From 3d44d773e8545116ff320f91f8104f335c497ebe Mon Sep 17 00:00:00 2001
From: Valentyne Stigloher <pixunil@posteo.eu>
Date: Sun, 14 Sep 2025 19:47:26 +0200
Subject: [PATCH] refactor: list available voices independent of locale configs

---
 app/components/Pronunciation.vue        |  46 +++---
 app/components/PronunciationInput.vue   |  88 +++++-------
 app/components/PronunciationSpeaker.vue |  10 +-
 locale/_base/config.suml                |   6 +-
 locale/ar/config.suml                   |  10 +-
 locale/config.ts                        |  43 +-----
 locale/en/config.suml                   |   6 +-
 locale/eo/config.suml                   |   1 -
 locale/es/config.suml                   |  10 +-
 locale/et/config.suml                   |   6 +-
 locale/fo/config.suml                   |   5 -
 locale/fr/config.suml                   |  10 +-
 locale/gl/config.suml                   |   3 +-
 locale/hbs/config.suml                  |  14 +-
 locale/he/config.suml                   |   7 +-
 locale/it/config.suml                   |   6 +-
 locale/ja/config.suml                   |   6 +-
 locale/ko/config.suml                   |   6 +-
 locale/nb/config.suml                   |   6 +-
 locale/nl/config.suml                   |   6 +-
 locale/nn/config.suml                   |   6 +-
 locale/pl/config.suml                   |   6 +-
 locale/pt/config.suml                   |  10 +-
 locale/ro/config.suml                   |   6 +-
 locale/ru/config.suml                   |   6 +-
 locale/sv/config.suml                   |   6 +-
 locale/tok/config.suml                  |   6 +-
 locale/tr/config.suml                   |   6 +-
 locale/vi/config.suml                   |   1 -
 locale/yi/config.suml                   |   1 -
 locale/zh/config.suml                   |   6 +-
 server/express/pronounce.ts             |  17 +--
 shared/pronunciation/voices.ts          | 177 ++++++++++++++++++++++++
 33 files changed, 272 insertions(+), 277 deletions(-)
 create mode 100644 shared/pronunciation/voices.ts

diff --git a/app/components/Pronunciation.vue b/app/components/Pronunciation.vue
index 03d8ff677..043d3d528 100644
--- a/app/components/Pronunciation.vue
+++ b/app/components/Pronunciation.vue
@@ -1,3 +1,22 @@
+<script setup lang="ts">
+import type { VoiceKey } from '#shared/pronunciation/voices.ts';
+import useConfig from '~/composables/useConfig.ts';
+
+defineProps<{
+    pronunciation: string;
+    text?: boolean;
+}>();
+
+const config = useConfig();
+
+const voices = computed((): VoiceKey[] => {
+    if (!config.pronunciation?.enabled) {
+        return [];
+    }
+    return config.pronunciation.voices;
+});
+</script>
+
 <template>
     <span class="text-nowrap">
         <span v-if="text" class="text-pronunciation">
@@ -13,33 +32,6 @@
     </span>
 </template>
 
-<script lang="ts">
-import { defineComponent } from 'vue';
-
-import useConfig from '../composables/useConfig.ts';
-
-export default defineComponent({
-    props: {
-        pronunciation: { required: true, type: String },
-        text: { default: false, type: Boolean },
-    },
-    setup() {
-        return {
-            config: useConfig(),
-        };
-    },
-    computed: {
-        voices(): string[] {
-            if (this.config.pronunciation?.enabled) {
-                return Object.keys(this.config.pronunciation.voices);
-            } else {
-                return [];
-            }
-        },
-    },
-});
-</script>
-
 <style lang="scss" scoped>
 @import "~/assets/variables";
 
diff --git a/app/components/PronunciationInput.vue b/app/components/PronunciationInput.vue
index 727e1647f..aebfb07d4 100644
--- a/app/components/PronunciationInput.vue
+++ b/app/components/PronunciationInput.vue
@@ -1,3 +1,41 @@
+<script setup lang="ts">
+import { escapePronunciationString, unescapePronunciationString } from '#shared/helpers.ts';
+import type { VoiceKey } from '#shared/pronunciation/voices.ts';
+import useConfig from '~/composables/useConfig.ts';
+
+const modelValue = defineModel<string | null>({ required: true });
+
+const config = useConfig();
+
+const rawPronunciation = computed({
+    get(): string {
+        if (modelValue.value) {
+            const phonemes = modelValue.value.substring(1, modelValue.value.length - 1);
+            return unescapePronunciationString(phonemes);
+        } else {
+            return '';
+        }
+    },
+    set(rawPronunciation: string) {
+        let pronunciation;
+        if (rawPronunciation) {
+            pronunciation = `/${escapePronunciationString(rawPronunciation)}/`;
+        } else {
+            pronunciation = null;
+        }
+
+        modelValue.value = pronunciation;
+    },
+});
+
+const voices = computed((): VoiceKey[] => {
+    if (!config.pronunciation?.enabled) {
+        return [];
+    }
+    return config.pronunciation.voices;
+});
+</script>
+
 <template>
     <div class="input-group input-group-sm w-auto">
         <span class="input-group-text">/</span>
@@ -17,53 +55,3 @@
         />
     </div>
 </template>
-
-<script lang="ts">
-import { defineComponent } from 'vue';
-import type { PropType } from 'vue';
-
-import useConfig from '../composables/useConfig.ts';
-
-import { escapePronunciationString, unescapePronunciationString } from '#shared/helpers.ts';
-
-export default defineComponent({
-    props: {
-        modelValue: { default: null, type: String as PropType<string | null> },
-    },
-    emits: ['update:modelValue'],
-    setup() {
-        return {
-            config: useConfig(),
-        };
-    },
-    computed: {
-        rawPronunciation: {
-            get(): string {
-                if (this.modelValue) {
-                    const phonemes = this.modelValue.substring(1, this.modelValue.length - 1);
-                    return unescapePronunciationString(phonemes);
-                } else {
-                    return '';
-                }
-            },
-            set(rawPronunciation: string) {
-                let pronunciation;
-                if (rawPronunciation) {
-                    pronunciation = `/${escapePronunciationString(rawPronunciation)}/`;
-                } else {
-                    pronunciation = null;
-                }
-
-                this.$emit('update:modelValue', pronunciation);
-            },
-        },
-        voices(): string[] {
-            if (this.config.pronunciation?.enabled) {
-                return Object.keys(this.config.pronunciation.voices);
-            } else {
-                return [];
-            }
-        },
-    },
-});
-</script>
diff --git a/app/components/PronunciationSpeaker.vue b/app/components/PronunciationSpeaker.vue
index 286f483ad..fcc7ac82e 100644
--- a/app/components/PronunciationSpeaker.vue
+++ b/app/components/PronunciationSpeaker.vue
@@ -1,9 +1,11 @@
 <script setup lang="ts">
 import { Base64 } from 'js-base64';
 
+import type { VoiceKey } from '#shared/pronunciation/voices.ts';
+
 const props = withDefaults(defineProps<{
     pronunciation?: string | null;
-    voice: string;
+    voice: VoiceKey;
     button?: boolean;
 }>(), {
     pronunciation: null,
@@ -58,16 +60,16 @@ const icon = computed((): string => {
 const config = useConfig();
 const name = computed((): string | null => {
     if (!config.pronunciation?.enabled) {
-        return props.voice;
+        return props.voice.toUpperCase();
     }
 
-    const voices = Object.keys(config.pronunciation.voices);
+    const voices = config.pronunciation.voices;
     if (voices.length === 1 && props.voice === voices[0]) {
         // don’t show voice name if it is considered the main voice for this locale
         return null;
     }
 
-    return props.voice;
+    return props.voice.toUpperCase();
 });
 </script>
 
diff --git a/locale/_base/config.suml b/locale/_base/config.suml
index 4e23ac97c..339946ded 100644
--- a/locale/_base/config.suml
+++ b/locale/_base/config.suml
@@ -99,11 +99,7 @@ pronouns:
 
 pronunciation:
     enabled: true
-    voices:
-        GB:
-            language: 'en-GB'
-            voice: 'Emma'
-            engine: 'neural'
+    voices: ['gb']
 
 sources:
     enabled: true
diff --git a/locale/ar/config.suml b/locale/ar/config.suml
index 23027610e..6d1d6ecb1 100644
--- a/locale/ar/config.suml
+++ b/locale/ar/config.suml
@@ -51,15 +51,7 @@ pronouns:
 
 pronunciation:
     enabled: true
-    voices:
-        AR:
-            language: 'arb'
-            voice: 'Zeina'
-            engine: 'standard'
-        AE:
-            language: 'ar-AE'
-            voice: 'Hala'
-            engine: 'neural'
+    voices: ['ar', 'ae']
 
 sources:
     enabled: false
diff --git a/locale/config.ts b/locale/config.ts
index e9170fa54..7648c915a 100644
--- a/locale/config.ts
+++ b/locale/config.ts
@@ -1,7 +1,6 @@
-import type { Engine, LanguageCode, VoiceId } from '@aws-sdk/client-polly';
-
 import type { Category } from '#shared/classes.ts';
 import type { GrammarTablesDefinition } from '#shared/language/grammarTables.ts';
+import type { VoiceKey } from '#shared/pronunciation/voices.ts';
 import type { LocaleCode } from '~~/locale/locales.ts';
 
 export type Toggable<T> = ({ enabled: true } & T) | { enabled: false } & Partial<T>;
@@ -404,45 +403,7 @@ interface PronunciationConfig {
      * whether to treat letter graphemes as phonemes. useful when the configured locale has no matching voice
      */
     ipa?: boolean;
-    voices: Record<string, PronunciationVoiceConfig>;
-}
-
-export type PronunciationVoiceConfig = AwsPollyPronunciationVoiceConfig | NarakeetPronunciationVoiceConfig;
-
-/**
- * @see https://docs.aws.amazon.com/polly/latest/dg/voicelist.html
- */
-export interface AwsPollyPronunciationVoiceConfig {
-    /**
-     * text-to-speech provider (aws_polly is the default, if not specified)
-     */
-    provider?: 'aws_polly';
-    /**
-     * language code
-     */
-    language: LanguageCode;
-    /**
-     * voice name
-     */
-    voice: VoiceId;
-    /**
-     * voice engine (default: 'standard')
-     */
-    engine?: Engine;
-}
-export interface NarakeetPronunciationVoiceConfig {
-    /**
-     * text-to-speech provider
-     */
-    provider: 'narakeet';
-    /**
-     * language code
-     */
-    language: string;
-    /**
-     * voice name
-     */
-    voice: string;
+    voices: VoiceKey[];
 }
 
 interface SourcesConfig {
diff --git a/locale/en/config.suml b/locale/en/config.suml
index 821883c44..4d6452aeb 100644
--- a/locale/en/config.suml
+++ b/locale/en/config.suml
@@ -149,11 +149,7 @@ pronouns:
 
 pronunciation:
     enabled: true
-    voices:
-        GB:
-            language: 'en-GB'
-            voice: 'Emma'
-            engine: 'neural'
+    voices: ['gb']
 
 sources:
     enabled: true
diff --git a/locale/eo/config.suml b/locale/eo/config.suml
index 37c0964ed..391875c9e 100644
--- a/locale/eo/config.suml
+++ b/locale/eo/config.suml
@@ -97,7 +97,6 @@ pronouns:
 
 pronunciation:
     enabled: false
-    voices: {}
 
 sources:
     enabled: true
diff --git a/locale/es/config.suml b/locale/es/config.suml
index c499117cb..f511e604d 100644
--- a/locale/es/config.suml
+++ b/locale/es/config.suml
@@ -128,15 +128,7 @@ pronouns:
 
 pronunciation:
     enabled: true
-    voices:
-        ES:
-            language: 'es-ES'
-            voice: 'Lucia'
-            engine: 'standard'
-        MX:
-            language: 'es-MX'
-            voice: 'Mia'
-            engine: 'standard'
+    voices: ['es', 'mx']
 
 sources:
     enabled: true
diff --git a/locale/et/config.suml b/locale/et/config.suml
index beac10e81..b1e8d2807 100644
--- a/locale/et/config.suml
+++ b/locale/et/config.suml
@@ -37,11 +37,7 @@ pronouns:
 
 pronunciation:
     enabled: false
-    voices:
-        FI:
-            language: 'fi-FI'
-            voice: 'Suvi'
-            engine: 'neural'
+    voices: ['fi']
 
 sources:
     enabled: true
diff --git a/locale/fo/config.suml b/locale/fo/config.suml
index abff8f2fc..588661263 100644
--- a/locale/fo/config.suml
+++ b/locale/fo/config.suml
@@ -133,11 +133,6 @@ pronouns:
 
 pronunciation:
     enabled: false
-    voices:
-        GB:
-            language: 'en-GB'
-            voice: 'Emma'
-            engine: 'neural'
 
 sources:
     enabled: true
diff --git a/locale/fr/config.suml b/locale/fr/config.suml
index 17007485b..d02dd4ae5 100644
--- a/locale/fr/config.suml
+++ b/locale/fr/config.suml
@@ -65,15 +65,7 @@ pronouns:
 
 pronunciation:
     enabled: true
-    voices:
-        FR:
-            language: 'fr-FR'
-            voice: 'Lea'
-            engine: 'standard'
-        CA:
-            language: 'fr-CA'
-            voice: 'Gabrielle'
-            engine: 'neural'
+    voices: ['fr', 'ca']
 
 sources:
     enabled: true
diff --git a/locale/gl/config.suml b/locale/gl/config.suml
index 218a20992..95948e842 100644
--- a/locale/gl/config.suml
+++ b/locale/gl/config.suml
@@ -79,8 +79,7 @@ pronouns:
     shortMorphemes: 3
 
 pronunciation:
-    enabled: true
-    voices: {}
+    enabled: false
 
 sources:
     enabled: true
diff --git a/locale/hbs/config.suml b/locale/hbs/config.suml
index 0c84853ac..fa2bacd2d 100644
--- a/locale/hbs/config.suml
+++ b/locale/hbs/config.suml
@@ -98,19 +98,7 @@ pronouns:
 
 pronunciation:
     enabled: true
-    voices:
-        BOS:
-            language: 'bos'
-            voice: 'suada'
-            provider: 'narakeet'
-        HRV:
-            language: 'hrv'
-            voice: 'jasna'
-            provider: 'narakeet'
-        SRP:
-            language: 'srp'
-            voice: 'milica-latin'
-            provider: 'narakeet'
+    voices: ['bos', 'hrv', 'srp']
 
 sources:
     enabled: true
diff --git a/locale/he/config.suml b/locale/he/config.suml
index 29bf441ab..b0d75bc24 100644
--- a/locale/he/config.suml
+++ b/locale/he/config.suml
@@ -44,12 +44,7 @@ pronouns:
     others: 'Other pronouns'
 
 pronunciation:
-    enabled: true
-    voices:
-        GB:
-            language: 'en-GB'
-            voice: 'Emma'
-            engine: 'neural'
+    enabled: false
 
 sources:
     enabled: true
diff --git a/locale/it/config.suml b/locale/it/config.suml
index a922d069c..33f824b6f 100644
--- a/locale/it/config.suml
+++ b/locale/it/config.suml
@@ -324,11 +324,7 @@ pronouns:
 
 pronunciation:
     enabled: true
-    voices:
-        IT:
-            language: 'it-IT'
-            voice: 'Bianca'
-            engine: 'neural'
+    voices: ['it']
 
 sources:
     enabled: true
diff --git a/locale/ja/config.suml b/locale/ja/config.suml
index fe65d26a6..11a202799 100644
--- a/locale/ja/config.suml
+++ b/locale/ja/config.suml
@@ -45,11 +45,7 @@ pronouns:
 
 pronunciation:
     enabled: true
-    voices:
-        JA:
-            language: 'ja-JP'
-            voice: 'Mizuki'
-            engine: 'standard'
+    voices: ['ja']
 
 sources:
     enabled: true
diff --git a/locale/ko/config.suml b/locale/ko/config.suml
index e2cbc096d..7ecf5878a 100644
--- a/locale/ko/config.suml
+++ b/locale/ko/config.suml
@@ -40,11 +40,7 @@ pronouns:
 
 pronunciation:
     enabled: true
-    voices:
-        KO:
-            language: 'ko-KR'
-            voice: 'Seoyeon'
-            engine: 'neural'
+    voices: ['ko']
 
 sources:
     enabled: true
diff --git a/locale/nb/config.suml b/locale/nb/config.suml
index a61c94eaa..8dd1a2c88 100644
--- a/locale/nb/config.suml
+++ b/locale/nb/config.suml
@@ -120,11 +120,7 @@ pronouns:
 
 pronunciation:
     enabled: true
-    voices:
-        NO:
-            language: 'nb-NO'
-            voice: 'Liv'
-            engine: 'standard'
+    voices: ['no']
 
 sources:
     enabled: true
diff --git a/locale/nl/config.suml b/locale/nl/config.suml
index 10c13da1e..c6e06d6f1 100644
--- a/locale/nl/config.suml
+++ b/locale/nl/config.suml
@@ -71,11 +71,7 @@ pronouns:
 
 pronunciation:
     enabled: true
-    voices:
-        GB:
-            language: 'nl-NL'
-            voice: 'Ruben'
-            engine: 'standard'
+    voices: ['nl']
 
 sources:
     enabled: true
diff --git a/locale/nn/config.suml b/locale/nn/config.suml
index fac7e7fe4..e13757ef0 100644
--- a/locale/nn/config.suml
+++ b/locale/nn/config.suml
@@ -121,11 +121,7 @@ pronouns:
 
 pronunciation:
     enabled: true
-    voices:
-        NO:
-            language: 'nb-NO'
-            voice: 'Liv'
-            engine: 'standard'
+    voices: ['no']
 
 sources:
     enabled: true
diff --git a/locale/pl/config.suml b/locale/pl/config.suml
index 7f58a87be..2202d508a 100644
--- a/locale/pl/config.suml
+++ b/locale/pl/config.suml
@@ -322,11 +322,7 @@ pronouns:
 
 pronunciation:
     enabled: true
-    voices:
-        PL:
-            language: 'pl-PL'
-            voice: 'Ewa'
-            engine: 'standard'
+    voices: ['pl']
 
 sources:
     enabled: true
diff --git a/locale/pt/config.suml b/locale/pt/config.suml
index 495e8f585..b9e272103 100644
--- a/locale/pt/config.suml
+++ b/locale/pt/config.suml
@@ -81,15 +81,7 @@ pronouns:
 
 pronunciation:
     enabled: true
-    voices:
-        PT:
-            language: 'pt-PT'
-            voice: 'Cristiano'
-            engine: 'standard'
-        BR:
-            language: 'pt-BR'
-            voice: 'Vitoria'
-            engine: 'standard'
+    voices: ['pt', 'br']
 
 sources:
     enabled: true
diff --git a/locale/ro/config.suml b/locale/ro/config.suml
index ddc2b3dd5..a37854ef8 100644
--- a/locale/ro/config.suml
+++ b/locale/ro/config.suml
@@ -113,11 +113,7 @@ pronouns:
 
 pronunciation:
     enabled: true
-    voices:
-        RO:
-            language: 'ro-RO'
-            voice: 'Carmen'
-            engine: 'standard'
+    voices: ['ro']
 
 sources:
     enabled: true
diff --git a/locale/ru/config.suml b/locale/ru/config.suml
index 7dbafb26b..1806691b1 100644
--- a/locale/ru/config.suml
+++ b/locale/ru/config.suml
@@ -162,11 +162,7 @@ pronouns:
 
 pronunciation:
     enabled: true
-    voices:
-        GB:
-            language: 'ru-RU'
-            voice: 'Tatyana'
-            engine: 'standard'
+    voices: ['ru']
 
 sources:
     enabled: true
diff --git a/locale/sv/config.suml b/locale/sv/config.suml
index a9faf6eca..f5b797335 100644
--- a/locale/sv/config.suml
+++ b/locale/sv/config.suml
@@ -50,11 +50,7 @@ pronouns:
 
 pronunciation:
     enabled: true
-    voices:
-        SE:
-            language: 'sv-SE'
-            voice: 'Astrid'
-            engine: 'standard'
+    voices: ['se']
 
 sources:
     enabled: true
diff --git a/locale/tok/config.suml b/locale/tok/config.suml
index 72baf5bd3..4fac59595 100644
--- a/locale/tok/config.suml
+++ b/locale/tok/config.suml
@@ -29,11 +29,7 @@ pronouns:
 pronunciation:
     enabled: false
     ipa: true
-    voices:
-        TOK:
-            language: 'es-US'
-            voice: 'Lupe'
-            engine: 'standard'
+    voices: ['tok']
 
 sources:
     enabled: false
diff --git a/locale/tr/config.suml b/locale/tr/config.suml
index 479ba1f37..9d855515a 100644
--- a/locale/tr/config.suml
+++ b/locale/tr/config.suml
@@ -11,11 +11,7 @@ pronouns:
 
 pronunciation:
     enabled: true
-    voices:
-        TR:
-            language: 'tr-TR'
-            voice: 'Filiz'
-            engine: 'standard'
+    voices: ['tr']
 
 sources:
     enabled: false
diff --git a/locale/vi/config.suml b/locale/vi/config.suml
index 8940520e1..45b046260 100644
--- a/locale/vi/config.suml
+++ b/locale/vi/config.suml
@@ -83,7 +83,6 @@ pronouns:
 
 pronunciation:
     enabled: false
-    voices: {}
 
 sources:
     enabled: true
diff --git a/locale/yi/config.suml b/locale/yi/config.suml
index b415929d5..95dfca4b4 100644
--- a/locale/yi/config.suml
+++ b/locale/yi/config.suml
@@ -35,7 +35,6 @@ pronouns:
 
 pronunciation:
     enabled: false
-    voices: {}
 
 sources:
     enabled: true
diff --git a/locale/zh/config.suml b/locale/zh/config.suml
index fac22a1dd..004bc8cbc 100644
--- a/locale/zh/config.suml
+++ b/locale/zh/config.suml
@@ -35,11 +35,7 @@ pronouns:
 
 pronunciation:
     enabled: true
-    voices:
-        CN:
-            language: 'cmn-CN'
-            voice: 'Zhiyu'
-            engine: 'standard'
+    voices: ['cn']
 
 sources:
     enabled: true
diff --git a/server/express/pronounce.ts b/server/express/pronounce.ts
index f1313b55b..fc9aa4a03 100644
--- a/server/express/pronounce.ts
+++ b/server/express/pronounce.ts
@@ -14,11 +14,8 @@ import {
     convertPronunciationStringToNarakeetFormat,
     handleErrorAsync,
 } from '#shared/helpers.ts';
-import type {
-    PronunciationVoiceConfig,
-    AwsPollyPronunciationVoiceConfig,
-    NarakeetPronunciationVoiceConfig,
-} from '~~/locale/config.ts';
+import { voices } from '#shared/pronunciation/voices.ts';
+import type { VoiceKey, AwsPollyVoice, NarakeetVoice, Voice } from '#shared/pronunciation/voices.ts';
 import { getLocale, loadConfig, localesDict } from '~~/server/data.ts';
 
 const router = Router();
@@ -26,7 +23,7 @@ const router = Router();
 type ProviderKey = 'aws_polly' | 'narakeet';
 interface Provider {
     tokenised(text: string): string;
-    generate(textTokenised: string, voice: PronunciationVoiceConfig): Promise<[Uint8Array, string]>;
+    generate(textTokenised: string, voice: Voice): Promise<[Uint8Array, string]>;
 }
 
 const providers: Record<ProviderKey, Provider> = {
@@ -34,7 +31,7 @@ const providers: Record<ProviderKey, Provider> = {
         tokenised(text: string): string {
             return convertPronunciationStringToSsml(text);
         },
-        async generate(textTokenised: string, voice: AwsPollyPronunciationVoiceConfig): Promise<[Uint8Array, string]> {
+        async generate(textTokenised: string, voice: AwsPollyVoice): Promise<[Uint8Array, string]> {
             const polly = new Polly(awsConfig) as NodeJsClient<Polly>;
 
             const pollyResponse = await polly.synthesizeSpeech({
@@ -56,7 +53,7 @@ const providers: Record<ProviderKey, Provider> = {
         tokenised(text: string): string {
             return convertPronunciationStringToNarakeetFormat(text);
         },
-        async generate(textTokenised: string, voice: NarakeetPronunciationVoiceConfig): Promise<[Uint8Array, string]> {
+        async generate(textTokenised: string, voice: NarakeetVoice): Promise<[Uint8Array, string]> {
             const url = `https://api.narakeet.com/text-to-speech/mp3?voice=${voice.voice}&language=${voice.language}`;
 
             const response = await fetch(url, {
@@ -91,11 +88,11 @@ router.get('/pronounce/:voice/:pronunciation', handleErrorAsync(async (req, res)
     const locale = getLocale(getH3Event(req));
     const config = await loadConfig(locale);
 
-    const voice: PronunciationVoiceConfig | undefined = config.pronunciation?.voices?.[req.params.voice];
-    if (!voice) {
+    if (!(req.params.voice in voices)) {
         return res.status(404).json({ error: 'Not found' });
     }
 
+    const voice = voices[req.params.voice as VoiceKey];
     const provider = providers[(voice.provider || 'aws_polly') as ProviderKey];
     const tokenised = provider.tokenised(text);
     const key = `pronunciation/${config.locale}-${req.params.voice}/${sha1(tokenised)}.mp3`;
diff --git a/shared/pronunciation/voices.ts b/shared/pronunciation/voices.ts
new file mode 100644
index 000000000..8a75ea7d0
--- /dev/null
+++ b/shared/pronunciation/voices.ts
@@ -0,0 +1,177 @@
+import type { Engine, LanguageCode, VoiceId } from '@aws-sdk/client-polly';
+
+export type Voice = AwsPollyVoice | NarakeetVoice;
+
+/**
+ * @see https://docs.aws.amazon.com/polly/latest/dg/voicelist.html
+ */
+export interface AwsPollyVoice {
+    /**
+     * text-to-speech provider (aws_polly is the default, if not specified)
+     */
+    provider?: 'aws_polly';
+    /**
+     * language code
+     */
+    language: LanguageCode;
+    /**
+     * voice name
+     */
+    voice: VoiceId;
+    /**
+     * voice engine (default: 'standard')
+     */
+    engine?: Engine;
+}
+
+export interface NarakeetVoice {
+    /**
+     * text-to-speech provider
+     */
+    provider: 'narakeet';
+    /**
+     * language code
+     */
+    language: string;
+    /**
+     * voice name
+     */
+    voice: string;
+}
+
+const defineVoices = <T extends Record<string, Voice>>(voices: T) => voices as Readonly<Record<keyof T, Voice>>;
+
+export const voices = defineVoices({
+    ar: {
+        language: 'arb',
+        voice: 'Zeina',
+        engine: 'standard',
+    },
+    ae: {
+        language: 'ar-AE',
+        voice: 'Hala',
+        engine: 'neural',
+    },
+    de: {
+        language: 'de-DE',
+        voice: 'Vicki',
+        engine: 'standard',
+    },
+    gb: {
+        language: 'en-GB',
+        voice: 'Emma',
+        engine: 'neural',
+    },
+    es: {
+        language: 'es-ES',
+        voice: 'Lucia',
+        engine: 'standard',
+    },
+    mx: {
+        language: 'es-MX',
+        voice: 'Mia',
+        engine: 'standard',
+    },
+    fi: {
+        language: 'fi-FI',
+        voice: 'Suvi',
+        engine: 'neural',
+    },
+    fr: {
+        language: 'fr-FR',
+        voice: 'Lea',
+        engine: 'standard',
+    },
+    ca: {
+        language: 'fr-CA',
+        voice: 'Gabrielle',
+        engine: 'neural',
+    },
+    bos: {
+        language: 'bos',
+        voice: 'suada',
+        provider: 'narakeet',
+    },
+    hrv: {
+        language: 'hrv',
+        voice: 'jasna',
+        provider: 'narakeet',
+    },
+    srp: {
+        language: 'srp',
+        voice: 'milica-latin',
+        provider: 'narakeet',
+    },
+    it: {
+        language: 'it-IT',
+        voice: 'Bianca',
+        engine: 'neural',
+    },
+    ja: {
+        language: 'ja-JP',
+        voice: 'Mizuki',
+        engine: 'standard',
+    },
+    ko: {
+        language: 'ko-KR',
+        voice: 'Seoyeon',
+        engine: 'neural',
+    },
+    nb: {
+        language: 'nb-NO',
+        voice: 'Liv',
+        engine: 'standard',
+    },
+    nl: {
+        language: 'nl-NL',
+        voice: 'Ruben',
+        engine: 'standard',
+    },
+    pl: {
+        language: 'pl-PL',
+        voice: 'Ewa',
+        engine: 'standard',
+    },
+    pt: {
+        language: 'pt-PT',
+        voice: 'Cristiano',
+        engine: 'standard',
+    },
+    br: {
+        language: 'pt-BR',
+        voice: 'Vitoria',
+        engine: 'standard',
+    },
+    ro: {
+        language: 'ro-RO',
+        voice: 'Carmen',
+        engine: 'standard',
+    },
+    ru: {
+        language: 'ru-RU',
+        voice: 'Tatyana',
+        engine: 'standard',
+    },
+    se: {
+        language: 'sv-SE',
+        voice: 'Astrid',
+        engine: 'standard',
+    },
+    tok: {
+        language: 'es-US',
+        voice: 'Lupe',
+        engine: 'standard',
+    },
+    tr: {
+        language: 'tr-TR',
+        voice: 'Filiz',
+        engine: 'standard',
+    },
+    cn: {
+        language: 'cmn-CN',
+        voice: 'Zhiyu',
+        engine: 'standard',
+    },
+});
+
+export type VoiceKey = keyof typeof voices;