utilitas 1999.1.70 → 1999.1.72

package/lib/alan.mjs

@@ -1,8 +1,8 @@
 import { checkSearch, distill, search } from './web.mjs';
 import { create as createUoid } from './uoid.mjs';
-import { createWavHeader } from './media.mjs';
 import { end, loop } from './event.mjs';
 import { fileTypeFromBuffer } from 'file-type';
+import { packPcmToWav } from './media.mjs';
 import { v4 as uuidv4 } from 'uuid';
 
 import {
@@ -707,12 +707,8 @@ const packResp = async (resp, options) => {
         const str = simpleText.indexOf(x);
         str >= 0 && (simpleText = simpleText.slice(0, str).trim());
     });
-    audio && (audio = Buffer.isBuffer(audio) ? audio : await convert(audio, {
-        input: BASE64, expected: BUFFER,
-    })) && audio.length && (audio = Buffer.concat([
-        createWavHeader(audio.length), audio
-    ])) && (audio = await convert(audio, {
-        input: BUFFER, expected: BUFFER, ...options || {},
+    audio = await ignoreErrFunc(async () => await packPcmToWav(audio, {
+        input: Buffer.isBuffer(audio) ? BUFFER : BASE64, expected: BUFFER,
     }));
     if (images?.length) {
         for (let i in images) {
@@ -1592,6 +1588,7 @@ export {
     getSession,
     init,
     initChat,
+    k,
     listFiles,
     listGptFineTuningEvents,
     listGptFineTuningJobs,

package/lib/manifest.mjs

@@ -1,7 +1,7 @@
 const manifest = {
     "name": "utilitas",
     "description": "Just another common utility for JavaScript.",
-    "version": "1999.1.70",
+    "version": "1999.1.72",
     "private": false,
     "homepage": "https://github.com/Leask/utilitas",
     "main": "index.mjs",
@@ -30,7 +30,6 @@ const manifest = {
         "@ffprobe-installer/ffprobe": "^2.1.2",
         "@google-cloud/speech": "^7.1.0",
         "@google-cloud/storage": "^7.16.0",
-        "@google-cloud/text-to-speech": "^6.1.0",
         "@google-cloud/vision": "^5.1.0",
         "@google/genai": "^1.0.0",
         "@mozilla/readability": "github:mozilla/readability",

package/lib/media.mjs

@@ -38,6 +38,17 @@ const createWavHeader = (
     return header;
 };
 
+const packPcmToWav = async (audio, options) => {
+    (audio = await convert(audio, { ...options || {}, expected: BUFFER })) // DON'T override expected
+        && audio.length
+        && (audio = Buffer.concat([createWavHeader(audio.length), audio]))
+        && (audio = await convert(audio, {
+            expected: BUFFER, ...options || {}, input: BUFFER, // DON'T override input
+        }));
+    assert(audio, 'Failed to pack PCM to WAV.', 500);
+    return audio;
+};
+
 // https://codex.so/ffmpeg-node-js
 const getFfmpeg = async (options) => {
     const ffmpeg = await need('fluent-ffmpeg');
@@ -92,4 +103,5 @@ export {
     convertAudioTo16kNanoPcmWave,
     createWavHeader,
     getFfmpeg,
+    packPcmToWav,
 };

package/lib/speech.mjs

@@ -1,9 +1,9 @@
-import { DEFAULT_MODELS, OPENAI_VOICE } from './alan.mjs';
+import { DEFAULT_MODELS, OPENAI_VOICE, countTokens, k } from './alan.mjs';
 import { getApiKeyCredentials, hash } from './encryption.mjs';
-import { getFfmpeg } from './media.mjs';
+import { getFfmpeg, packPcmToWav } from './media.mjs';
 import { get } from './web.mjs';
 import { convert, getTempPath } from './storage.mjs';
-import { ensureString } from './utilitas.mjs';
+import { ensureString, mergeAtoB } from './utilitas.mjs';
 
 import {
     call, countKeys, ignoreErrFunc, inBrowser,
@@ -17,19 +17,32 @@ import {
 
 const _NEED = [
     '@google-cloud/speech',
-    '@google-cloud/text-to-speech',
+    '@google/genai',
     'OpenAI',
     'whisper-node',
 ];
 
 const WHISPER_DEFAULT_MODEL = 'base';
 const errorMessage = 'Invalid audio data.';
-const [BUFFER, STREAM, BASE64, FILE, clients, languageCode, audioEncoding, suffix, SPEAKER, cleanup]
-    = ['BUFFER', 'STREAM', 'BASE64', 'FILE', {}, 'en-US', 'OGG_OPUS', 'ogg', 'SPEAKER', true];
-const [GPT_4O_MIMI_TTS, GPT_4O_TRANSCRIBE, OPENAI_TTS_MAX_LENGTH]
-    = ['gpt-4o-mini-tts', 'gpt-4o-transcribe', 4096];
-const [defaultOpenAITtsModel, defaultOpenAISttModel]
-    = [GPT_4O_MIMI_TTS, GPT_4O_TRANSCRIBE];
+
+const [
+    BUFFER, STREAM, BASE64, FILE, clients, languageCode, audioEncoding, suffix,
+    SPEAKER, cleanup, wav,
+] = [
+        'BUFFER', 'STREAM', 'BASE64', 'FILE', {}, 'en-US', 'OGG_OPUS', 'ogg',
+        'SPEAKER', true, 'wav'
+    ];
+
+const [
+    GPT_4O_MIMI_TTS, GPT_4O_TRANSCRIBE, GEMINI_25_PRO_TTS, GEMINI_25_FLASH_TTS,
+    OPENAI_TTS_MAX_LENGTH,
+] = [
+        'gpt-4o-mini-tts', 'gpt-4o-transcribe', 'gemini-2.5-pro-preview-tts',
+        'gemini-2.5-flash-preview-tts', 4096
+    ];
+
+const [defaultOpenAITtsModel, defaultOpenAISttModel, defaultGeminiTtsModel]
+    = [GPT_4O_MIMI_TTS, GPT_4O_TRANSCRIBE, GEMINI_25_PRO_TTS];
 
 const WHISPER_MODELS = [
     // npx whisper-node download tiny.en
@@ -104,13 +117,14 @@ const init = async (options) => {
                 break;
             case 'GOOGLE':
                 clients._provider = provider;
-                const sslCreds = await getApiKeyCredentials(options);
                 if (options?.tts) {
-                    const tts = (await need('@google-cloud/text-to-speech')).default;
-                    clients.tts = new tts.TextToSpeechClient({ sslCreds });
+                    let { GoogleGenAI } = await need('@google/genai');
+                    let client = new GoogleGenAI(options);
+                    clients.tts = client.models.generateContent;
                 }
                 if (options?.stt) {
                     const stt = (await need('@google-cloud/speech')).default;
+                    const sslCreds = await getApiKeyCredentials(options);
                     clients.stt = new stt.SpeechClient({ sslCreds });
                 }
                 break;
@@ -159,15 +173,29 @@ const ttsOpenAI = async (input, options) => {
     return await convert(buffer, { suffix, ...options || {} });
 };
 
-const ttsGoogle = async (text, options) => {
+// https://ai.google.dev/gemini-api/docs/speech-generation#voices
+const ttsGoogle = async (contents, options) => {
     assert(clients.tts, 'Google TTS API has not been initialized.', 500);
-    assert(text, 'Text is required.', 400);
-    const [response] = await clients.tts.synthesizeSpeech({
-        input: { text, ...options?.input || {} },
-        voice: { languageCode, name: 'en-US-Wavenet-F', ...options?.voice || {} },
-        audioConfig: { audioEncoding, ...options?.audioConfig || {} },
+    assert(contents, 'Text is required.', 400);
+    assert(await countTokens(contents) <= k(32), 'Text is too long.', 400);
+    const resp = await clients.tts({
+        model: options?.model || defaultGeminiTtsModel, contents,
+        config: mergeAtoB(options?.config, {
+            responseModalities: ['AUDIO'],
+            speechConfig: {
+                voiceConfig: {
+                    prebuiltVoiceConfig: {
+                        voiceName: options?.voice || 'Leda',
+                    },
+                },
+            },
+        }),
+    });
+    const rawAudio = resp?.candidates?.[0]?.content?.parts?.[0]?.inlineData;
+    assert(rawAudio, 'Failed to generate audio.', 500);
+    return options?.raw ? rawAudio : await packPcmToWav(rawAudio?.data, {
+        input: BASE64, expected: 'FILE', suffix: wav, ...options || {},
     });
-    return await convert(response.audioContent, { suffix, ...options || {} });
 };
 
 const ttsSay = async (text, options) => {
@@ -275,13 +303,16 @@ const stt = async (audio, options) => {
 export default init;
 export {
     _NEED,
+    OPENAI_TTS_MAX_LENGTH,
     checkSay,
     checkWhisper,
-    init, OPENAI_TTS_MAX_LENGTH, stt, sttGoogle,
+    init,
+    stt,
+    sttGoogle,
     sttOpenAI,
     sttWhisper,
     tts,
     ttsGoogle,
     ttsOpenAI,
-    ttsSay
+    ttsSay,
 };

package/package.json

@@ -1,7 +1,7 @@
 {
     "name": "utilitas",
     "description": "Just another common utility for JavaScript.",
-    "version": "1999.1.70",
+    "version": "1999.1.72",
     "private": false,
     "homepage": "https://github.com/Leask/utilitas",
     "main": "index.mjs",
@@ -41,7 +41,6 @@
         "@ffprobe-installer/ffprobe": "^2.1.2",
         "@google-cloud/speech": "^7.1.0",
         "@google-cloud/storage": "^7.16.0",
-        "@google-cloud/text-to-speech": "^6.1.0",
         "@google-cloud/vision": "^5.1.0",
         "@google/genai": "^1.0.0",
         "@mozilla/readability": "github:mozilla/readability",