Merge pull request #678 from nature-heart-software/dev

release

.dockerignore

@@ -0,0 +1,3 @@
+node_modules
+npm-debug.log
+yarn-error.log

.yarnrc

@@ -0,0 +1,2 @@
+httpTimeout 600000
+network-timeout 600000

Dockerfile

@@ -0,0 +1,33 @@
+FROM mcr.microsoft.com/windows/servercore:ltsc2019
+
+WORKDIR /app
+
+RUN icacls "C:\app" /grant:r everyone:(OI)(CI)F /t
+
+COPY ./scripts/install-nodejs.ps1 ./scripts/install-nodejs.ps1
+RUN powershell -Command ./scripts/install-nodejs.ps1
+
+COPY ./scripts/install-yarn.ps1 ./scripts/install-yarn.ps1
+RUN powershell -Command ./scripts/install-yarn.ps1
+
+COPY ./scripts/install-python.ps1 ./scripts/install-python.ps1
+RUN powershell -Command ./scripts/install-python.ps1
+
+COPY ./scripts/install-vs-build-tools.ps1 ./scripts/install-vs-build-tools.ps1
+RUN powershell -Command ./scripts/install-vs-build-tools.ps1
+
+ENV PATH="C:\nodejs;C:\Program Files (x86)\Yarn\bin;C:\Python310;C:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Tools\MSVC\14.30.30705\bin\Hostx64\x64;${PATH}"
+
+ENV PYTHON="C:\Python310\python.exe"
+
+ENV NODE_OPTIONS="--max-old-space-size=4096"
+
+RUN npm config set python C:\Python310\python.exe
+
+RUN npm config set msvs_version 2022
+
+COPY ./ ./
+
+COPY ./lerna.json.prod ./lerna.json
+
+RUN npm run build

README.md

@@ -7,7 +7,7 @@
 </h1>
 
 <p align="center">
-  Your speech Assistant ✨
+  Your speech assistant ✨
 </p>
 
 <p align="center">
@@ -87,7 +87,7 @@ Here's a list of all the text-to-speech engines that are supported in Izabela:
 | Izabela (multiple engines) | Yes                                 | [Download](https://github.com/nature-heart-software/izabela/blob/dev/assets/izabela-sample.mp3?raw=true)             | https://github.com/Weilbyte/tiktok-tts                                        |
 | Amazon Polly               | Yes                                 | [Download](https://github.com/nature-heart-software/izabela/blob/dev/assets/amazon-polly-sample.mp3?raw=true)        | https://aws.amazon.com/polly/                                                 |
 | Google Cloud TTS           | Yes                                 | [Download](https://github.com/nature-heart-software/izabela/blob/dev/assets/google-cloud-tts-sample.mp3?raw=true)    | https://cloud.google.com/text-to-speech                                       |
-| IBM Watson TTS             | Yes                                 |                                                                                                                      | https://www.ibm.com/cloud/watson-text-to-speech                               |
+| IBM Watson TTS             | Temporarily unavailable             |                                                                                                                      | https://www.ibm.com/cloud/watson-text-to-speech                               |
 | Microsoft Azure TTS        | Yes                                 | [Download](https://github.com/nature-heart-software/izabela/blob/dev/assets/microsoft-azure-tts-sample.mp3?raw=true) | https://azure.microsoft.com/en-us/products/cognitive-services/text-to-speech/ |
 | Say                        | Included by default                 | [Download](https://github.com/nature-heart-software/izabela/blob/dev/assets/say-sample.mp3?raw=true)                 | https://github.com/Marak/say.js/                                              |
 | Sam                        | Included by default                 | [Download](https://github.com/nature-heart-software/izabela/blob/dev/assets/sam-sample.mp3?raw=true)                 | https://github.com/discordier/sam / https://github.com/Imrane03/better-sam    |
@@ -101,3 +101,14 @@ Find guides and API documentation on the [Wiki](https://github.com/nature-heart-
 ## Resources
 
 - [Figma](https://www.figma.com/proto/U4A6IwSY8T4W2tm2agW92S/Izabela-v1.0.0?node-id=103%3A4&scaling=min-zoom&page-id=103%3A3&starting-point-node-id=103%3A4)
+
+## Development
+
+The project requires the following in order to run:
+
+- Node 16
+- Yarn
+- Python 3.10
+- Visual Studio Build Tools 2019 or higher
+  - Desktop development with C++
+  - Node.js build tools

apps/app-server/src/plugins/speech-apis/el.ts

@@ -1,13 +1,11 @@
 import { RequestHandler } from 'express'
 import axios from 'axios'
 import { handleError } from '../../utils/requests'
-import hash from 'object-hash'
 
 const plugin: Izabela.Server.Plugin = ({ app }) => {
   const api = axios.create({
     baseURL: 'https://api.elevenlabs.io/v1',
   })
-  let settingsCache: string = ''
   const listVoicesHandler: RequestHandler = async (
     {
       body: {
@@ -30,32 +28,53 @@ const plugin: Izabela.Server.Plugin = ({ app }) => {
     }
   }
 
+  const listModelsHandler: RequestHandler = async (
+    {
+      body: {
+        credentials: { apiKey },
+      },
+    },
+    res,
+  ) => {
+    try {
+      const { data: models } = await api.get('/models', {
+        headers: {
+          'xi-api-key': apiKey,
+        },
+      })
+      res.status(200).json(models)
+    } catch (e: any) {
+      handleError(res, 'Internal server error', e.message, 500)
+    }
+  }
+
   const synthesizeSpeechHandler: RequestHandler = async (
     {
       body: {
         credentials: { apiKey },
-        payload: { text, voice, stability, similarity_boost },
+        payload: {
+          text,
+          voice,
+          stability,
+          similarity_boost,
+          use_speaker_boost,
+          style,
+          model_id,
+        },
       },
     },
     res,
   ) => {
     try {
-      const settings = {
+      const voice_settings = {
         stability,
         similarity_boost,
-      }
-      const hashedSettings = hash({ ...settings, voice_id: voice.voice_id })
-      if (settingsCache !== hashedSettings) {
-        await api.post(`/voices/${voice.voice_id}/settings/edit`, settings, {
-          headers: {
-            'xi-api-key': apiKey,
-          },
-        })
-        settingsCache = hashedSettings
+        use_speaker_boost,
+        style,
       }
       const { data } = await api.post(
         `/text-to-speech/${voice.voice_id}/stream`,
-        { text },
+        { model_id, text, voice_settings },
         {
           headers: {
             'xi-api-key': apiKey,
@@ -73,6 +92,7 @@ const plugin: Izabela.Server.Plugin = ({ app }) => {
     }
   }
   app.post('/api/tts/elevenlabs/list-voices', listVoicesHandler)
+  app.post('/api/tts/elevenlabs/list-models', listModelsHandler)
   app.post('/api/tts/elevenlabs/synthesize-speech', synthesizeSpeechHandler)
 }
 

apps/app/src/electron/app.ts

@@ -72,6 +72,10 @@ const App = () => {
   }
 
   const onQuit = () => {
+    /* fixes app.quit(): https://stackoverflow.com/a/75369483 */
+    ElectronWindowManager.getInstances().forEach(({ window }) => {
+      window.removeAllListeners()
+    })
     destroyWinMouse()
   }
 

apps/app/src/features/speech/components/inputs/NvSpeechEngineSelect.vue

@@ -6,23 +6,29 @@ import { NvSelect } from '@packages/ui'
 import { computed } from 'vue'
 import { useSpeechEngineManager } from '@/modules/speech-engine-manager'
 import { orderBy } from 'lodash'
+import { groupOptions } from '@/utils/select'
+import { capitalize } from '@/utils/text'
 
 const { engines } = useSpeechEngineManager()
 const options = computed(() =>
-  orderBy(
-    engines.value.map((engine) => {
-      const disabled = engine.hasCredentials ? !engine.hasCredentials() : false
-      return {
-        disabled,
-        label: engine.name,
-        value: engine.id,
-        attrs: {
-          title: disabled ? 'Requires credentials' : '',
-        },
-      }
-    }),
-    ['disabled', 'label'],
-    ['asc', 'asc'],
+  groupOptions(
+    orderBy(
+      engines.value.map((engine) => {
+        const disabled = engine.hasCredentials ? !engine.hasCredentials() : false
+        return {
+          disabled,
+          label: engine.name,
+          value: engine.id,
+          category: capitalize(engine.category),
+          attrs: {
+            title: disabled ? 'Requires credentials' : '',
+          },
+        }
+      }),
+      ['disabled', 'label'],
+      ['asc', 'asc'],
+    ),
+    'category',
   ),
 )
 </script>

apps/app/src/modules/speech-engine-manager/types.ts

@@ -9,6 +9,7 @@ export type Payload = { [key: string]: any }
 export interface SpeechEngine {
   id: string
   name: string
+  category: 'cloud' | 'local' | 'other'
   getVoiceName: (voice: any) => string
   getSelectedVoice: () => any
   getCredentials: () => Credentials

apps/app/src/plugins/speech-engines/amazon-polly/NvVoiceSelect.vue

@@ -8,17 +8,34 @@
       ...$attrs,
     }"
     valueKey="Id"
-  />
+  >
+    <template #optionAfter="{ option, hover }">
+      <span v-show="(!option.children && hover) || favoriteVoiceIds.includes(option.id)">
+        <NvButton
+          :icon-name="favoriteVoiceIds.includes(option.id) ? 'times' : 'heart'"
+          :title="
+            favoriteVoiceIds.includes(option.id) ? 'Remove from favorites' : 'Add to favorites'
+          "
+          size="sm"
+          type="default"
+          @mousedown.prevent.stop="
+            setProperty('favoriteVoiceIds', xor(favoriteVoiceIds, [option.id]))
+          "
+        />
+      </span>
+    </template>
+  </NvSelect>
 </template>
 <script lang="ts" setup>
 import { computed, watch } from 'vue'
 import { useQueryClient } from 'vue-query'
 import { purify } from '@packages/toolbox'
-import { orderBy } from 'lodash'
-import { NvSelect } from '@packages/ui'
+import { orderBy, xor } from 'lodash'
+import { NvButton, NvSelect } from '@packages/ui'
 import { useSpeechStore } from '@/features/speech/store'
+import { groupOptions } from '@/utils/select'
 import { useListVoicesQuery } from './hooks'
-import { getVoiceName, LIST_VOICES_QUERY_KEY } from './shared'
+import { getVoiceCategory, getVoiceId, getVoiceName, LIST_VOICES_QUERY_KEY } from './shared'
 import { getProperty, setProperty } from './store'
 
 const queryClient = useQueryClient()
@@ -40,12 +57,32 @@ const { data, isFetching } = useListVoicesQuery(computedParams, {
   enabled: canFetch,
 })
 const voices = computed(() => orderBy(data.value || [], ['LanguageCode', 'Name']))
-const options = computed(() =>
-  voices.value.map((voice) => ({
-    label: getVoiceName(voice),
-    value: voice,
-  })),
-)
+const getOptionFromVoice = (voice: any) => ({
+  id: getVoiceId(voice),
+  label: getVoiceName(voice),
+  value: voice,
+  category: getVoiceCategory(voice),
+})
+
+const options = computed(() => {
+  const localOptions = groupOptions(voices.value.map(getOptionFromVoice), 'category')
+  const favoriteVoiceIds = getProperty('favoriteVoiceIds')
+  if (favoriteVoiceIds) {
+    const favoriteVoices = voices.value.filter((voice: any) =>
+      favoriteVoiceIds.includes(getVoiceId(voice)),
+    )
+    if (favoriteVoices.length) {
+      localOptions.unshift({
+        label: 'Favorites',
+        children: favoriteVoices.map(getOptionFromVoice),
+      })
+    }
+  }
+  return localOptions
+})
+
+const favoriteVoiceIds = computed<string[]>(() => getProperty('favoriteVoiceIds'))
+
 watch(
   () => [getProperty('identityPoolId', true), getProperty('region')],
   () => canFetch.value && queryClient.refetchQueries(LIST_VOICES_QUERY_KEY),

apps/app/src/plugins/speech-engines/amazon-polly/register.ts

@@ -15,6 +15,7 @@ const getSelectedVoice = () => getProperty('selectedVoice')
 registerEngine({
   id: ENGINE_ID,
   name: ENGINE_NAME,
+  category: 'cloud',
   getSelectedVoice,
   getVoiceName,
   getCredentials,

apps/app/src/plugins/speech-engines/amazon-polly/shared.ts

@@ -2,3 +2,5 @@ export const ENGINE_ID = 'aptts' as const
 export const ENGINE_NAME = 'Amazon Polly' as const
 export const LIST_VOICES_QUERY_KEY = 'aptts-list-voices' as const
 export const getVoiceName = (voice: any) => `${voice.LanguageCode} ${voice.Name} - ${voice.Gender}`
+export const getVoiceId = (voice: any) => voice.Id
+export const getVoiceCategory = (voice: any) => voice.LanguageName

apps/app/src/plugins/speech-engines/amazon-polly/store.ts

@@ -12,5 +12,6 @@ export const { setProperty, getProperty } = definePluginStore(ENGINE_ID, {
     Name: 'Amy',
     SupportedEngines: ['neural', 'standard'],
   },
+  favoriteVoiceIds: [],
   useLocalCredentials: false,
 })