support deepgram (#369)

* support deepgram

* wip

src/containers/internal/views/applications/speech-selection.tsx

@@ -25,6 +25,7 @@ import {
   VENDOR_SONIOX,
   VENDOR_WELLSAID,
   VENDOR_WHISPER,
+  useTtsModels,
 } from "src/vendor";
 import {
   LabelOptions,
@@ -92,11 +93,16 @@ export const SpeechProviderSelection = ({
 
   const currentVendor = useRef(synthVendor);
 
+  const ttsModels = useTtsModels();
+
   useEffect(() => {
     currentVendor.current = synthVendor;
     if (!synthesis || synthVendor.startsWith("custom:")) {
       return;
     }
+    if (synthVendor === VENDOR_DEEPGRAM) {
+      return;
+    }
     const voiceOpts = synthesis[synthVendor as keyof SynthesisVendors]
       .filter((lang: VoiceLanguage) => {
         // ELEVENLABS has same voice for all lange, take voices from the 1st language
@@ -206,7 +212,6 @@ export const SpeechProviderSelection = ({
             value={synthVendor}
             options={ttsVendorOptions.filter(
               (vendor) =>
-                vendor.value != VENDOR_DEEPGRAM &&
                 vendor.value != VENDOR_ASSEMBLYAI &&
                 vendor.value != VENDOR_SONIOX &&
                 vendor.value !== VENDOR_CUSTOM &&
@@ -223,6 +228,12 @@ export const SpeechProviderSelection = ({
                 return;
               }
 
+              /** DEEPGRAM only support voice */
+              if (vendor.toString().startsWith(VENDOR_DEEPGRAM)) {
+                setSynthVoice("");
+                return;
+              }
+
               /** When using Google and en-US, ensure "Standard-C" is used as default */
               if (
                 e.target.value === VENDOR_GOOGLE &&
@@ -282,8 +293,21 @@ export const SpeechProviderSelection = ({
               />
             </>
           )}
+          {ttsModels && synthVendor === VENDOR_DEEPGRAM && (
+            <>
+              <label htmlFor="synthesis_lang">Model</label>
+              <Selector
+                id="synthesis_voice"
+                name="synthesis_voice"
+                value={synthVoice}
+                options={ttsModels[synthVendor]}
+                onChange={(e) => setSynthVoice(e.target.value)}
+              />
+            </>
+          )}
           {synthVendor &&
             !synthVendor.toString().startsWith(VENDOR_CUSTOM) &&
+            synthVendor !== VENDOR_DEEPGRAM &&
             synthLang && (
               <>
                 <label htmlFor="synthesis_lang">Language</label>

src/containers/internal/views/speech-services/form.tsx

@@ -639,8 +639,7 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
         </fieldset>
         {vendor && (
           <fieldset>
-            {vendor !== VENDOR_DEEPGRAM &&
-              vendor !== VENDOR_ASSEMBLYAI &&
+            {vendor !== VENDOR_ASSEMBLYAI &&
               vendor !== VENDOR_COBALT &&
               vendor !== VENDOR_SONIOX &&
               vendor != VENDOR_CUSTOM && (

src/vendor/index.tsx

@@ -94,14 +94,22 @@ export const useTtsModels = () => {
     Promise.all([
       import("./speech-synthsis-models/elevenlabs-models"),
       import("./speech-synthsis-models/whisper-models"),
-    ]).then(([{ default: elevenlabs }, { default: whisper }]) => {
-      if (!ignore) {
-        setModels({
-          elevenlabs,
-          whisper,
-        });
+      import("./speech-synthsis-models/deepgram-models"),
+    ]).then(
+      ([
+        { default: elevenlabs },
+        { default: whisper },
+        { default: deepgram },
+      ]) => {
+        if (!ignore) {
+          setModels({
+            elevenlabs,
+            whisper,
+            deepgram,
+          });
+        }
       }
-    });
+    );
     return function cleanup() {
       ignore = true;
     };
@@ -207,6 +215,8 @@ export const useSpeechVendors = () => {
               nvidia: nvidiaynthesis,
               elevenlabs: elevenLabsSynthesis,
               whisper: whisperSynthesis,
+              // Deepgram just have model which includes language, voice
+              deepgram: [],
             },
             recognizers: {
               aws: awsRecognizer,

src/vendor/speech-synthsis-models/deepgram-models.ts

@@ -0,0 +1,13 @@
+import type { Model } from "../types";
+
+export const models: Model[] = [
+  { name: "Aurora English (US) Female", value: "alpha-aurora-en-v2" },
+  { name: "Asteria English (US) Female", value: "alpha-asteria-en-v2" },
+  { name: "Artemis English (UK) Female", value: "alpha-artemis-en-v3" },
+  { name: "Andromeda English (US) Female", value: "alpha-andromeda-en-v3" },
+  { name: "Stella	English (UK) Female", value: "alpha-stella-en-v2" },
+  { name: "Orion	English (US)	Male", value: "alpha-orion-en-v2" },
+  { name: "Atlas	English (US)	Male", value: "alpha-atlas-en-v3" },
+];
+
+export default models;

src/vendor/types.ts

@@ -73,6 +73,7 @@ export interface RegionVendors {
 export interface TtsModels {
   elevenlabs: Model[];
   whisper: Model[];
+  deepgram: Model[];
 }
 
 export interface RecognizerVendors {
@@ -98,6 +99,7 @@ export interface SynthesisVendors {
   nvidia: VoiceLanguage[];
   elevenlabs: VoiceLanguage[];
   whisper: VoiceLanguage[];
+  deepgram: VoiceLanguage[];
 }
 
 export interface MSRawSpeech {