ben.carrasco/jambonz-webapp
1
0
Fork 0
mirror of https://github.com/jambonz/jambonz-webapp.git synced 2026-07-04 19:21:58 +00:00
Code Issues Packages Projects Releases Wiki Activity

support mod_playht_tts (#423)

Browse Source 
* support mod_playht_tts

* wip

* wip

* fix code style
This commit is contained in:
Hoan Luu Huu
2024-04-08 21:24:06 +07:00
committed by GitHub
parent ea2713a021
commit bc68eb8e71
5 changed files with 126 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/api/constants.ts
+14 -1
View File
@@ -4,6 +4,7 @@ import type {
LimitField,
LimitUnitOption,
PasswordSettings,
PlayHTOptions,
SelectorOptions,
SipGateway,
SmppGateway,
@@ -211,7 +212,7 @@ export const GOOGLE_CUSTOM_VOICES_REPORTED_USAGE = [
{ name: "REALTIME", value: "REALTIME" },
{ name: "OFFLINE", value: "OFFLINE" },
];
// Eleven Labs options
// ElevenLabs options
export const DEFAULT_ELEVENLABS_OPTIONS: Partial<ElevenLabsOptions> = {
optimize_streaming_latency: 3,
voice_settings: {
@@ -220,6 +221,18 @@ export const DEFAULT_ELEVENLABS_OPTIONS: Partial<ElevenLabsOptions> = {
use_speaker_boost: true,
},
};
// PlayHT options
export const DEFAULT_PLAYHT_OPTIONS: Partial<PlayHTOptions> = {
quality: "medium",
speed: 1,
seed: 1,
temperature: 1,
emotion: "female_happy",
voice_guidance: 3,
style_guidance: 20,
text_guidance: 1,
};
/** Password Length options */
export const PASSWORD_MIN = 8;
src/api/types.ts
+13
View File
@@ -390,6 +390,7 @@ export interface SpeechCredential {
region: null | string;
aws_region: null | string;
api_key: null | string;
user_id: null | string;
access_key_id: null | string;
secret_access_key: null | string;
service_key: null | string;
@@ -415,6 +416,7 @@ export interface SpeechCredential {
label: null | string;
cobalt_server_uri: null | string;
model_id: null | string;
voice_engine: null | string;
model: null | string;
options: null | string;
deepgram_stt_uri: null | string;
@@ -718,3 +720,14 @@ export interface ElevenLabsOptions {
use_speaker_boost: boolean;
}>;
}
export interface PlayHTOptions {
quality: string;
speed: number;
seed: number;
temperature: number;
emotion: string;
voice_guidance: number;
style_guidance: number;
text_guidance: number;
}
src/containers/internal/views/speech-services/form.tsx
+92 -6
View File
@@ -40,6 +40,7 @@ import {
VENDOR_ELEVENLABS,
VENDOR_ASSEMBLYAI,
VENDOR_WHISPER,
VENDOR_PLAYHT,
} from "src/vendor";
import { MSG_REQUIRED_FIELDS } from "src/constants";
import {
@@ -69,6 +70,7 @@ import { setAccountFilter, setLocation } from "src/store/localStore";
import {
DEFAULT_ELEVENLABS_OPTIONS,
DEFAULT_GOOGLE_CUSTOM_VOICES_REPORTED_USAGE,
DEFAULT_PLAYHT_OPTIONS,
DISABLE_CUSTOM_SPEECH,
GOOGLE_CUSTOM_VOICES_REPORTED_USAGE,
} from "src/api/constants";
@@ -93,6 +95,7 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
);
const [region, setRegion] = useState("");
const [apiKey, setApiKey] = useState("");
const [userId, setUserId] = useState("");
const [accessKeyId, setAccessKeyId] = useState("");
const [secretAccessKey, setSecretAccessKey] = useState("");
const [clientId, setClientId] = useState("");
@@ -179,6 +182,30 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
});
};
const getDefaultVendorOptions = () => {
if (vendor) {
switch (vendor) {
case VENDOR_ELEVENLABS:
return DEFAULT_ELEVENLABS_OPTIONS;
case VENDOR_PLAYHT:
return DEFAULT_PLAYHT_OPTIONS;
}
}
return "";
};
const getDefaultVendorApiDoc = () => {
if (vendor) {
switch (vendor) {
case VENDOR_ELEVENLABS:
return "https://elevenlabs.io/docs/api-reference/streaming";
case VENDOR_PLAYHT:
return "https://docs.play.ht/reference/api-generate-tts-audio-stream";
}
}
return "";
};
const handlePutGoogleCustomVoices = () => {
if (!credential || !credential.data) {
return;
@@ -297,9 +324,13 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
...((vendor === VENDOR_ELEVENLABS || vendor === VENDOR_WHISPER) && {
model_id: ttsModelId || null,
}),
...(vendor === VENDOR_ELEVENLABS && {
...((vendor === VENDOR_ELEVENLABS || vendor === VENDOR_PLAYHT) && {
options: options || null,
}),
...(vendor === VENDOR_PLAYHT &&
ttsModelId && {
voice_engine: ttsModelId,
}),
...(vendor === VENDOR_DEEPGRAM && {
deepgram_stt_uri: deepgramSttUri || null,
deepgram_stt_use_tls: deepgramSttUseTls ? 1 : 0,
@@ -345,10 +376,15 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
vendor === VENDOR_ASSEMBLYAI ||
vendor === VENDOR_SONIOX ||
vendor === VENDOR_ELEVENLABS ||
vendor === VENDOR_PLAYHT ||
vendor === VENDOR_WHISPER
? apiKey
: null,
}),
...(vendor === VENDOR_PLAYHT &&
userId && {
user_id: userId,
}),
riva_server_uri: vendor == VENDOR_NVIDIA ? rivaServerUri : null,
})
.then(({ json }) => {
@@ -379,7 +415,11 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
};
useEffect(() => {
if (vendor === VENDOR_ELEVENLABS || vendor === VENDOR_WHISPER) {
if (
vendor === VENDOR_ELEVENLABS ||
vendor === VENDOR_WHISPER ||
vendor === VENDOR_PLAYHT
) {
getSpeechSupportedLanguagesAndVoices(
currentServiceProvider?.service_provider_sid,
vendor,
@@ -389,7 +429,9 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
setTtsModels(json.models);
if (
json.models.length > 0 &&
(vendor === VENDOR_ELEVENLABS || vendor === VENDOR_WHISPER)
(vendor === VENDOR_ELEVENLABS ||
vendor === VENDOR_WHISPER ||
vendor === VENDOR_PLAYHT)
) {
setTtsModelId(json.models[0].value);
}
@@ -566,6 +608,14 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
}
setInitialDeepgramOnpremCheck(hasValue(credential?.data?.deepgram_stt_uri));
setIsDeepgramOnpremEnabled(hasValue(credential?.data?.deepgram_stt_uri));
if (credential?.data?.user_id) {
setUserId(credential.data.user_id);
}
if (credential?.data?.voice_engine) {
setTtsModelId(credential.data.voice_engine);
}
}, [credential]);
const updateCustomVoices = (
@@ -692,6 +742,7 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
{vendor !== VENDOR_WELLSAID &&
vendor !== VENDOR_CUSTOM &&
vendor !== VENDOR_WHISPER &&
vendor !== VENDOR_PLAYHT &&
vendor !== VENDOR_ELEVENLABS && (
<label htmlFor="use_for_stt" className="chk">
<input
@@ -1133,12 +1184,33 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
/>
</fieldset>
)}
{(vendor === VENDOR_WELLSAID ||
vendor === VENDOR_ASSEMBLYAI ||
vendor == VENDOR_ELEVENLABS ||
vendor === VENDOR_WHISPER ||
vendor === VENDOR_PLAYHT ||
vendor === VENDOR_SONIOX) && (
<fieldset>
{vendor === VENDOR_PLAYHT && (
<>
<label htmlFor={`${vendor}_userid`}>
User ID<span>*</span>
</label>
<input
id="playht_user_id"
type="text"
name="playht_user_id"
placeholder="User ID"
required
value={userId}
onChange={(e) => {
setUserId(e.target.value);
}}
disabled={credential ? true : false}
/>
</>
)}
<label htmlFor={`${vendor}_apikey`}>
API key<span>*</span>
</label>
@@ -1153,6 +1225,20 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
/>
</fieldset>
)}
{vendor === VENDOR_PLAYHT && ttsModels.length > 0 && (
<fieldset>
<label htmlFor={`${vendor}_tts_model_id`}>Model</label>
<Selector
id={"tts_model_id"}
name={"tts_model_id"}
value={ttsModelId}
options={ttsModels}
onChange={(e) => {
setTtsModelId(e.target.value);
}}
/>
</fieldset>
)}
{vendor === VENDOR_DEEPGRAM && (
<fieldset>
<label htmlFor={`${vendor}_apikey`}>
@@ -1184,7 +1270,7 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
/>
</fieldset>
)}
{vendor === VENDOR_ELEVENLABS && (
{(vendor === VENDOR_ELEVENLABS || vendor === VENDOR_PLAYHT) && (
<fieldset>
<Checkzone
hidden
@@ -1196,7 +1282,7 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
setOptions(
tmpOptions
? tmpOptions
: JSON.stringify(DEFAULT_ELEVENLABS_OPTIONS, null, 2),
: JSON.stringify(getDefaultVendorOptions(), null, 2),
);
}
if (!e.target.checked) {
@@ -1219,7 +1305,7 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
}}
>
<a
href="https://elevenlabs.io/docs/api-reference/streaming"
href={getDefaultVendorApiDoc()}
target="_blank"
rel="noopener noreferrer"
style={{ marginRight: "10px", fontSize: "16px" }}
src/vendor/index.tsx
Vendored
+5
View File
@@ -20,6 +20,7 @@ export const VENDOR_COBALT = "cobalt";
export const VENDOR_ELEVENLABS = "elevenlabs";
export const VENDOR_ASSEMBLYAI = "assemblyai";
export const VENDOR_WHISPER = "whisper";
export const VENDOR_PLAYHT = "playht";
export const vendors: VendorOptions[] = [
{
@@ -78,6 +79,10 @@ export const vendors: VendorOptions[] = [
name: "Whisper",
value: VENDOR_WHISPER,
},
{
name: "PlayHT",
value: VENDOR_PLAYHT,
},
].sort((a, b) => a.name.localeCompare(b.name)) as VendorOptions[];
export const useRegionVendors = () => {
src/vendor/types.ts
Vendored
+2 -1
View File
@@ -12,7 +12,8 @@ export type Vendor =
| "Custom"
| "ElevenLabs"
| "assemblyai"
| "whisper";
| "whisper"
| "playht";
export interface VendorOptions {
name: Vendor;