@argo-video/cli 0.12.0 → 0.12.2
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.
|
@@ -10,6 +10,8 @@ export interface TransformersEngineOptions {
|
|
|
10
10
|
speakerEmbeddings?: string;
|
|
11
11
|
/** Number of inference steps (higher = better quality). Default: 10 */
|
|
12
12
|
numInferenceSteps?: number;
|
|
13
|
+
/** Default language code for models that require language tags (e.g., Supertonic-TTS-2 needs 'en'). Wraps text as `<lang>text</lang>`. */
|
|
14
|
+
lang?: string;
|
|
13
15
|
}
|
|
14
16
|
export declare class TransformersEngine implements TTSEngine {
|
|
15
17
|
private pipeline;
|
|
@@ -19,6 +21,7 @@ export declare class TransformersEngine implements TTSEngine {
|
|
|
19
21
|
private device;
|
|
20
22
|
private speakerEmbeddings?;
|
|
21
23
|
private numInferenceSteps;
|
|
24
|
+
private lang?;
|
|
22
25
|
constructor(options?: TransformersEngineOptions);
|
|
23
26
|
describe(): TTSEngineMetadata;
|
|
24
27
|
private getPipeline;
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"transformers.d.ts","sourceRoot":"","sources":["../../../src/tts/engines/transformers.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,SAAS,EAAE,gBAAgB,EAAE,iBAAiB,EAAE,MAAM,cAAc,CAAC;AAGnF,MAAM,WAAW,yBAAyB;IACxC,oFAAoF;IACpF,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,wCAAwC;IACxC,KAAK,CAAC,EAAE,MAAM,GAAG,MAAM,GAAG,IAAI,GAAG,IAAI,GAAG,OAAO,CAAC;IAChD,wDAAwD;IACxD,MAAM,CAAC,EAAE,MAAM,GAAG,QAAQ,GAAG,KAAK,GAAG,IAAI,CAAC;IAC1C,2EAA2E;IAC3E,iBAAiB,CAAC,EAAE,MAAM,CAAC;IAC3B,uEAAuE;IACvE,iBAAiB,CAAC,EAAE,MAAM,CAAC;
|
|
1
|
+
{"version":3,"file":"transformers.d.ts","sourceRoot":"","sources":["../../../src/tts/engines/transformers.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,SAAS,EAAE,gBAAgB,EAAE,iBAAiB,EAAE,MAAM,cAAc,CAAC;AAGnF,MAAM,WAAW,yBAAyB;IACxC,oFAAoF;IACpF,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,wCAAwC;IACxC,KAAK,CAAC,EAAE,MAAM,GAAG,MAAM,GAAG,IAAI,GAAG,IAAI,GAAG,OAAO,CAAC;IAChD,wDAAwD;IACxD,MAAM,CAAC,EAAE,MAAM,GAAG,QAAQ,GAAG,KAAK,GAAG,IAAI,CAAC;IAC1C,2EAA2E;IAC3E,iBAAiB,CAAC,EAAE,MAAM,CAAC;IAC3B,uEAAuE;IACvE,iBAAiB,CAAC,EAAE,MAAM,CAAC;IAC3B,0IAA0I;IAC1I,IAAI,CAAC,EAAE,MAAM,CAAC;CACf;AAED,qBAAa,kBAAmB,YAAW,SAAS;IAClD,OAAO,CAAC,QAAQ,CAAa;IAC7B,OAAO,CAAC,WAAW,CAA6B;IAChD,OAAO,CAAC,KAAK,CAAS;IACtB,OAAO,CAAC,KAAK,CAAS;IACtB,OAAO,CAAC,MAAM,CAAgB;IAC9B,OAAO,CAAC,iBAAiB,CAAC,CAAS;IACnC,OAAO,CAAC,iBAAiB,CAAS;IAClC,OAAO,CAAC,IAAI,CAAC,CAAS;gBAEV,OAAO,CAAC,EAAE,yBAAyB;IAS/C,QAAQ,IAAI,iBAAiB;YAIf,WAAW;IAyCnB,QAAQ,CAAC,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,gBAAgB,GAAG,OAAO,CAAC,MAAM,CAAC;CA2EzE"}
|
|
@@ -7,12 +7,14 @@ export class TransformersEngine {
|
|
|
7
7
|
device;
|
|
8
8
|
speakerEmbeddings;
|
|
9
9
|
numInferenceSteps;
|
|
10
|
+
lang;
|
|
10
11
|
constructor(options) {
|
|
11
12
|
this.model = options?.model ?? 'onnx-community/Supertonic-TTS-ONNX';
|
|
12
13
|
this.dtype = options?.dtype ?? 'q8';
|
|
13
14
|
this.device = options?.device ?? null;
|
|
14
15
|
this.speakerEmbeddings = options?.speakerEmbeddings;
|
|
15
16
|
this.numInferenceSteps = options?.numInferenceSteps ?? 10;
|
|
17
|
+
this.lang = options?.lang;
|
|
16
18
|
}
|
|
17
19
|
describe() {
|
|
18
20
|
return { engine: 'transformers', model: this.model, dtype: this.dtype };
|
|
@@ -81,13 +83,18 @@ export class TransformersEngine {
|
|
|
81
83
|
};
|
|
82
84
|
if (speaker)
|
|
83
85
|
baseOpts.speaker_embeddings = speaker;
|
|
86
|
+
// Wrap text in language tags if lang is set (per-scene or engine default).
|
|
87
|
+
// Models like Supertonic-TTS-2 require <en>text</en> format — without
|
|
88
|
+
// these tags the model produces garbled audio.
|
|
89
|
+
const lang = options.lang ?? this.lang;
|
|
84
90
|
const chunks = splitTextForTTS(text);
|
|
85
91
|
const audioChunks = [];
|
|
86
92
|
let sampleRate = TARGET_RATE;
|
|
87
93
|
for (const chunk of chunks) {
|
|
94
|
+
const input = lang ? `<${lang}>${chunk}</${lang}>` : chunk;
|
|
88
95
|
let audio;
|
|
89
96
|
try {
|
|
90
|
-
audio = await tts(
|
|
97
|
+
audio = await tts(input, baseOpts);
|
|
91
98
|
}
|
|
92
99
|
catch (err) {
|
|
93
100
|
throw new Error(`Transformers TTS failed to generate audio for text "${chunk.substring(0, 80)}..." ` +
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"transformers.js","sourceRoot":"","sources":["../../../src/tts/engines/transformers.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,eAAe,EAAE,aAAa,EAAE,MAAM,cAAc,CAAC;
|
|
1
|
+
{"version":3,"file":"transformers.js","sourceRoot":"","sources":["../../../src/tts/engines/transformers.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,eAAe,EAAE,aAAa,EAAE,MAAM,cAAc,CAAC;AAiB9D,MAAM,OAAO,kBAAkB;IACrB,QAAQ,GAAQ,IAAI,CAAC;IACrB,WAAW,GAAwB,IAAI,CAAC;IACxC,KAAK,CAAS;IACd,KAAK,CAAS;IACd,MAAM,CAAgB;IACtB,iBAAiB,CAAU;IAC3B,iBAAiB,CAAS;IAC1B,IAAI,CAAU;IAEtB,YAAY,OAAmC;QAC7C,IAAI,CAAC,KAAK,GAAG,OAAO,EAAE,KAAK,IAAI,oCAAoC,CAAC;QACpE,IAAI,CAAC,KAAK,GAAG,OAAO,EAAE,KAAK,IAAI,IAAI,CAAC;QACpC,IAAI,CAAC,MAAM,GAAG,OAAO,EAAE,MAAM,IAAI,IAAI,CAAC;QACtC,IAAI,CAAC,iBAAiB,GAAG,OAAO,EAAE,iBAAiB,CAAC;QACpD,IAAI,CAAC,iBAAiB,GAAG,OAAO,EAAE,iBAAiB,IAAI,EAAE,CAAC;QAC1D,IAAI,CAAC,IAAI,GAAG,OAAO,EAAE,IAAI,CAAC;IAC5B,CAAC;IAED,QAAQ;QACN,OAAO,EAAE,MAAM,EAAE,cAAc,EAAE,KAAK,EAAE,IAAI,CAAC,KAAK,EAAE,KAAK,EAAE,IAAI,CAAC,KAAK,EAAE,CAAC;IAC1E,CAAC;IAEO,KAAK,CAAC,WAAW;QACvB,IAAI,IAAI,CAAC,QAAQ;YAAE,OAAO,IAAI,CAAC,QAAQ,CAAC;QACxC,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE,CAAC;YACtB,IAAI,CAAC,WAAW,GAAG,CAAC,KAAK,IAAI,EAAE;gBAC7B,IAAI,QAAa,CAAC;gBAClB,IAAI,CAAC;oBACH,CAAC,EAAE,QAAQ,EAAE,GAAG,MAAM,MAAM,CAAC,2BAA2B,CAAC,CAAC,CAAC;gBAC7D,CAAC;gBAAC,MAAM,CAAC;oBACP,MAAM,IAAI,KAAK,CACb,4EAA4E;wBAC5E,kDAAkD,CACnD,CAAC;gBACJ,CAAC;gBACD,IAAI,CAAC;oBACH,IAAI,CAAC,QAAQ,GAAG,MAAM,QAAQ,CAAC,gBAAgB,EAAE,IAAI,CAAC,KAAK,EAAE;wBAC3D,KAAK,EAAE,IAAI,CAAC,KAAK;wBACjB,MAAM,EAAE,IAAI,CAAC,MAAM;qBACpB,CAAC,CAAC;gBACL,CAAC;gBAAC,OAAO,GAAG,EAAE,CAAC;oBACb,IAAI,CAAC,WAAW,GAAG,IAAI,CAAC;oBACxB,MAAM,GAAG,GAAI,GAAa,CAAC,OAAO,CAAC;oBACnC,iEAAiE;oBACjE,IAAI,GAAG,CAAC,QAAQ,CAAC,WAAW,CAAC,IAAI,GAAG,CAAC,QAAQ,CAAC,uBAAuB,CAAC,EAAE,CAAC;wBACvE,MAAM,IAAI,KAAK,CACb,yBAAyB,IAAI,CAAC,KAAK,iBAAiB,IAAI,CAAC,KAAK,sCAAsC;4BACpG,6CAA6C;4BAC7C,mFAAmF,IAAI,CAAC,KAAK,qBAAqB,CACnH,CAAC;oBACJ,CAAC;oBACD,MAAM,IAAI,KAAK,CACb,0DAA0D,IAAI,CAAC,KAAK,YAAY,IAAI,CAAC,KAAK,KAAK;wBAC/F,yEAAyE;wBACzE,mBAAmB,GAAG,EAAE,CACzB,CAAC;gBACJ,CAAC;gBACD,OAAO,IAAI,CAAC,QAAQ,CAAC;YACvB,CAAC,CAAC,EAAE,CAAC;QACP,CAAC;QACD,OAAO,IAAI,CAAC,WAAW,CAAC;IAC1B,CAAC;IAED,KAAK,CAAC,QAAQ,CAAC,IAAY,EAAE,OAAyB;QACpD,IAAI,CAAC,IAAI,EAAE,IAAI,EAAE;YAAE,MAAM,IAAI,KAAK,CAAC,4BAA4B,CAAC,CAAC;QACjE,MAAM,GAAG,GAAG,MAAM,IAAI,CAAC,WAAW,EAAE,CAAC;QAErC,MAAM,EAAE,eAAe,EAAE,GAAG,MAAM,MAAM,CAAC,cAAc,CAAC,CAAC;QACzD,MAAM,WAAW,GAAG,KAAK,CAAC;QAE1B,iCAAiC;QACjC,oFAAoF;QACpF,0EAA0E;QAC1E,sEAAsE;QACtE,IAAI,OAAO,GAAG,IAAI,CAAC,iBAAiB,CAAC;QACrC,IAAI,OAAO,CAAC,KAAK,EAAE,CAAC;YAClB,IAAI,OAAO,CAAC,KAAK,CAAC,QAAQ,CAAC,GAAG,CAAC,IAAI,OAAO,CAAC,KAAK,CAAC,QAAQ,CAAC,IAAI,CAAC,IAAI,OAAO,CAAC,KAAK,CAAC,QAAQ,CAAC,MAAM,CAAC,EAAE,CAAC;gBAClG,OAAO,GAAG,OAAO,CAAC,KAAK,CAAC;YAC1B,CAAC;iBAAM,IAAI,CAAC,IAAI,CAAC,iBAAiB,EAAE,CAAC;gBACnC,OAAO,CAAC,IAAI,CACV,mBAAmB,OAAO,CAAC,KAAK,kEAAkE;oBAClG,WAAW,IAAI,CAAC,KAAK,qCAAqC;oBAC1D,+GAA+G,CAChH,CAAC;YACJ,CAAC;QACH,CAAC;QACD,MAAM,QAAQ,GAA4B;YACxC,KAAK,EAAE,OAAO,CAAC,KAAK,IAAI,GAAG;YAC3B,mBAAmB,EAAE,IAAI,CAAC,iBAAiB;SAC5C,CAAC;QACF,IAAI,OAAO;YAAE,QAAQ,CAAC,kBAAkB,GAAG,OAAO,CAAC;QAEnD,2EAA2E;QAC3E,sEAAsE;QACtE,+CAA+C;QAC/C,MAAM,IAAI,GAAG,OAAO,CAAC,IAAI,IAAI,IAAI,CAAC,IAAI,CAAC;QAEvC,MAAM,MAAM,GAAG,eAAe,CAAC,IAAI,CAAC,CAAC;QACrC,MAAM,WAAW,GAAmB,EAAE,CAAC;QACvC,IAAI,UAAU,GAAG,WAAW,CAAC;QAE7B,KAAK,MAAM,KAAK,IAAI,MAAM,EAAE,CAAC;YAC3B,MAAM,KAAK,GAAG,IAAI,CAAC,CAAC,CAAC,IAAI,IAAI,IAAI,KAAK,KAAK,IAAI,GAAG,CAAC,CAAC,CAAC,KAAK,CAAC;YAC3D,IAAI,KAAU,CAAC;YACf,IAAI,CAAC;gBACH,KAAK,GAAG,MAAM,GAAG,CAAC,KAAK,EAAE,QAAQ,CAAC,CAAC;YACrC,CAAC;YAAC,OAAO,GAAG,EAAE,CAAC;gBACb,MAAM,IAAI,KAAK,CACb,uDAAuD,KAAK,CAAC,SAAS,CAAC,CAAC,EAAE,EAAE,CAAC,OAAO;oBACpF,WAAW,IAAI,CAAC,KAAK,sBAAuB,GAAa,CAAC,OAAO,EAAE,CACpE,CAAC;YACJ,CAAC;YAED,MAAM,OAAO,GAAiB,KAAK,CAAC,KAAK,IAAI,KAAK,CAAC,IAAI,CAAC;YACxD,IAAI,CAAC,OAAO,IAAI,CAAC,CAAC,OAAO,YAAY,YAAY,CAAC,EAAE,CAAC;gBACnD,MAAM,IAAI,KAAK,CACb,gEAAgE,IAAI,CAAC,KAAK,IAAI;oBAC9E,qDAAqD,CACtD,CAAC;YACJ,CAAC;YACD,UAAU,GAAG,KAAK,CAAC,aAAa,IAAI,WAAW,CAAC;YAChD,WAAW,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC;QAC5B,CAAC;QAED,uCAAuC;QACvC,MAAM,QAAQ,GAAG,aAAa,CAAC,WAAW,EAAE,UAAU,CAAC,CAAC;QAExD,kEAAkE;QAClE,mEAAmE;QACnE,2FAA2F;QAC3F,IAAI,UAAU,KAAK,WAAW,EAAE,CAAC;YAC/B,MAAM,EAAE,YAAY,EAAE,GAAG,MAAM,MAAM,CAAC,cAAc,CAAC,CAAC;YACtD,MAAM,SAAS,GAAG,eAAe,CAAC,QAAQ,EAAE,UAAU,CAAC,CAAC;YACxD,OAAO,YAAY,CAAC,SAAS,CAAC,CAAC;QACjC,CAAC;QAED,OAAO,eAAe,CAAC,QAAQ,EAAE,UAAU,CAAC,CAAC;IAC/C,CAAC;CACF"}
|