTS does not support SSML, stripped tags from input');
|
||||
}
|
||||
// Use instructions as prompt for Gemini TTS style control, options.prompt can override
|
||||
const prompt = options?.prompt || instructions;
|
||||
input = {
|
||||
text: inputText,
|
||||
...(prompt && { prompt })
|
||||
};
|
||||
} else {
|
||||
input = text.startsWith('<speak>') ? { ssml: text } : { text };
|
||||
}
|
||||
|
||||
// Build voice selection params based on voice type
|
||||
let voiceParams;
|
||||
if (isGemini) {
|
||||
voiceParams = {
|
||||
languageCode: language || 'en-US',
|
||||
name: voice,
|
||||
modelName: model
|
||||
};
|
||||
} else if (isVoiceCloning) {
|
||||
voiceParams = {
|
||||
languageCode: language,
|
||||
voiceClone: {
|
||||
voiceCloningKey: voice.voice_cloning_key
|
||||
}
|
||||
};
|
||||
} else {
|
||||
voiceParams = {
|
||||
...(typeof voice === 'string' && { name: voice }),
|
||||
...(typeof voice === 'object' && { customVoice: voice }),
|
||||
languageCode: language,
|
||||
ssmlGender: gender || 'SSML_VOICE_GENDER_UNSPECIFIED'
|
||||
},
|
||||
audioConfig: {audioEncoding: 'MP3'}
|
||||
};
|
||||
Object.assign(opts, {input: text.startsWith('<speak>') ? {ssml: text} : {text}});
|
||||
};
|
||||
}
|
||||
|
||||
// Build audio config based on voice type
|
||||
let audioConfig;
|
||||
let extension;
|
||||
let sampleRate;
|
||||
if (isVoiceCloning) {
|
||||
audioConfig = { audioEncoding: 'LINEAR16', sampleRateHertz: 24000 };
|
||||
extension = 'wav';
|
||||
sampleRate = 24000;
|
||||
} else {
|
||||
audioConfig = { audioEncoding: 'MP3' };
|
||||
extension = 'mp3';
|
||||
sampleRate = 8000;
|
||||
}
|
||||
|
||||
const opts = { input, voice: voiceParams, audioConfig };
|
||||
|
||||
try {
|
||||
const responses = await client.synthesizeSpeech(opts);
|
||||
logger.debug({ opts }, 'synthGoogle: request');
|
||||
const [response] = await client.synthesizeSpeech(opts);
|
||||
stats.increment('tts.count', ['vendor:google', 'accepted:yes']);
|
||||
client.close();
|
||||
return {
|
||||
audioContent: responses[0].audioContent,
|
||||
extension: 'mp3',
|
||||
sampleRate: 8000
|
||||
audioContent: response.audioContent,
|
||||
extension,
|
||||
sampleRate
|
||||
};
|
||||
} catch (err) {
|
||||
console.error(err);
|
||||
logger.info({err, opts}, 'synthAudio: Error synthesizing speech using google');
|
||||
logger.info({ err, opts }, 'synthAudio: Error synthesizing speech using google');
|
||||
stats.increment('tts.count', ['vendor:google', 'accepted:no']);
|
||||
client && client.close();
|
||||
throw err;
|
||||
|
||||
1376
package-lock.json
generated
1376
package-lock.json
generated
File diff suppressed because it is too large
Load Diff
@@ -30,7 +30,7 @@
|
||||
"@aws-sdk/client-polly": "^3.496.0",
|
||||
"@aws-sdk/client-sts": "^3.496.0",
|
||||
"@cartesia/cartesia-js": "^2.2.7",
|
||||
"@google-cloud/text-to-speech": "^5.5.0",
|
||||
"@google-cloud/text-to-speech": "^6.4.0",
|
||||
"@grpc/grpc-js": "^1.9.14",
|
||||
"@jambonz/realtimedb-helpers": "^0.8.7",
|
||||
"bent": "^7.3.12",
|
||||
|
||||
385
test/synth.js
385
test/synth.js
@@ -41,6 +41,7 @@ test('Google speech synth tests', async(t) => {
|
||||
gender: 'FEMALE',
|
||||
text: 'This is a test. This is only a test',
|
||||
salt: 'foo.bar',
|
||||
renderForCaching: true,
|
||||
});
|
||||
t.ok(!opts.servedFromCache, `successfully synthesized google audio to ${opts.filePath}`);
|
||||
|
||||
@@ -55,6 +56,7 @@ test('Google speech synth tests', async(t) => {
|
||||
language: 'en-GB',
|
||||
gender: 'FEMALE',
|
||||
text: 'This is a test. This is only a test',
|
||||
renderForCaching: true,
|
||||
});
|
||||
t.ok(opts.servedFromCache, `successfully retrieved cached google audio from ${opts.filePath}`);
|
||||
|
||||
@@ -78,6 +80,7 @@ test('Google speech synth tests', async(t) => {
|
||||
language: 'en-GB',
|
||||
gender: 'FEMALE',
|
||||
text: 'This is a test. This is only a test',
|
||||
renderForCaching: true,
|
||||
});
|
||||
t.ok(!opts.servedFromCache, `successfully synthesized google audio regardless of current cache to ${opts.filePath}`);
|
||||
} catch (err) {
|
||||
@@ -114,7 +117,8 @@ GCP_CUSTOM_VOICE_FILE nor GCP_CUSTOM_VOICE_JSON_KEY provided, GCP_CUSTOM_VOICE_M
|
||||
voice: {
|
||||
reportedUsage: 'REALTIME',
|
||||
model: process.env.GCP_CUSTOM_VOICE_MODEL
|
||||
}
|
||||
},
|
||||
renderForCaching: true,
|
||||
});
|
||||
t.ok(!opts.servedFromCache, `successfully synthesized google custom voice audio to ${opts.filePath}`);
|
||||
} catch (err) {
|
||||
@@ -132,7 +136,7 @@ test('Google speech voice cloning synth tests', async(t) => {
|
||||
!process.env.GCP_CUSTOM_VOICE_JSON_KEY ||
|
||||
!process.env.GCP_VOICE_CLONING_FILE &&
|
||||
!process.env.GCP_VOICE_CLONING_JSON_KEY) {
|
||||
t.pass(`skipping google speech synth tests since neither
|
||||
t.pass(`skipping google speech synth tests since neither
|
||||
GCP_CUSTOM_VOICE_FILE nor GCP_CUSTOM_VOICE_JSON_KEY provided,
|
||||
GCP_VOICE_CLONING_FILE nor GCP_VOICE_CLONING_JSON_KEY is not provided`);
|
||||
return t.end();
|
||||
@@ -156,7 +160,8 @@ GCP_VOICE_CLONING_FILE nor GCP_VOICE_CLONING_JSON_KEY is not provided`);
|
||||
text: 'This is a test. This is only a test. This is a test. This is only a test. This is a test. This is only a test',
|
||||
voice: {
|
||||
voice_cloning_key
|
||||
}
|
||||
},
|
||||
renderForCaching: true,
|
||||
});
|
||||
t.ok(!opts.servedFromCache, `successfully synthesized google voice cloning audio to ${opts.filePath}`);
|
||||
} catch (err) {
|
||||
@@ -166,6 +171,380 @@ GCP_VOICE_CLONING_FILE nor GCP_VOICE_CLONING_JSON_KEY is not provided`);
|
||||
client.quit();
|
||||
});
|
||||
|
||||
test('Google Gemini TTS synth tests', async(t) => {
|
||||
const fn = require('..');
|
||||
const {synthAudio, client} = fn(opts, logger);
|
||||
|
||||
if (!process.env.GCP_FILE && !process.env.GCP_JSON_KEY) {
|
||||
t.pass('skipping Google Gemini TTS synth tests since neither GCP_FILE nor GCP_JSON_KEY provided');
|
||||
return t.end();
|
||||
}
|
||||
try {
|
||||
const str = process.env.GCP_JSON_KEY || fs.readFileSync(process.env.GCP_FILE);
|
||||
const creds = JSON.parse(str);
|
||||
const geminiModel = process.env.GCP_GEMINI_TTS_MODEL || 'gemini-2.5-flash-tts';
|
||||
|
||||
// Test Gemini TTS with model and instructions (both required for Gemini)
|
||||
let result = await synthAudio(stats, {
|
||||
vendor: 'google',
|
||||
credentials: {
|
||||
credentials: {
|
||||
client_email: creds.client_email,
|
||||
private_key: creds.private_key,
|
||||
},
|
||||
},
|
||||
language: 'en-US',
|
||||
voice: 'Kore',
|
||||
model: geminiModel,
|
||||
text: 'Hello, this is a test of Google Gemini text to speech.',
|
||||
instructions: 'Speak clearly and naturally.',
|
||||
renderForCaching: true,
|
||||
});
|
||||
t.ok(!result.servedFromCache, `successfully synthesized Google Gemini TTS audio to ${result.filePath}`);
|
||||
t.ok(result.filePath.endsWith('.mp3'), 'Gemini TTS audio file has correct extension');
|
||||
|
||||
// Test Gemini TTS with different voice and instructions
|
||||
result = await synthAudio(stats, {
|
||||
vendor: 'google',
|
||||
credentials: {
|
||||
credentials: {
|
||||
client_email: creds.client_email,
|
||||
private_key: creds.private_key,
|
||||
},
|
||||
},
|
||||
language: 'en-US',
|
||||
voice: 'Charon',
|
||||
model: geminiModel,
|
||||
text: 'Welcome to our service. How can I help you today?',
|
||||
instructions: 'Speak in a warm, friendly and professional tone.',
|
||||
renderForCaching: true,
|
||||
});
|
||||
t.ok(!result.servedFromCache, `successfully synthesized Gemini TTS with instructions to ${result.filePath}`);
|
||||
|
||||
// Test cache retrieval
|
||||
result = await synthAudio(stats, {
|
||||
vendor: 'google',
|
||||
credentials: {
|
||||
credentials: {
|
||||
client_email: creds.client_email,
|
||||
private_key: creds.private_key,
|
||||
},
|
||||
},
|
||||
language: 'en-US',
|
||||
voice: 'Kore',
|
||||
model: geminiModel,
|
||||
text: 'Hello, this is a test of Google Gemini text to speech.',
|
||||
instructions: 'Speak clearly and naturally.',
|
||||
renderForCaching: true,
|
||||
});
|
||||
t.ok(result.servedFromCache, `successfully retrieved Gemini TTS audio from cache ${result.filePath}`);
|
||||
|
||||
// Test SSML stripping (Gemini doesn't support SSML)
|
||||
result = await synthAudio(stats, {
|
||||
vendor: 'google',
|
||||
credentials: {
|
||||
credentials: {
|
||||
client_email: creds.client_email,
|
||||
private_key: creds.private_key,
|
||||
},
|
||||
},
|
||||
language: 'en-US',
|
||||
voice: 'Leda',
|
||||
model: geminiModel,
|
||||
text: '<speak>This SSML should be stripped for Gemini TTS.</speak>',
|
||||
instructions: 'Speak naturally.',
|
||||
disableTtsCache: true,
|
||||
renderForCaching: true,
|
||||
});
|
||||
t.ok(!result.servedFromCache, `successfully synthesized Gemini TTS with SSML stripped to ${result.filePath}`);
|
||||
|
||||
} catch (err) {
|
||||
console.error(err);
|
||||
t.end(err);
|
||||
}
|
||||
client.quit();
|
||||
});
|
||||
|
||||
test('Google TTS streaming tests (!JAMBONES_DISABLE_TTS_STREAMING)', async(t) => {
|
||||
// Ensure streaming is enabled (default behavior)
|
||||
delete process.env.JAMBONES_DISABLE_TTS_STREAMING;
|
||||
|
||||
// Clear require cache to reload config with new env var
|
||||
delete require.cache[require.resolve('../lib/config')];
|
||||
delete require.cache[require.resolve('../lib/synth-audio')];
|
||||
delete require.cache[require.resolve('..')];
|
||||
|
||||
const fn = require('..');
|
||||
const {synthAudio, client} = fn(opts, logger);
|
||||
|
||||
if (!process.env.GCP_FILE && !process.env.GCP_JSON_KEY) {
|
||||
t.pass('skipping Google TTS streaming tests since neither GCP_FILE nor GCP_JSON_KEY provided');
|
||||
return t.end();
|
||||
}
|
||||
|
||||
try {
|
||||
const str = process.env.GCP_JSON_KEY || fs.readFileSync(process.env.GCP_FILE);
|
||||
const creds = JSON.parse(str);
|
||||
const geminiModel = process.env.GCP_GEMINI_TTS_MODEL || 'gemini-2.5-flash-tts';
|
||||
|
||||
// Test 1: Standard voice streaming (use_live_api=0)
|
||||
let result = await synthAudio(stats, {
|
||||
vendor: 'google',
|
||||
credentials: {
|
||||
credentials: {
|
||||
client_email: creds.client_email,
|
||||
private_key: creds.private_key,
|
||||
},
|
||||
},
|
||||
language: 'en-US',
|
||||
voice: 'en-US-Wavenet-D',
|
||||
gender: 'MALE',
|
||||
text: 'This is a test of standard voice streaming.',
|
||||
disableTtsCache: true
|
||||
});
|
||||
t.ok(result.filePath.startsWith('say:'), 'Standard voice returns streaming say: path');
|
||||
t.ok(result.filePath.includes('vendor=google'), 'Standard voice streaming path contains vendor=google');
|
||||
t.ok(result.filePath.includes('use_live_api=0'), 'Standard voice uses use_live_api=0');
|
||||
t.ok(result.filePath.includes('use_gemini_tts=0'), 'Standard voice uses use_gemini_tts=0');
|
||||
t.ok(result.filePath.includes('voice=en-US-Wavenet-D'), 'Standard voice streaming path contains voice');
|
||||
// Verify credentials are base64 encoded (no raw JSON braces that would break FreeSWitch parsing)
|
||||
t.ok(result.filePath.includes('credentials='), 'Standard voice streaming path contains credentials');
|
||||
t.ok(!result.filePath.includes('credentials={'), 'Credentials are not raw JSON (base64 encoded)');
|
||||
|
||||
// Test 2: HD voice streaming (use_live_api=1)
|
||||
result = await synthAudio(stats, {
|
||||
vendor: 'google',
|
||||
credentials: {
|
||||
credentials: {
|
||||
client_email: creds.client_email,
|
||||
private_key: creds.private_key,
|
||||
},
|
||||
},
|
||||
language: 'en-US',
|
||||
voice: 'en-US-Chirp3-HD-Charon',
|
||||
text: 'This is a test of HD voice streaming.',
|
||||
disableTtsCache: true
|
||||
});
|
||||
t.ok(result.filePath.startsWith('say:'), 'HD voice returns streaming say: path');
|
||||
t.ok(result.filePath.includes('vendor=google'), 'HD voice streaming path contains vendor=google');
|
||||
t.ok(result.filePath.includes('use_live_api=1'), 'HD voice uses use_live_api=1 (Live API)');
|
||||
t.ok(result.filePath.includes('use_gemini_tts=0'), 'HD voice uses use_gemini_tts=0');
|
||||
t.ok(result.filePath.includes('voice=en-US-Chirp3-HD-Charon'), 'HD voice streaming path contains voice');
|
||||
|
||||
// Test 3: Gemini TTS streaming (use_live_api=1)
|
||||
result = await synthAudio(stats, {
|
||||
vendor: 'google',
|
||||
credentials: {
|
||||
credentials: {
|
||||
client_email: creds.client_email,
|
||||
private_key: creds.private_key,
|
||||
},
|
||||
},
|
||||
language: 'en-US',
|
||||
voice: 'Kore',
|
||||
model: geminiModel,
|
||||
text: 'This is a test of Gemini TTS streaming.',
|
||||
instructions: 'Speak naturally.',
|
||||
disableTtsCache: true
|
||||
});
|
||||
t.ok(result.filePath.startsWith('say:'), 'Gemini TTS returns streaming say: path');
|
||||
t.ok(result.filePath.includes('vendor=google'), 'Gemini TTS streaming path contains vendor=google');
|
||||
t.ok(result.filePath.includes('use_live_api=0'), 'Gemini TTS uses use_live_api=0');
|
||||
t.ok(result.filePath.includes('use_gemini_tts=1'), 'Gemini TTS uses use_gemini_tts=1');
|
||||
t.ok(result.filePath.includes(`model_name=${geminiModel}`), 'Gemini TTS streaming path contains model_name');
|
||||
t.ok(result.filePath.includes('prompt=Speak naturally.'), 'Gemini TTS streaming path contains prompt');
|
||||
|
||||
// Test 4: Gemini TTS with SSML stripping in streaming mode
|
||||
result = await synthAudio(stats, {
|
||||
vendor: 'google',
|
||||
credentials: {
|
||||
credentials: {
|
||||
client_email: creds.client_email,
|
||||
private_key: creds.private_key,
|
||||
},
|
||||
},
|
||||
language: 'en-US',
|
||||
voice: 'Leda',
|
||||
model: geminiModel,
|
||||
text: '<speak>This SSML should be stripped.</speak>',
|
||||
instructions: 'Speak naturally.',
|
||||
disableTtsCache: true
|
||||
});
|
||||
t.ok(result.filePath.startsWith('say:'), 'Gemini TTS with SSML returns streaming say: path');
|
||||
t.ok(!result.filePath.includes('<speak>'), 'SSML tags are stripped from streaming path');
|
||||
t.ok(result.filePath.includes('This SSML should be stripped.'), 'Text content is preserved after SSML stripping');
|
||||
|
||||
// Test 5: Gemini TTS with prompt containing special characters
|
||||
result = await synthAudio(stats, {
|
||||
vendor: 'google',
|
||||
credentials: {
|
||||
credentials: {
|
||||
client_email: creds.client_email,
|
||||
private_key: creds.private_key,
|
||||
},
|
||||
},
|
||||
language: 'en-US',
|
||||
voice: 'Kore',
|
||||
model: geminiModel,
|
||||
text: 'Testing special characters in prompt.',
|
||||
options: { prompt: 'Speak in a warm, friendly tone' },
|
||||
disableTtsCache: true
|
||||
});
|
||||
t.ok(result.filePath.startsWith('say:'), 'Gemini TTS with special chars returns streaming say: path');
|
||||
// Commas in prompt should be replaced with semicolons
|
||||
t.ok(result.filePath.includes('prompt=Speak in a warm; friendly tone'), 'Commas in prompt are escaped to semicolons');
|
||||
|
||||
// Test 6: options.useLiveApi override (force live api on standard voice)
|
||||
result = await synthAudio(stats, {
|
||||
vendor: 'google',
|
||||
credentials: {
|
||||
credentials: {
|
||||
client_email: creds.client_email,
|
||||
private_key: creds.private_key,
|
||||
},
|
||||
},
|
||||
language: 'en-US',
|
||||
voice: 'en-US-Wavenet-D',
|
||||
text: 'Testing useLiveApi option override.',
|
||||
options: { useLiveApi: true },
|
||||
disableTtsCache: true
|
||||
});
|
||||
t.ok(result.filePath.includes('use_live_api=1'), 'options.useLiveApi=true overrides default for standard voice');
|
||||
t.ok(result.filePath.includes('use_gemini_tts=0'), 'use_gemini_tts remains 0 for standard voice');
|
||||
|
||||
// Test 7: options.useGeminiTts override (force gemini tts without model)
|
||||
result = await synthAudio(stats, {
|
||||
vendor: 'google',
|
||||
credentials: {
|
||||
credentials: {
|
||||
client_email: creds.client_email,
|
||||
private_key: creds.private_key,
|
||||
},
|
||||
},
|
||||
language: 'en-US',
|
||||
voice: 'Kore',
|
||||
text: 'Testing useGeminiTts option override.',
|
||||
options: { useGeminiTts: true },
|
||||
disableTtsCache: true
|
||||
});
|
||||
t.ok(result.filePath.includes('use_gemini_tts=1'), 'options.useGeminiTts=true overrides default');
|
||||
t.ok(result.filePath.includes('use_live_api=0'), 'use_live_api remains 0 without HD voice');
|
||||
|
||||
// Test 8: Both options override together
|
||||
result = await synthAudio(stats, {
|
||||
vendor: 'google',
|
||||
credentials: {
|
||||
credentials: {
|
||||
client_email: creds.client_email,
|
||||
private_key: creds.private_key,
|
||||
},
|
||||
},
|
||||
language: 'en-US',
|
||||
voice: 'en-US-Wavenet-D',
|
||||
text: 'Testing both options override.',
|
||||
options: { useLiveApi: true, useGeminiTts: true },
|
||||
disableTtsCache: true
|
||||
});
|
||||
t.ok(result.filePath.includes('use_live_api=1'), 'options.useLiveApi=true works with useGeminiTts');
|
||||
t.ok(result.filePath.includes('use_gemini_tts=1'), 'options.useGeminiTts=true works with useLiveApi');
|
||||
|
||||
} catch (err) {
|
||||
console.error(err);
|
||||
t.end(err);
|
||||
}
|
||||
client.quit();
|
||||
});
|
||||
|
||||
test('Google TTS non-streaming tests (JAMBONES_DISABLE_TTS_STREAMING=true)', async(t) => {
|
||||
// Enable streaming disable flag
|
||||
process.env.JAMBONES_DISABLE_TTS_STREAMING = 'true';
|
||||
|
||||
// Clear require cache to reload config with new env var
|
||||
delete require.cache[require.resolve('../lib/config')];
|
||||
delete require.cache[require.resolve('../lib/synth-audio')];
|
||||
delete require.cache[require.resolve('..')];
|
||||
|
||||
const fn = require('..');
|
||||
const {synthAudio, client} = fn(opts, logger);
|
||||
|
||||
if (!process.env.GCP_FILE && !process.env.GCP_JSON_KEY) {
|
||||
t.pass('skipping Google TTS non-streaming tests since neither GCP_FILE nor GCP_JSON_KEY provided');
|
||||
delete process.env.JAMBONES_DISABLE_TTS_STREAMING;
|
||||
return t.end();
|
||||
}
|
||||
|
||||
try {
|
||||
const str = process.env.GCP_JSON_KEY || fs.readFileSync(process.env.GCP_FILE);
|
||||
const creds = JSON.parse(str);
|
||||
const geminiModel = process.env.GCP_GEMINI_TTS_MODEL || 'gemini-2.5-flash-tts';
|
||||
|
||||
// Test 1: Standard voice falls back to non-streaming API
|
||||
let result = await synthAudio(stats, {
|
||||
vendor: 'google',
|
||||
credentials: {
|
||||
credentials: {
|
||||
client_email: creds.client_email,
|
||||
private_key: creds.private_key,
|
||||
},
|
||||
},
|
||||
language: 'en-US',
|
||||
voice: 'en-US-Wavenet-D',
|
||||
gender: 'MALE',
|
||||
text: 'This is a test with streaming disabled.',
|
||||
disableTtsCache: true
|
||||
});
|
||||
t.ok(!result.filePath.startsWith('say:'), 'Standard voice does NOT return streaming say: path when disabled');
|
||||
t.ok(result.filePath.endsWith('.mp3'), 'Standard voice returns mp3 file path');
|
||||
|
||||
// Test 2: HD voice falls back to non-streaming API
|
||||
result = await synthAudio(stats, {
|
||||
vendor: 'google',
|
||||
credentials: {
|
||||
credentials: {
|
||||
client_email: creds.client_email,
|
||||
private_key: creds.private_key,
|
||||
},
|
||||
},
|
||||
language: 'en-US',
|
||||
voice: 'en-US-Chirp3-HD-Charon',
|
||||
text: 'This is a test of HD voice with streaming disabled.',
|
||||
disableTtsCache: true
|
||||
});
|
||||
t.ok(!result.filePath.startsWith('say:'), 'HD voice does NOT return streaming say: path when disabled');
|
||||
t.ok(result.filePath.endsWith('.mp3'), 'HD voice returns mp3 file path');
|
||||
|
||||
// Test 3: Gemini TTS falls back to non-streaming API
|
||||
result = await synthAudio(stats, {
|
||||
vendor: 'google',
|
||||
credentials: {
|
||||
credentials: {
|
||||
client_email: creds.client_email,
|
||||
private_key: creds.private_key,
|
||||
},
|
||||
},
|
||||
language: 'en-US',
|
||||
voice: 'Kore',
|
||||
model: geminiModel,
|
||||
text: 'This is a test of Gemini TTS with streaming disabled.',
|
||||
instructions: 'Speak naturally.',
|
||||
disableTtsCache: true
|
||||
});
|
||||
t.ok(!result.filePath.startsWith('say:'), 'Gemini TTS does NOT return streaming say: path when disabled');
|
||||
t.ok(result.filePath.endsWith('.mp3'), 'Gemini TTS returns mp3 file path');
|
||||
|
||||
} catch (err) {
|
||||
console.error(err);
|
||||
t.end(err);
|
||||
} finally {
|
||||
// Clean up: restore default behavior
|
||||
delete process.env.JAMBONES_DISABLE_TTS_STREAMING;
|
||||
delete require.cache[require.resolve('../lib/config')];
|
||||
delete require.cache[require.resolve('../lib/synth-audio')];
|
||||
delete require.cache[require.resolve('..')];
|
||||
}
|
||||
client.quit();
|
||||
});
|
||||
|
||||
test('AWS speech synth tests', async(t) => {
|
||||
const fn = require('..');
|
||||
const {synthAudio, client} = fn(opts, logger);
|
||||
|
||||
Reference in New Issue
Block a user