@speech-sdk/core 0.0.3 → 0.0.5
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +37 -0
- package/dist/audio-tags.d.ts +6 -0
- package/dist/audio-tags.d.ts.map +1 -0
- package/dist/audio-tags.js +17 -0
- package/dist/audio-tags.js.map +1 -0
- package/dist/errors.d.ts +1 -1
- package/dist/errors.d.ts.map +1 -1
- package/dist/errors.js +5 -5
- package/dist/errors.js.map +1 -1
- package/dist/generate-speech.d.ts +2 -2
- package/dist/generate-speech.d.ts.map +1 -1
- package/dist/generate-speech.js +29 -6
- package/dist/generate-speech.js.map +1 -1
- package/dist/index.d.ts +6 -5
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +4 -2
- package/dist/index.js.map +1 -1
- package/dist/provider-utils.d.ts.map +1 -1
- package/dist/provider-utils.js +16 -12
- package/dist/provider-utils.js.map +1 -1
- package/dist/providers/cartesia/index.d.ts +9 -3
- package/dist/providers/cartesia/index.d.ts.map +1 -1
- package/dist/providers/cartesia/index.js +159 -15
- package/dist/providers/cartesia/index.js.map +1 -1
- package/dist/providers/deepgram/index.d.ts +2 -2
- package/dist/providers/deepgram/index.d.ts.map +1 -1
- package/dist/providers/deepgram/index.js +16 -9
- package/dist/providers/deepgram/index.js.map +1 -1
- package/dist/providers/elevenlabs/index.d.ts +9 -5
- package/dist/providers/elevenlabs/index.d.ts.map +1 -1
- package/dist/providers/elevenlabs/index.js +166 -33
- package/dist/providers/elevenlabs/index.js.map +1 -1
- package/dist/providers/fal/index.d.ts +6 -6
- package/dist/providers/fal/index.d.ts.map +1 -1
- package/dist/providers/fal/index.js +52 -17
- package/dist/providers/fal/index.js.map +1 -1
- package/dist/providers/fish-audio/index.d.ts +6 -2
- package/dist/providers/fish-audio/index.d.ts.map +1 -1
- package/dist/providers/fish-audio/index.js +24 -10
- package/dist/providers/fish-audio/index.js.map +1 -1
- package/dist/providers/google/index.d.ts +3 -3
- package/dist/providers/google/index.d.ts.map +1 -1
- package/dist/providers/google/index.js +80 -15
- package/dist/providers/google/index.js.map +1 -1
- package/dist/providers/hume/index.d.ts +3 -3
- package/dist/providers/hume/index.d.ts.map +1 -1
- package/dist/providers/hume/index.js +43 -15
- package/dist/providers/hume/index.js.map +1 -1
- package/dist/providers/mistral/index.d.ts +2 -2
- package/dist/providers/mistral/index.d.ts.map +1 -1
- package/dist/providers/mistral/index.js +33 -16
- package/dist/providers/mistral/index.js.map +1 -1
- package/dist/providers/murf/index.d.ts +3 -3
- package/dist/providers/murf/index.d.ts.map +1 -1
- package/dist/providers/murf/index.js +62 -14
- package/dist/providers/murf/index.js.map +1 -1
- package/dist/providers/openai/index.d.ts +4 -4
- package/dist/providers/openai/index.d.ts.map +1 -1
- package/dist/providers/openai/index.js +99 -18
- package/dist/providers/openai/index.js.map +1 -1
- package/dist/providers/resemble/index.d.ts +2 -2
- package/dist/providers/resemble/index.d.ts.map +1 -1
- package/dist/providers/resemble/index.js +41 -10
- package/dist/providers/resemble/index.js.map +1 -1
- package/dist/providers/unreal-speech/index.d.ts +2 -2
- package/dist/providers/unreal-speech/index.d.ts.map +1 -1
- package/dist/providers/unreal-speech/index.js +20 -13
- package/dist/providers/unreal-speech/index.js.map +1 -1
- package/dist/resolve-provider.d.ts +1 -1
- package/dist/resolve-provider.d.ts.map +1 -1
- package/dist/resolve-provider.js +34 -34
- package/dist/resolve-provider.js.map +1 -1
- package/dist/speech-provider.d.ts +10 -6
- package/dist/speech-provider.d.ts.map +1 -1
- package/dist/speech-result.d.ts +4 -3
- package/dist/speech-result.d.ts.map +1 -1
- package/dist/speech-result.js +9 -7
- package/dist/speech-result.js.map +1 -1
- package/dist/types.d.ts +7 -7
- package/dist/types.d.ts.map +1 -1
- package/package.json +6 -2
|
@@ -1,16 +1,23 @@
|
|
|
1
|
-
import {
|
|
1
|
+
import { handleErrorResponse, resolveApiKey } from "../../provider-utils.js";
|
|
2
2
|
export class DeepgramSpeechProvider {
|
|
3
|
-
id =
|
|
4
|
-
defaultModel =
|
|
3
|
+
id = "deepgram";
|
|
4
|
+
defaultModel = "aura-2";
|
|
5
5
|
models = [
|
|
6
|
-
{
|
|
6
|
+
{
|
|
7
|
+
id: "aura-2",
|
|
8
|
+
audioTags: false,
|
|
9
|
+
languages: ["en", "es", "de", "fr", "it", "ja", "nl"],
|
|
10
|
+
releaseDate: "2025-04-15",
|
|
11
|
+
openSource: false,
|
|
12
|
+
inlineVoiceCloning: false,
|
|
13
|
+
},
|
|
7
14
|
];
|
|
8
15
|
apiKey;
|
|
9
16
|
baseURL;
|
|
10
17
|
fetchFn;
|
|
11
18
|
constructor(config) {
|
|
12
19
|
this.apiKey = config.apiKey;
|
|
13
|
-
this.baseURL = config.baseURL ??
|
|
20
|
+
this.baseURL = config.baseURL ?? "https://api.deepgram.com/v1";
|
|
14
21
|
this.fetchFn = config.fetch ?? globalThis.fetch;
|
|
15
22
|
}
|
|
16
23
|
async generate(options) {
|
|
@@ -23,10 +30,10 @@ export class DeepgramSpeechProvider {
|
|
|
23
30
|
text: options.text,
|
|
24
31
|
};
|
|
25
32
|
const response = await this.fetchFn(url, {
|
|
26
|
-
method:
|
|
33
|
+
method: "POST",
|
|
27
34
|
headers: {
|
|
28
|
-
|
|
29
|
-
|
|
35
|
+
"Content-Type": "application/json",
|
|
36
|
+
Authorization: `Token ${resolveApiKey(this.apiKey, "DEEPGRAM_API_KEY", "Deepgram")}`,
|
|
30
37
|
...options.headers,
|
|
31
38
|
},
|
|
32
39
|
body: JSON.stringify(body),
|
|
@@ -34,7 +41,7 @@ export class DeepgramSpeechProvider {
|
|
|
34
41
|
});
|
|
35
42
|
await handleErrorResponse(response, `deepgram/${options.modelId}`);
|
|
36
43
|
const arrayBuffer = await response.arrayBuffer();
|
|
37
|
-
const mediaType = response.headers.get(
|
|
44
|
+
const mediaType = response.headers.get("content-type") ?? "audio/mpeg";
|
|
38
45
|
return {
|
|
39
46
|
audio: new Uint8Array(arrayBuffer),
|
|
40
47
|
mediaType,
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/providers/deepgram/index.ts"],"names":[],"mappings":"
|
|
1
|
+
{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/providers/deepgram/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,mBAAmB,EAAE,aAAa,EAAE,MAAM,yBAAyB,CAAC;AAS7E,MAAM,OAAO,sBAAsB;IACxB,EAAE,GAAG,UAAU,CAAC;IAChB,YAAY,GAAG,QAAQ,CAAC;IAExB,MAAM,GAAG;QAChB;YACE,EAAE,EAAE,QAAQ;YACZ,SAAS,EAAE,KAAK;YAChB,SAAS,EAAE,CAAC,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,CAAC;YACrD,WAAW,EAAE,YAAY;YACzB,UAAU,EAAE,KAAK;YACjB,kBAAkB,EAAE,KAAK;SAC1B;KACO,CAAC;IAEM,MAAM,CAAqB;IAC3B,OAAO,CAAS;IAChB,OAAO,CAA0B;IAElD,YAAY,MAAoC;QAC9C,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC,OAAO,IAAI,6BAA6B,CAAC;QAC/D,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC,KAAK,IAAI,UAAU,CAAC,KAAK,CAAC;IAClD,CAAC;IAED,KAAK,CAAC,QAAQ,CAAC,OAOd;QAKC,MAAM,UAAU,GAAG,OAAO,CAAC,KAAK;YAC9B,CAAC,CAAC,GAAG,OAAO,CAAC,OAAO,IAAI,OAAO,CAAC,KAAK,EAAE;YACvC,CAAC,CAAC,OAAO,CAAC,OAAO,CAAC;QAEpB,MAAM,GAAG,GAAG,GAAG,IAAI,CAAC,OAAO,gBAAgB,kBAAkB,CAAC,UAAU,CAAC,EAAE,CAAC;QAE5E,MAAM,IAAI,GAA4B;YACpC,GAAG,OAAO,CAAC,eAAe;YAC1B,IAAI,EAAE,OAAO,CAAC,IAAI;SACnB,CAAC;QAEF,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,EAAE;YACvC,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,cAAc,EAAE,kBAAkB;gBAClC,aAAa,EAAE,SAAS,aAAa,CAAC,IAAI,CAAC,MAAM,EAAE,kBAAkB,EAAE,UAAU,CAAC,EAAE;gBACpF,GAAG,OAAO,CAAC,OAAO;aACnB;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC;YAC1B,MAAM,EAAE,OAAO,CAAC,WAAW;SAC5B,CAAC,CAAC;QAEH,MAAM,mBAAmB,CAAC,QAAQ,EAAE,YAAY,OAAO,CAAC,OAAO,EAAE,CAAC,CAAC;QAEnE,MAAM,WAAW,GAAG,MAAM,QAAQ,CAAC,WAAW,EAAE,CAAC;QACjD,MAAM,SAAS,GAAG,QAAQ,CAAC,OAAO,CAAC,GAAG,CAAC,cAAc,CAAC,IAAI,YAAY,CAAC;QAEvE,OAAO;YACL,KAAK,EAAE,IAAI,UAAU,CAAC,WAAW,CAAC;YAClC,SAAS;SACV,CAAC;IACJ,CAAC;CACF;AAED,MAAM,UAAU,cAAc,CAAC,SAAuC,EAAE;IACtE,MAAM,QAAQ,GAAG,IAAI,sBAAsB,CAAC,MAAM,CAAC,CAAC;IAEpD,OAAO,SAAS,QAAQ,CAAC,OAAgB;QACvC,OAAO;YACL,QAAQ;YACR,OAAO,EAAE,OAAO,IAAI,QAAQ,CAAC,YAAY;SAC1C,CAAC;IACJ,CAAC,CAAC;AACJ,CAAC"}
|
|
@@ -1,4 +1,4 @@
|
|
|
1
|
-
import type {
|
|
1
|
+
import type { ResolvedModel, SpeechProvider } from "../../speech-provider.js";
|
|
2
2
|
export interface ElevenLabsSpeechProviderConfig {
|
|
3
3
|
apiKey?: string;
|
|
4
4
|
baseURL?: string;
|
|
@@ -12,37 +12,41 @@ export declare class ElevenLabsSpeechProvider implements SpeechProvider<string,
|
|
|
12
12
|
private static readonly V3_LANGUAGES;
|
|
13
13
|
readonly models: readonly [{
|
|
14
14
|
readonly id: "eleven_v3";
|
|
15
|
+
readonly audioTags: true;
|
|
15
16
|
readonly languages: readonly ["af", "ar", "hy", "as", "az", "be", "bn", "bs", "bg", "ca", "ceb", "ny", "hr", "cs", "da", "nl", "en", "et", "fil", "fi", "fr", "gl", "ka", "de", "el", "gu", "ha", "he", "hi", "hu", "is", "id", "ga", "it", "ja", "jv", "kn", "kk", "ky", "ko", "lv", "ln", "lt", "lb", "mk", "ms", "ml", "zh", "mr", "ne", "no", "ps", "fa", "pl", "pt", "pa", "ro", "ru", "sr", "sd", "sk", "sl", "so", "es", "sw", "sv", "ta", "te", "th", "tr", "uk", "ur", "vi", "cy"];
|
|
16
17
|
readonly releaseDate: "2025-06-08";
|
|
17
18
|
readonly openSource: false;
|
|
18
19
|
readonly inlineVoiceCloning: false;
|
|
19
|
-
readonly zeroDataRetention: true;
|
|
20
20
|
}, {
|
|
21
21
|
readonly id: "eleven_multilingual_v2";
|
|
22
|
+
readonly audioTags: false;
|
|
22
23
|
readonly languages: readonly ["ar", "bg", "cs", "da", "de", "el", "en", "es", "fi", "fil", "fr", "he", "hi", "hr", "id", "it", "ja", "ko", "ms", "nl", "pl", "pt", "ro", "ru", "sk", "sv", "ta", "uk", "zh"];
|
|
23
24
|
readonly releaseDate: "2023-08-22";
|
|
24
25
|
readonly openSource: false;
|
|
25
26
|
readonly inlineVoiceCloning: false;
|
|
26
|
-
readonly zeroDataRetention: true;
|
|
27
27
|
}, {
|
|
28
28
|
readonly id: "eleven_flash_v2_5";
|
|
29
|
+
readonly audioTags: false;
|
|
29
30
|
readonly languages: readonly ["ar", "bg", "cs", "da", "de", "el", "en", "es", "fi", "fil", "fr", "he", "hi", "hr", "id", "it", "ja", "ko", "ms", "nl", "pl", "pt", "ro", "ru", "sk", "sv", "ta", "uk", "zh", "hu", "no", "vi"];
|
|
30
31
|
readonly releaseDate: "2024-12-01";
|
|
31
32
|
readonly openSource: false;
|
|
32
33
|
readonly inlineVoiceCloning: false;
|
|
33
|
-
readonly zeroDataRetention: true;
|
|
34
34
|
}, {
|
|
35
35
|
readonly id: "eleven_flash_v2";
|
|
36
|
+
readonly audioTags: false;
|
|
36
37
|
readonly languages: readonly ["en"];
|
|
37
38
|
readonly releaseDate: "2024-12-01";
|
|
38
39
|
readonly openSource: false;
|
|
39
40
|
readonly inlineVoiceCloning: false;
|
|
40
|
-
readonly zeroDataRetention: true;
|
|
41
41
|
}];
|
|
42
42
|
private readonly apiKey;
|
|
43
43
|
private readonly baseURL;
|
|
44
44
|
private readonly fetchFn;
|
|
45
45
|
constructor(config: ElevenLabsSpeechProviderConfig);
|
|
46
|
+
processAudioTags(text: string, modelId: string): {
|
|
47
|
+
text: string;
|
|
48
|
+
warnings: string[];
|
|
49
|
+
};
|
|
46
50
|
generate(options: {
|
|
47
51
|
modelId: string;
|
|
48
52
|
text: string;
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/providers/elevenlabs/index.ts"],"names":[],"mappings":"
|
|
1
|
+
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/providers/elevenlabs/index.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAE,aAAa,EAAE,cAAc,EAAE,MAAM,0BAA0B,CAAC;AAE9E,MAAM,WAAW,8BAA8B;IAC7C,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,KAAK,CAAC,EAAE,OAAO,UAAU,CAAC,KAAK,CAAC;CACjC;AAED,qBAAa,wBACX,YAAW,cAAc,CAAC,MAAM,EAAE,MAAM,CAAC;IAEzC,QAAQ,CAAC,EAAE,gBAAgB;IAC3B,QAAQ,CAAC,YAAY,4BAA4B;IAEjD,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,YAAY,CA8BzB;IAEX,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,oBAAoB,CAKjC;IAEX,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,YAAY,CA2EzB;IAEX,QAAQ,CAAC,MAAM;;;;;;;;;;;;;;;;;;;;;;;;;;;;OAiCJ;IAEX,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAqB;IAC5C,OAAO,CAAC,QAAQ,CAAC,OAAO,CAAS;IACjC,OAAO,CAAC,QAAQ,CAAC,OAAO,CAA0B;gBAEtC,MAAM,EAAE,8BAA8B;IAMlD,gBAAgB,CACd,IAAI,EAAE,MAAM,EACZ,OAAO,EAAE,MAAM,GACd;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,QAAQ,EAAE,MAAM,EAAE,CAAA;KAAE;IAOjC,QAAQ,CAAC,OAAO,EAAE;QACtB,OAAO,EAAE,MAAM,CAAC;QAChB,IAAI,EAAE,MAAM,CAAC;QACb,KAAK,CAAC,EAAE,MAAM,CAAC;QACf,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;QAC1C,WAAW,CAAC,EAAE,WAAW,CAAC;QAC1B,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;KAClC,GAAG,OAAO,CAAC;QACV,KAAK,EAAE,UAAU,CAAC;QAClB,SAAS,EAAE,MAAM,CAAC;QAClB,gBAAgB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;KAC5C,CAAC;CAoEH;AAED,wBAAgB,gBAAgB,CAAC,MAAM,GAAE,8BAAmC,IAG/C,UAAU,MAAM,KAAG,aAAa,CAAC,MAAM,CAAC,CAMpE"}
|
|
@@ -1,43 +1,173 @@
|
|
|
1
|
-
import {
|
|
2
|
-
import {
|
|
1
|
+
import { stripAudioTags } from "../../audio-tags.js";
|
|
2
|
+
import { SpeechSDKError } from "../../errors.js";
|
|
3
|
+
import { handleErrorResponse, resolveApiKey } from "../../provider-utils.js";
|
|
3
4
|
export class ElevenLabsSpeechProvider {
|
|
4
|
-
id =
|
|
5
|
-
defaultModel =
|
|
5
|
+
id = "elevenlabs";
|
|
6
|
+
defaultModel = "eleven_multilingual_v2";
|
|
6
7
|
static V2_LANGUAGES = [
|
|
7
|
-
|
|
8
|
-
|
|
9
|
-
|
|
8
|
+
"ar",
|
|
9
|
+
"bg",
|
|
10
|
+
"cs",
|
|
11
|
+
"da",
|
|
12
|
+
"de",
|
|
13
|
+
"el",
|
|
14
|
+
"en",
|
|
15
|
+
"es",
|
|
16
|
+
"fi",
|
|
17
|
+
"fil",
|
|
18
|
+
"fr",
|
|
19
|
+
"he",
|
|
20
|
+
"hi",
|
|
21
|
+
"hr",
|
|
22
|
+
"id",
|
|
23
|
+
"it",
|
|
24
|
+
"ja",
|
|
25
|
+
"ko",
|
|
26
|
+
"ms",
|
|
27
|
+
"nl",
|
|
28
|
+
"pl",
|
|
29
|
+
"pt",
|
|
30
|
+
"ro",
|
|
31
|
+
"ru",
|
|
32
|
+
"sk",
|
|
33
|
+
"sv",
|
|
34
|
+
"ta",
|
|
35
|
+
"uk",
|
|
36
|
+
"zh",
|
|
10
37
|
];
|
|
11
38
|
static FLASH_V2_5_LANGUAGES = [
|
|
12
|
-
...ElevenLabsSpeechProvider.V2_LANGUAGES,
|
|
39
|
+
...ElevenLabsSpeechProvider.V2_LANGUAGES,
|
|
40
|
+
"hu",
|
|
41
|
+
"no",
|
|
42
|
+
"vi",
|
|
13
43
|
];
|
|
14
44
|
static V3_LANGUAGES = [
|
|
15
|
-
|
|
16
|
-
|
|
17
|
-
|
|
18
|
-
|
|
19
|
-
|
|
20
|
-
|
|
21
|
-
|
|
22
|
-
|
|
45
|
+
"af",
|
|
46
|
+
"ar",
|
|
47
|
+
"hy",
|
|
48
|
+
"as",
|
|
49
|
+
"az",
|
|
50
|
+
"be",
|
|
51
|
+
"bn",
|
|
52
|
+
"bs",
|
|
53
|
+
"bg",
|
|
54
|
+
"ca",
|
|
55
|
+
"ceb",
|
|
56
|
+
"ny",
|
|
57
|
+
"hr",
|
|
58
|
+
"cs",
|
|
59
|
+
"da",
|
|
60
|
+
"nl",
|
|
61
|
+
"en",
|
|
62
|
+
"et",
|
|
63
|
+
"fil",
|
|
64
|
+
"fi",
|
|
65
|
+
"fr",
|
|
66
|
+
"gl",
|
|
67
|
+
"ka",
|
|
68
|
+
"de",
|
|
69
|
+
"el",
|
|
70
|
+
"gu",
|
|
71
|
+
"ha",
|
|
72
|
+
"he",
|
|
73
|
+
"hi",
|
|
74
|
+
"hu",
|
|
75
|
+
"is",
|
|
76
|
+
"id",
|
|
77
|
+
"ga",
|
|
78
|
+
"it",
|
|
79
|
+
"ja",
|
|
80
|
+
"jv",
|
|
81
|
+
"kn",
|
|
82
|
+
"kk",
|
|
83
|
+
"ky",
|
|
84
|
+
"ko",
|
|
85
|
+
"lv",
|
|
86
|
+
"ln",
|
|
87
|
+
"lt",
|
|
88
|
+
"lb",
|
|
89
|
+
"mk",
|
|
90
|
+
"ms",
|
|
91
|
+
"ml",
|
|
92
|
+
"zh",
|
|
93
|
+
"mr",
|
|
94
|
+
"ne",
|
|
95
|
+
"no",
|
|
96
|
+
"ps",
|
|
97
|
+
"fa",
|
|
98
|
+
"pl",
|
|
99
|
+
"pt",
|
|
100
|
+
"pa",
|
|
101
|
+
"ro",
|
|
102
|
+
"ru",
|
|
103
|
+
"sr",
|
|
104
|
+
"sd",
|
|
105
|
+
"sk",
|
|
106
|
+
"sl",
|
|
107
|
+
"so",
|
|
108
|
+
"es",
|
|
109
|
+
"sw",
|
|
110
|
+
"sv",
|
|
111
|
+
"ta",
|
|
112
|
+
"te",
|
|
113
|
+
"th",
|
|
114
|
+
"tr",
|
|
115
|
+
"uk",
|
|
116
|
+
"ur",
|
|
117
|
+
"vi",
|
|
118
|
+
"cy",
|
|
23
119
|
];
|
|
24
120
|
models = [
|
|
25
|
-
{
|
|
26
|
-
|
|
27
|
-
|
|
28
|
-
|
|
121
|
+
{
|
|
122
|
+
id: "eleven_v3",
|
|
123
|
+
audioTags: true,
|
|
124
|
+
languages: ElevenLabsSpeechProvider.V3_LANGUAGES,
|
|
125
|
+
releaseDate: "2025-06-08",
|
|
126
|
+
openSource: false,
|
|
127
|
+
inlineVoiceCloning: false,
|
|
128
|
+
},
|
|
129
|
+
{
|
|
130
|
+
id: "eleven_multilingual_v2",
|
|
131
|
+
audioTags: false,
|
|
132
|
+
languages: ElevenLabsSpeechProvider.V2_LANGUAGES,
|
|
133
|
+
releaseDate: "2023-08-22",
|
|
134
|
+
openSource: false,
|
|
135
|
+
inlineVoiceCloning: false,
|
|
136
|
+
},
|
|
137
|
+
{
|
|
138
|
+
id: "eleven_flash_v2_5",
|
|
139
|
+
audioTags: false,
|
|
140
|
+
languages: ElevenLabsSpeechProvider.FLASH_V2_5_LANGUAGES,
|
|
141
|
+
releaseDate: "2024-12-01",
|
|
142
|
+
openSource: false,
|
|
143
|
+
inlineVoiceCloning: false,
|
|
144
|
+
},
|
|
145
|
+
{
|
|
146
|
+
id: "eleven_flash_v2",
|
|
147
|
+
audioTags: false,
|
|
148
|
+
languages: ["en"],
|
|
149
|
+
releaseDate: "2024-12-01",
|
|
150
|
+
openSource: false,
|
|
151
|
+
inlineVoiceCloning: false,
|
|
152
|
+
},
|
|
29
153
|
];
|
|
30
154
|
apiKey;
|
|
31
155
|
baseURL;
|
|
32
156
|
fetchFn;
|
|
33
157
|
constructor(config) {
|
|
34
158
|
this.apiKey = config.apiKey;
|
|
35
|
-
this.baseURL = config.baseURL ??
|
|
159
|
+
this.baseURL = config.baseURL ?? "https://api.elevenlabs.io";
|
|
36
160
|
this.fetchFn = config.fetch ?? globalThis.fetch;
|
|
37
161
|
}
|
|
162
|
+
processAudioTags(text, modelId) {
|
|
163
|
+
if (this.models.some((m) => m.id === modelId && m.audioTags)) {
|
|
164
|
+
return { text, warnings: [] };
|
|
165
|
+
}
|
|
166
|
+
return stripAudioTags(text, `elevenlabs/${modelId}`);
|
|
167
|
+
}
|
|
38
168
|
async generate(options) {
|
|
39
169
|
if (!options.voice) {
|
|
40
|
-
throw new SpeechSDKError(
|
|
170
|
+
throw new SpeechSDKError("ElevenLabs requires a voice ID. Pass it via the voice option.");
|
|
41
171
|
}
|
|
42
172
|
const providerOptions = options.providerOptions ?? {};
|
|
43
173
|
const { output_format, enable_logging, optimize_streaming_latency, ...bodyOptions } = providerOptions;
|
|
@@ -47,22 +177,25 @@ export class ElevenLabsSpeechProvider {
|
|
|
47
177
|
model_id: options.modelId,
|
|
48
178
|
};
|
|
49
179
|
const queryParams = new URLSearchParams();
|
|
50
|
-
if (output_format != null)
|
|
51
|
-
queryParams.set(
|
|
52
|
-
|
|
53
|
-
|
|
54
|
-
|
|
55
|
-
|
|
180
|
+
if (output_format != null) {
|
|
181
|
+
queryParams.set("output_format", String(output_format));
|
|
182
|
+
}
|
|
183
|
+
if (enable_logging != null) {
|
|
184
|
+
queryParams.set("enable_logging", String(enable_logging));
|
|
185
|
+
}
|
|
186
|
+
if (optimize_streaming_latency != null) {
|
|
187
|
+
queryParams.set("optimize_streaming_latency", String(optimize_streaming_latency));
|
|
188
|
+
}
|
|
56
189
|
let url = `${this.baseURL}/v1/text-to-speech/${options.voice}`;
|
|
57
190
|
const queryString = queryParams.toString();
|
|
58
191
|
if (queryString) {
|
|
59
192
|
url += `?${queryString}`;
|
|
60
193
|
}
|
|
61
194
|
const response = await this.fetchFn(url, {
|
|
62
|
-
method:
|
|
195
|
+
method: "POST",
|
|
63
196
|
headers: {
|
|
64
|
-
|
|
65
|
-
|
|
197
|
+
"Content-Type": "application/json",
|
|
198
|
+
"xi-api-key": resolveApiKey(this.apiKey, "ELEVENLABS_API_KEY", "ElevenLabs"),
|
|
66
199
|
...options.headers,
|
|
67
200
|
},
|
|
68
201
|
body: JSON.stringify(body),
|
|
@@ -70,8 +203,8 @@ export class ElevenLabsSpeechProvider {
|
|
|
70
203
|
});
|
|
71
204
|
await handleErrorResponse(response, `elevenlabs/${options.modelId}`);
|
|
72
205
|
const arrayBuffer = await response.arrayBuffer();
|
|
73
|
-
const mediaType = response.headers.get(
|
|
74
|
-
const requestId = response.headers.get(
|
|
206
|
+
const mediaType = response.headers.get("content-type") ?? "audio/mpeg";
|
|
207
|
+
const requestId = response.headers.get("request-id");
|
|
75
208
|
return {
|
|
76
209
|
audio: new Uint8Array(arrayBuffer),
|
|
77
210
|
mediaType,
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/providers/elevenlabs/index.ts"],"names":[],"mappings":"
|
|
1
|
+
{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/providers/elevenlabs/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,cAAc,EAAE,MAAM,qBAAqB,CAAC;AACrD,OAAO,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AACjD,OAAO,EAAE,mBAAmB,EAAE,aAAa,EAAE,MAAM,yBAAyB,CAAC;AAS7E,MAAM,OAAO,wBAAwB;IAG1B,EAAE,GAAG,YAAY,CAAC;IAClB,YAAY,GAAG,wBAAwB,CAAC;IAEzC,MAAM,CAAU,YAAY,GAAG;QACrC,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,KAAK;QACL,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;KACI,CAAC;IAEH,MAAM,CAAU,oBAAoB,GAAG;QAC7C,GAAG,wBAAwB,CAAC,YAAY;QACxC,IAAI;QACJ,IAAI;QACJ,IAAI;KACI,CAAC;IAEH,MAAM,CAAU,YAAY,GAAG;QACrC,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,KAAK;QACL,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,KAAK;QACL,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;QACJ,IAAI;KACI,CAAC;IAEF,MAAM,GAAG;QAChB;YACE,EAAE,EAAE,WAAW;YACf,SAAS,EAAE,IAAI;YACf,SAAS,EAAE,wBAAwB,CAAC,YAAY;YAChD,WAAW,EAAE,YAAY;YACzB,UAAU,EAAE,KAAK;YACjB,kBAAkB,EAAE,KAAK;SAC1B;QACD;YACE,EAAE,EAAE,wBAAwB;YAC5B,SAAS,EAAE,KAAK;YAChB,SAAS,EAAE,wBAAwB,CAAC,YAAY;YAChD,WAAW,EAAE,YAAY;YACzB,UAAU,EAAE,KAAK;YACjB,kBAAkB,EAAE,KAAK;SAC1B;QACD;YACE,EAAE,EAAE,mBAAmB;YACvB,SAAS,EAAE,KAAK;YAChB,SAAS,EAAE,wBAAwB,CAAC,oBAAoB;YACxD,WAAW,EAAE,YAAY;YACzB,UAAU,EAAE,KAAK;YACjB,kBAAkB,EAAE,KAAK;SAC1B;QACD;YACE,EAAE,EAAE,iBAAiB;YACrB,SAAS,EAAE,KAAK;YAChB,SAAS,EAAE,CAAC,IAAI,CAAU;YAC1B,WAAW,EAAE,YAAY;YACzB,UAAU,EAAE,KAAK;YACjB,kBAAkB,EAAE,KAAK;SAC1B;KACO,CAAC;IAEM,MAAM,CAAqB;IAC3B,OAAO,CAAS;IAChB,OAAO,CAA0B;IAElD,YAAY,MAAsC;QAChD,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC,OAAO,IAAI,2BAA2B,CAAC;QAC7D,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC,KAAK,IAAI,UAAU,CAAC,KAAK,CAAC;IAClD,CAAC;IAED,gBAAgB,CACd,IAAY,EACZ,OAAe;QAEf,IAAI,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,OAAO,IAAI,CAAC,CAAC,SAAS,CAAC,EAAE,CAAC;YAC7D,OAAO,EAAE,IAAI,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC;QAChC,CAAC;QACD,OAAO,cAAc,CAAC,IAAI,EAAE,cAAc,OAAO,EAAE,CAAC,CAAC;IACvD,CAAC;IAED,KAAK,CAAC,QAAQ,CAAC,OAOd;QAKC,IAAI,CAAC,OAAO,CAAC,KAAK,EAAE,CAAC;YACnB,MAAM,IAAI,cAAc,CACtB,+DAA+D,CAChE,CAAC;QACJ,CAAC;QAED,MAAM,eAAe,GAAG,OAAO,CAAC,eAAe,IAAI,EAAE,CAAC;QACtD,MAAM,EACJ,aAAa,EACb,cAAc,EACd,0BAA0B,EAC1B,GAAG,WAAW,EACf,GAAG,eAA0C,CAAC;QAE/C,MAAM,IAAI,GAA4B;YACpC,GAAG,WAAW;YACd,IAAI,EAAE,OAAO,CAAC,IAAI;YAClB,QAAQ,EAAE,OAAO,CAAC,OAAO;SAC1B,CAAC;QAEF,MAAM,WAAW,GAAG,IAAI,eAAe,EAAE,CAAC;QAC1C,IAAI,aAAa,IAAI,IAAI,EAAE,CAAC;YAC1B,WAAW,CAAC,GAAG,CAAC,eAAe,EAAE,MAAM,CAAC,aAAa,CAAC,CAAC,CAAC;QAC1D,CAAC;QACD,IAAI,cAAc,IAAI,IAAI,EAAE,CAAC;YAC3B,WAAW,CAAC,GAAG,CAAC,gBAAgB,EAAE,MAAM,CAAC,cAAc,CAAC,CAAC,CAAC;QAC5D,CAAC;QACD,IAAI,0BAA0B,IAAI,IAAI,EAAE,CAAC;YACvC,WAAW,CAAC,GAAG,CACb,4BAA4B,EAC5B,MAAM,CAAC,0BAA0B,CAAC,CACnC,CAAC;QACJ,CAAC;QAED,IAAI,GAAG,GAAG,GAAG,IAAI,CAAC,OAAO,sBAAsB,OAAO,CAAC,KAAK,EAAE,CAAC;QAC/D,MAAM,WAAW,GAAG,WAAW,CAAC,QAAQ,EAAE,CAAC;QAC3C,IAAI,WAAW,EAAE,CAAC;YAChB,GAAG,IAAI,IAAI,WAAW,EAAE,CAAC;QAC3B,CAAC;QAED,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,EAAE;YACvC,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,cAAc,EAAE,kBAAkB;gBAClC,YAAY,EAAE,aAAa,CACzB,IAAI,CAAC,MAAM,EACX,oBAAoB,EACpB,YAAY,CACb;gBACD,GAAG,OAAO,CAAC,OAAO;aACnB;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC;YAC1B,MAAM,EAAE,OAAO,CAAC,WAAW;SAC5B,CAAC,CAAC;QAEH,MAAM,mBAAmB,CAAC,QAAQ,EAAE,cAAc,OAAO,CAAC,OAAO,EAAE,CAAC,CAAC;QAErE,MAAM,WAAW,GAAG,MAAM,QAAQ,CAAC,WAAW,EAAE,CAAC;QACjD,MAAM,SAAS,GAAG,QAAQ,CAAC,OAAO,CAAC,GAAG,CAAC,cAAc,CAAC,IAAI,YAAY,CAAC;QACvE,MAAM,SAAS,GAAG,QAAQ,CAAC,OAAO,CAAC,GAAG,CAAC,YAAY,CAAC,CAAC;QAErD,OAAO;YACL,KAAK,EAAE,IAAI,UAAU,CAAC,WAAW,CAAC;YAClC,SAAS;YACT,gBAAgB,EAAE,SAAS,CAAC,CAAC,CAAC,EAAE,SAAS,EAAE,CAAC,CAAC,CAAC,SAAS;SACxD,CAAC;IACJ,CAAC;;AAGH,MAAM,UAAU,gBAAgB,CAAC,SAAyC,EAAE;IAC1E,MAAM,QAAQ,GAAG,IAAI,wBAAwB,CAAC,MAAM,CAAC,CAAC;IAEtD,OAAO,SAAS,UAAU,CAAC,OAAgB;QACzC,OAAO;YACL,QAAQ;YACR,OAAO,EAAE,OAAO,IAAI,QAAQ,CAAC,YAAY;SAC1C,CAAC;IACJ,CAAC,CAAC;AACJ,CAAC"}
|
|
@@ -1,4 +1,4 @@
|
|
|
1
|
-
import type {
|
|
1
|
+
import type { ResolvedModel, SpeechProvider } from "../../speech-provider.js";
|
|
2
2
|
export interface FalSpeechProviderConfig {
|
|
3
3
|
apiKey?: string;
|
|
4
4
|
baseURL?: string;
|
|
@@ -11,39 +11,39 @@ export declare class FalSpeechProvider implements SpeechProvider<string, string
|
|
|
11
11
|
readonly defaultModel = "";
|
|
12
12
|
readonly models: readonly [{
|
|
13
13
|
readonly id: "f5-tts";
|
|
14
|
+
readonly audioTags: false;
|
|
14
15
|
readonly languages: readonly ["en", "zh", "fr", "it", "hi", "ja", "ru", "es", "fi"];
|
|
15
16
|
readonly releaseDate: "2024-10-08";
|
|
16
17
|
readonly openSource: true;
|
|
17
18
|
readonly inlineVoiceCloning: true;
|
|
18
|
-
readonly zeroDataRetention: false;
|
|
19
19
|
}, {
|
|
20
20
|
readonly id: "kokoro";
|
|
21
|
+
readonly audioTags: false;
|
|
21
22
|
readonly languages: readonly ["en", "fr", "ko", "ja", "zh"];
|
|
22
23
|
readonly releaseDate: "2025-01-27";
|
|
23
24
|
readonly openSource: true;
|
|
24
25
|
readonly inlineVoiceCloning: false;
|
|
25
|
-
readonly zeroDataRetention: false;
|
|
26
26
|
}, {
|
|
27
27
|
readonly id: "dia-tts";
|
|
28
|
+
readonly audioTags: false;
|
|
28
29
|
readonly languages: readonly ["en"];
|
|
29
30
|
readonly releaseDate: "2025-04-21";
|
|
30
31
|
readonly openSource: true;
|
|
31
32
|
readonly inlineVoiceCloning: true;
|
|
32
|
-
readonly zeroDataRetention: false;
|
|
33
33
|
}, {
|
|
34
34
|
readonly id: "orpheus-tts";
|
|
35
|
+
readonly audioTags: false;
|
|
35
36
|
readonly languages: readonly ["en", "es", "fr", "de", "it", "pt", "zh"];
|
|
36
37
|
readonly releaseDate: "2025-03-18";
|
|
37
38
|
readonly openSource: true;
|
|
38
39
|
readonly inlineVoiceCloning: false;
|
|
39
|
-
readonly zeroDataRetention: false;
|
|
40
40
|
}, {
|
|
41
41
|
readonly id: "index-tts-2";
|
|
42
|
+
readonly audioTags: false;
|
|
42
43
|
readonly languages: readonly ["en", "zh"];
|
|
43
44
|
readonly releaseDate: "2025-09-08";
|
|
44
45
|
readonly openSource: true;
|
|
45
46
|
readonly inlineVoiceCloning: true;
|
|
46
|
-
readonly zeroDataRetention: false;
|
|
47
47
|
}];
|
|
48
48
|
private readonly apiKey;
|
|
49
49
|
private readonly baseURL;
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/providers/fal/index.ts"],"names":[],"mappings":"
|
|
1
|
+
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/providers/fal/index.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,EAAE,aAAa,EAAE,cAAc,EAAE,MAAM,0BAA0B,CAAC;AAE9E,MAAM,WAAW,uBAAuB;IACtC,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,KAAK,CAAC,EAAE,OAAO,UAAU,CAAC,KAAK,CAAC;CACjC;AAED,qBAAa,iBACX,YAAW,cAAc,CAAC,MAAM,EAAE,MAAM,GAAG;IAAE,GAAG,EAAE,MAAM,CAAA;CAAE,CAAC;IAE3D,QAAQ,CAAC,EAAE,YAAY;IACvB,QAAQ,CAAC,YAAY,MAAM;IAE3B,QAAQ,CAAC,MAAM;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;OAyCJ;IAEX,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAqB;IAC5C,OAAO,CAAC,QAAQ,CAAC,OAAO,CAAS;IACjC,OAAO,CAAC,QAAQ,CAAC,OAAO,CAA0B;gBAEtC,MAAM,EAAE,uBAAuB;IAMrC,QAAQ,CAAC,OAAO,EAAE;QACtB,OAAO,EAAE,MAAM,CAAC;QAChB,IAAI,EAAE,MAAM,CAAC;QACb,KAAK,CAAC,EAAE,MAAM,GAAG;YAAE,GAAG,EAAE,MAAM,CAAA;SAAE,CAAC;QACjC,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;QAC1C,WAAW,CAAC,EAAE,WAAW,CAAC;QAC1B,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;KAClC,GAAG,OAAO,CAAC;QACV,KAAK,EAAE,UAAU,CAAC;QAClB,SAAS,EAAE,MAAM,CAAC;QAClB,gBAAgB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;KAC5C,CAAC;CAwDH;AAED,wBAAgB,SAAS,CAAC,MAAM,GAAE,uBAA4B,IAI1D,UAAU,MAAM,KACf,aAAa,CAAC,MAAM,GAAG;IAAE,GAAG,EAAE,MAAM,CAAA;CAAE,CAAC,CAM3C"}
|
|
@@ -1,21 +1,56 @@
|
|
|
1
|
-
import {
|
|
2
|
-
import {
|
|
1
|
+
import { ApiError } from "../../errors.js";
|
|
2
|
+
import { handleErrorResponse, resolveApiKey } from "../../provider-utils.js";
|
|
3
3
|
export class FalSpeechProvider {
|
|
4
|
-
id =
|
|
5
|
-
defaultModel =
|
|
4
|
+
id = "fal-ai";
|
|
5
|
+
defaultModel = "";
|
|
6
6
|
models = [
|
|
7
|
-
{
|
|
8
|
-
|
|
9
|
-
|
|
10
|
-
|
|
11
|
-
|
|
7
|
+
{
|
|
8
|
+
id: "f5-tts",
|
|
9
|
+
audioTags: false,
|
|
10
|
+
languages: ["en", "zh", "fr", "it", "hi", "ja", "ru", "es", "fi"],
|
|
11
|
+
releaseDate: "2024-10-08",
|
|
12
|
+
openSource: true,
|
|
13
|
+
inlineVoiceCloning: true,
|
|
14
|
+
},
|
|
15
|
+
{
|
|
16
|
+
id: "kokoro",
|
|
17
|
+
audioTags: false,
|
|
18
|
+
languages: ["en", "fr", "ko", "ja", "zh"],
|
|
19
|
+
releaseDate: "2025-01-27",
|
|
20
|
+
openSource: true,
|
|
21
|
+
inlineVoiceCloning: false,
|
|
22
|
+
},
|
|
23
|
+
{
|
|
24
|
+
id: "dia-tts",
|
|
25
|
+
audioTags: false,
|
|
26
|
+
languages: ["en"],
|
|
27
|
+
releaseDate: "2025-04-21",
|
|
28
|
+
openSource: true,
|
|
29
|
+
inlineVoiceCloning: true,
|
|
30
|
+
},
|
|
31
|
+
{
|
|
32
|
+
id: "orpheus-tts",
|
|
33
|
+
audioTags: false,
|
|
34
|
+
languages: ["en", "es", "fr", "de", "it", "pt", "zh"],
|
|
35
|
+
releaseDate: "2025-03-18",
|
|
36
|
+
openSource: true,
|
|
37
|
+
inlineVoiceCloning: false,
|
|
38
|
+
},
|
|
39
|
+
{
|
|
40
|
+
id: "index-tts-2",
|
|
41
|
+
audioTags: false,
|
|
42
|
+
languages: ["en", "zh"],
|
|
43
|
+
releaseDate: "2025-09-08",
|
|
44
|
+
openSource: true,
|
|
45
|
+
inlineVoiceCloning: true,
|
|
46
|
+
},
|
|
12
47
|
];
|
|
13
48
|
apiKey;
|
|
14
49
|
baseURL;
|
|
15
50
|
fetchFn;
|
|
16
51
|
constructor(config) {
|
|
17
52
|
this.apiKey = config.apiKey;
|
|
18
|
-
this.baseURL = config.baseURL ??
|
|
53
|
+
this.baseURL = config.baseURL ?? "https://fal.run";
|
|
19
54
|
this.fetchFn = config.fetch ?? globalThis.fetch;
|
|
20
55
|
}
|
|
21
56
|
async generate(options) {
|
|
@@ -28,25 +63,25 @@ export class FalSpeechProvider {
|
|
|
28
63
|
text: options.text,
|
|
29
64
|
};
|
|
30
65
|
if (options.voice != null) {
|
|
31
|
-
if (typeof options.voice ===
|
|
66
|
+
if (typeof options.voice === "string") {
|
|
32
67
|
body.voice = options.voice;
|
|
33
68
|
}
|
|
34
|
-
else if (
|
|
69
|
+
else if ("url" in options.voice) {
|
|
35
70
|
body.audio_url = options.voice.url;
|
|
36
71
|
}
|
|
37
72
|
}
|
|
38
73
|
const response = await this.fetchFn(url, {
|
|
39
|
-
method:
|
|
74
|
+
method: "POST",
|
|
40
75
|
headers: {
|
|
41
|
-
|
|
42
|
-
|
|
76
|
+
"Content-Type": "application/json",
|
|
77
|
+
Authorization: `Key ${resolveApiKey(this.apiKey, "FAL_API_KEY", "fal")}`,
|
|
43
78
|
...options.headers,
|
|
44
79
|
},
|
|
45
80
|
body: JSON.stringify(body),
|
|
46
81
|
signal: options.abortSignal,
|
|
47
82
|
});
|
|
48
83
|
await handleErrorResponse(response, `fal-ai/${options.modelId}`);
|
|
49
|
-
const json = await response.json();
|
|
84
|
+
const json = (await response.json());
|
|
50
85
|
const audioResponse = await this.fetchFn(json.audio.url, {
|
|
51
86
|
signal: options.abortSignal,
|
|
52
87
|
});
|
|
@@ -60,7 +95,7 @@ export class FalSpeechProvider {
|
|
|
60
95
|
const arrayBuffer = await audioResponse.arrayBuffer();
|
|
61
96
|
return {
|
|
62
97
|
audio: new Uint8Array(arrayBuffer),
|
|
63
|
-
mediaType:
|
|
98
|
+
mediaType: "audio/mpeg",
|
|
64
99
|
};
|
|
65
100
|
}
|
|
66
101
|
}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/providers/fal/index.ts"],"names":[],"mappings":"
|
|
1
|
+
{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/providers/fal/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,QAAQ,EAAE,MAAM,iBAAiB,CAAC;AAC3C,OAAO,EAAE,mBAAmB,EAAE,aAAa,EAAE,MAAM,yBAAyB,CAAC;AAS7E,MAAM,OAAO,iBAAiB;IAGnB,EAAE,GAAG,QAAQ,CAAC;IACd,YAAY,GAAG,EAAE,CAAC;IAElB,MAAM,GAAG;QAChB;YACE,EAAE,EAAE,QAAQ;YACZ,SAAS,EAAE,KAAK;YAChB,SAAS,EAAE,CAAC,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,CAAC;YACjE,WAAW,EAAE,YAAY;YACzB,UAAU,EAAE,IAAI;YAChB,kBAAkB,EAAE,IAAI;SACzB;QACD;YACE,EAAE,EAAE,QAAQ;YACZ,SAAS,EAAE,KAAK;YAChB,SAAS,EAAE,CAAC,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,CAAC;YACzC,WAAW,EAAE,YAAY;YACzB,UAAU,EAAE,IAAI;YAChB,kBAAkB,EAAE,KAAK;SAC1B;QACD;YACE,EAAE,EAAE,SAAS;YACb,SAAS,EAAE,KAAK;YAChB,SAAS,EAAE,CAAC,IAAI,CAAC;YACjB,WAAW,EAAE,YAAY;YACzB,UAAU,EAAE,IAAI;YAChB,kBAAkB,EAAE,IAAI;SACzB;QACD;YACE,EAAE,EAAE,aAAa;YACjB,SAAS,EAAE,KAAK;YAChB,SAAS,EAAE,CAAC,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,CAAC;YACrD,WAAW,EAAE,YAAY;YACzB,UAAU,EAAE,IAAI;YAChB,kBAAkB,EAAE,KAAK;SAC1B;QACD;YACE,EAAE,EAAE,aAAa;YACjB,SAAS,EAAE,KAAK;YAChB,SAAS,EAAE,CAAC,IAAI,EAAE,IAAI,CAAC;YACvB,WAAW,EAAE,YAAY;YACzB,UAAU,EAAE,IAAI;YAChB,kBAAkB,EAAE,IAAI;SACzB;KACO,CAAC;IAEM,MAAM,CAAqB;IAC3B,OAAO,CAAS;IAChB,OAAO,CAA0B;IAElD,YAAY,MAA+B;QACzC,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC,OAAO,IAAI,iBAAiB,CAAC;QACnD,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC,KAAK,IAAI,UAAU,CAAC,KAAK,CAAC;IAClD,CAAC;IAED,KAAK,CAAC,QAAQ,CAAC,OAOd;QAKC,IAAI,CAAC,OAAO,CAAC,OAAO,EAAE,CAAC;YACrB,MAAM,IAAI,KAAK,CACb,yFAAyF,CAC1F,CAAC;QACJ,CAAC;QAED,MAAM,GAAG,GAAG,GAAG,IAAI,CAAC,OAAO,WAAW,OAAO,CAAC,OAAO,EAAE,CAAC;QAExD,MAAM,IAAI,GAA4B;YACpC,GAAG,OAAO,CAAC,eAAe;YAC1B,IAAI,EAAE,OAAO,CAAC,IAAI;SACnB,CAAC;QAEF,IAAI,OAAO,CAAC,KAAK,IAAI,IAAI,EAAE,CAAC;YAC1B,IAAI,OAAO,OAAO,CAAC,KAAK,KAAK,QAAQ,EAAE,CAAC;gBACtC,IAAI,CAAC,KAAK,GAAG,OAAO,CAAC,KAAK,CAAC;YAC7B,CAAC;iBAAM,IAAI,KAAK,IAAI,OAAO,CAAC,KAAK,EAAE,CAAC;gBAClC,IAAI,CAAC,SAAS,GAAG,OAAO,CAAC,KAAK,CAAC,GAAG,CAAC;YACrC,CAAC;QACH,CAAC;QAED,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,EAAE;YACvC,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,cAAc,EAAE,kBAAkB;gBAClC,aAAa,EAAE,OAAO,aAAa,CAAC,IAAI,CAAC,MAAM,EAAE,aAAa,EAAE,KAAK,CAAC,EAAE;gBACxE,GAAG,OAAO,CAAC,OAAO;aACnB;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC;YAC1B,MAAM,EAAE,OAAO,CAAC,WAAW;SAC5B,CAAC,CAAC;QAEH,MAAM,mBAAmB,CAAC,QAAQ,EAAE,UAAU,OAAO,CAAC,OAAO,EAAE,CAAC,CAAC;QAEjE,MAAM,IAAI,GAAG,CAAC,MAAM,QAAQ,CAAC,IAAI,EAAE,CAA+B,CAAC;QAEnE,MAAM,aAAa,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,IAAI,CAAC,KAAK,CAAC,GAAG,EAAE;YACvD,MAAM,EAAE,OAAO,CAAC,WAAW;SAC5B,CAAC,CAAC;QAEH,IAAI,CAAC,aAAa,CAAC,EAAE,EAAE,CAAC;YACtB,MAAM,IAAI,QAAQ,CAAC,cAAc,aAAa,CAAC,MAAM,EAAE,EAAE;gBACvD,UAAU,EAAE,aAAa,CAAC,MAAM;gBAChC,KAAK,EAAE,UAAU,OAAO,CAAC,OAAO,EAAE;gBAClC,YAAY,EAAE,MAAM,aAAa,CAAC,IAAI,EAAE,CAAC,KAAK,CAAC,GAAG,EAAE,CAAC,SAAS,CAAC;aAChE,CAAC,CAAC;QACL,CAAC;QAED,MAAM,WAAW,GAAG,MAAM,aAAa,CAAC,WAAW,EAAE,CAAC;QAEtD,OAAO;YACL,KAAK,EAAE,IAAI,UAAU,CAAC,WAAW,CAAC;YAClC,SAAS,EAAE,YAAY;SACxB,CAAC;IACJ,CAAC;CACF;AAED,MAAM,UAAU,SAAS,CAAC,SAAkC,EAAE;IAC5D,MAAM,QAAQ,GAAG,IAAI,iBAAiB,CAAC,MAAM,CAAC,CAAC;IAE/C,OAAO,SAAS,GAAG,CACjB,OAAgB;QAEhB,OAAO;YACL,QAAQ;YACR,OAAO,EAAE,OAAO,IAAI,QAAQ,CAAC,YAAY;SAC1C,CAAC;IACJ,CAAC,CAAC;AACJ,CAAC"}
|
|
@@ -1,4 +1,4 @@
|
|
|
1
|
-
import type {
|
|
1
|
+
import type { ResolvedModel, SpeechProvider } from "../../speech-provider.js";
|
|
2
2
|
export interface FishAudioSpeechProviderConfig {
|
|
3
3
|
apiKey?: string;
|
|
4
4
|
baseURL?: string;
|
|
@@ -9,16 +9,20 @@ export declare class FishAudioSpeechProvider implements SpeechProvider<string, s
|
|
|
9
9
|
readonly defaultModel = "s2-pro";
|
|
10
10
|
readonly models: readonly [{
|
|
11
11
|
readonly id: "s2-pro";
|
|
12
|
+
readonly audioTags: true;
|
|
12
13
|
readonly languages: readonly ["ja", "en", "zh", "ko", "es", "pt", "ar", "ru", "fr", "de"];
|
|
13
14
|
readonly releaseDate: "2026-03-09";
|
|
14
15
|
readonly openSource: true;
|
|
15
16
|
readonly inlineVoiceCloning: true;
|
|
16
|
-
readonly zeroDataRetention: false;
|
|
17
17
|
}];
|
|
18
18
|
private readonly apiKey;
|
|
19
19
|
private readonly baseURL;
|
|
20
20
|
private readonly fetchFn;
|
|
21
21
|
constructor(config: FishAudioSpeechProviderConfig);
|
|
22
|
+
processAudioTags(text: string, modelId: string): {
|
|
23
|
+
text: string;
|
|
24
|
+
warnings: string[];
|
|
25
|
+
};
|
|
22
26
|
generate(options: {
|
|
23
27
|
modelId: string;
|
|
24
28
|
text: string;
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/providers/fish-audio/index.ts"],"names":[],"mappings":"
|
|
1
|
+
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/providers/fish-audio/index.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,EAAE,aAAa,EAAE,cAAc,EAAE,MAAM,0BAA0B,CAAC;AAE9E,MAAM,WAAW,6BAA6B;IAC5C,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,KAAK,CAAC,EAAE,OAAO,UAAU,CAAC,KAAK,CAAC;CACjC;AAED,qBAAa,uBAAwB,YAAW,cAAc,CAAC,MAAM,EAAE,MAAM,CAAC;IAC5E,QAAQ,CAAC,EAAE,gBAAgB;IAC3B,QAAQ,CAAC,YAAY,YAAY;IAEjC,QAAQ,CAAC,MAAM;;;;;;;OASJ;IAEX,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAqB;IAC5C,OAAO,CAAC,QAAQ,CAAC,OAAO,CAAS;IACjC,OAAO,CAAC,QAAQ,CAAC,OAAO,CAA0B;gBAEtC,MAAM,EAAE,6BAA6B;IAMjD,gBAAgB,CACd,IAAI,EAAE,MAAM,EACZ,OAAO,EAAE,MAAM,GACd;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,QAAQ,EAAE,MAAM,EAAE,CAAA;KAAE;IAOjC,QAAQ,CAAC,OAAO,EAAE;QACtB,OAAO,EAAE,MAAM,CAAC;QAChB,IAAI,EAAE,MAAM,CAAC;QACb,KAAK,CAAC,EAAE,MAAM,CAAC;QACf,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;QAC1C,WAAW,CAAC,EAAE,WAAW,CAAC;QAC1B,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;KAClC,GAAG,OAAO,CAAC;QACV,KAAK,EAAE,UAAU,CAAC;QAClB,SAAS,EAAE,MAAM,CAAC;QAClB,gBAAgB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;KAC5C,CAAC;CAkCH;AAED,wBAAgB,eAAe,CAAC,MAAM,GAAE,6BAAkC,IAG9C,UAAU,MAAM,KAAG,aAAa,CAAC,MAAM,CAAC,CAMnE"}
|