@omnimedia/omnitool 1.1.0-3 → 1.1.0-5
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/package.json +11 -9
- package/s/context.ts +1 -1
- package/s/demo/demo.bundle.ts +6 -2
- package/s/demo/routines/filmstrip-test.ts +2 -2
- package/s/demo/routines/transcriber-test.ts +34 -0
- package/s/demo/routines/transitions-test.ts +43 -0
- package/s/driver/fns/host.ts +7 -6
- package/s/driver/fns/schematic.ts +1 -1
- package/s/driver/fns/work.ts +116 -119
- package/s/driver/utils/load-decoder-source.ts +3 -4
- package/s/features/speech/transcribe/default-spec.ts +11 -0
- package/s/features/speech/transcribe/parts/load-pipe.ts +19 -0
- package/s/features/speech/transcribe/parts/prep-audio.ts +23 -0
- package/s/features/speech/transcribe/parts/transcribe.ts +70 -0
- package/s/features/speech/transcribe/transcriber.ts +46 -0
- package/s/features/speech/transcribe/types.ts +82 -0
- package/s/features/speech/transcribe/worker.bundle.ts +40 -0
- package/s/features/transition/parts/fragment.ts +24 -0
- package/s/features/transition/parts/types.ts +94 -0
- package/s/features/transition/parts/uniforms.ts +29 -0
- package/s/features/transition/parts/vertex.ts +31 -0
- package/s/features/transition/transition.ts +60 -0
- package/s/timeline/utils/checksum.ts +2 -1
- package/s/tools/common/loader.ts +26 -0
- package/s/tools/common/transformer-pipeline.ts +26 -0
- package/s/tools/speech-recognition/common/model.ts +26 -0
- package/s/tools/speech-recognition/whisper/fns/host.ts +25 -0
- package/s/tools/speech-recognition/whisper/fns/schematic.ts +23 -0
- package/s/tools/speech-recognition/whisper/fns/work.ts +91 -0
- package/s/tools/speech-recognition/whisper/parts/types.ts +38 -0
- package/s/tools/speech-recognition/whisper/parts/worker.bundle.ts +7 -0
- package/s/tools/speech-recognition/whisper/tool.ts +70 -0
- package/x/context.js +1 -1
- package/x/context.js.map +1 -1
- package/x/demo/demo.bundle.js +6 -2
- package/x/demo/demo.bundle.js.map +1 -1
- package/x/demo/demo.bundle.min.js +6 -6
- package/x/demo/demo.bundle.min.js.map +4 -4
- package/x/demo/routines/filmstrip-test.d.ts +1 -1
- package/x/demo/routines/filmstrip-test.js +2 -2
- package/x/demo/routines/filmstrip-test.js.map +1 -1
- package/x/demo/routines/transcriber-test.d.ts +4 -0
- package/x/demo/routines/transcriber-test.js +33 -0
- package/x/demo/routines/transcriber-test.js.map +1 -0
- package/x/demo/routines/transitions-test.d.ts +5 -0
- package/x/demo/routines/transitions-test.js +35 -0
- package/x/demo/routines/transitions-test.js.map +1 -0
- package/x/driver/driver.worker.bundle.min.js +80 -80
- package/x/driver/driver.worker.bundle.min.js.map +4 -4
- package/x/driver/fns/host.js +3 -3
- package/x/driver/fns/host.js.map +1 -1
- package/x/driver/fns/schematic.d.ts +1 -1
- package/x/driver/fns/work.js +8 -8
- package/x/driver/fns/work.js.map +1 -1
- package/x/driver/utils/load-decoder-source.d.ts +2 -1
- package/x/driver/utils/load-decoder-source.js +2 -3
- package/x/driver/utils/load-decoder-source.js.map +1 -1
- package/x/features/speech/transcribe/default-spec.d.ts +2 -0
- package/x/features/speech/transcribe/default-spec.js +8 -0
- package/x/features/speech/transcribe/default-spec.js.map +1 -0
- package/x/features/speech/transcribe/parts/load-pipe.d.ts +2 -0
- package/x/features/speech/transcribe/parts/load-pipe.js +13 -0
- package/x/features/speech/transcribe/parts/load-pipe.js.map +1 -0
- package/x/features/speech/transcribe/parts/prep-audio.d.ts +5 -0
- package/x/features/speech/transcribe/parts/prep-audio.js +21 -0
- package/x/features/speech/transcribe/parts/prep-audio.js.map +1 -0
- package/x/features/speech/transcribe/parts/transcribe.d.ts +5 -0
- package/x/features/speech/transcribe/parts/transcribe.js +56 -0
- package/x/features/speech/transcribe/parts/transcribe.js.map +1 -0
- package/x/features/speech/transcribe/transcriber.d.ts +5 -0
- package/x/features/speech/transcribe/transcriber.js +33 -0
- package/x/features/speech/transcribe/transcriber.js.map +1 -0
- package/x/features/speech/transcribe/types.d.ts +66 -0
- package/x/features/speech/transcribe/types.js +2 -0
- package/x/features/speech/transcribe/types.js.map +1 -0
- package/x/features/speech/transcribe/worker.bundle.d.ts +1 -0
- package/x/features/speech/transcribe/worker.bundle.js +33 -0
- package/x/features/speech/transcribe/worker.bundle.js.map +1 -0
- package/x/features/speech/transcribe/worker.bundle.min.js +2916 -0
- package/x/features/speech/transcribe/worker.bundle.min.js.map +7 -0
- package/x/features/transition/parts/fragment.d.ts +1 -0
- package/x/features/transition/parts/fragment.js +25 -0
- package/x/features/transition/parts/fragment.js.map +1 -0
- package/x/features/transition/parts/types.d.ts +23 -0
- package/x/features/transition/parts/types.js +2 -0
- package/x/features/transition/parts/types.js.map +1 -0
- package/x/features/transition/parts/uniforms.d.ts +31 -0
- package/x/features/transition/parts/uniforms.js +27 -0
- package/x/features/transition/parts/uniforms.js.map +1 -0
- package/x/features/transition/parts/vertex.d.ts +1 -0
- package/x/features/transition/parts/vertex.js +32 -0
- package/x/features/transition/parts/vertex.js.map +1 -0
- package/x/features/transition/transition.d.ts +5 -0
- package/x/features/transition/transition.js +50 -0
- package/x/features/transition/transition.js.map +1 -0
- package/x/index.html +2 -2
- package/x/timeline/utils/checksum.js +2 -1
- package/x/timeline/utils/checksum.js.map +1 -1
- package/x/tools/common/loader.d.ts +19 -0
- package/x/tools/common/loader.js +18 -0
- package/x/tools/common/loader.js.map +1 -0
- package/x/tools/common/transformer-pipeline.d.ts +8 -0
- package/x/tools/common/transformer-pipeline.js +24 -0
- package/x/tools/common/transformer-pipeline.js.map +1 -0
- package/x/tools/speech-recognition/common/model.d.ts +14 -0
- package/x/tools/speech-recognition/common/model.js +16 -0
- package/x/tools/speech-recognition/common/model.js.map +1 -0
- package/x/tools/speech-recognition/whisper/fns/host.d.ts +13 -0
- package/x/tools/speech-recognition/whisper/fns/host.js +19 -0
- package/x/tools/speech-recognition/whisper/fns/host.js.map +1 -0
- package/x/tools/speech-recognition/whisper/fns/schematic.d.ts +19 -0
- package/x/tools/speech-recognition/whisper/fns/schematic.js +2 -0
- package/x/tools/speech-recognition/whisper/fns/schematic.js.map +1 -0
- package/x/tools/speech-recognition/whisper/fns/work.d.ts +12 -0
- package/x/tools/speech-recognition/whisper/fns/work.js +74 -0
- package/x/tools/speech-recognition/whisper/fns/work.js.map +1 -0
- package/x/tools/speech-recognition/whisper/parts/types.d.ts +31 -0
- package/x/tools/speech-recognition/whisper/parts/types.js +2 -0
- package/x/tools/speech-recognition/whisper/parts/types.js.map +1 -0
- package/x/tools/speech-recognition/whisper/parts/worker.bundle.d.ts +1 -0
- package/x/tools/speech-recognition/whisper/parts/worker.bundle.js +4 -0
- package/x/tools/speech-recognition/whisper/parts/worker.bundle.js.map +1 -0
- package/x/tools/speech-recognition/whisper/parts/worker.bundle.min.js +8 -0
- package/x/tools/speech-recognition/whisper/parts/worker.bundle.min.js.map +7 -0
- package/x/tools/speech-recognition/whisper/tool.d.ts +12 -0
- package/x/tools/speech-recognition/whisper/tool.js +63 -0
- package/x/tools/speech-recognition/whisper/tool.js.map +1 -0
package/x/driver/fns/host.js
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
import { Comrade } from "@e280/comrade";
|
|
2
|
-
export const setupDriverHost = (machina) => Comrade.host(
|
|
2
|
+
export const setupDriverHost = (machina) => (Comrade.host(_shell => ({
|
|
3
3
|
async world() {
|
|
4
4
|
machina.count++;
|
|
5
|
-
}
|
|
6
|
-
}));
|
|
5
|
+
},
|
|
6
|
+
})));
|
|
7
7
|
//# sourceMappingURL=host.js.map
|
package/x/driver/fns/host.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"host.js","sourceRoot":"","sources":["../../../s/driver/fns/host.ts"],"names":[],"mappings":"AACA,OAAO,EAAC,OAAO,EAAC,MAAM,eAAe,CAAA;AAIrC,MAAM,CAAC,MAAM,eAAe,GAAG,CAAC,OAAgB,EAAE,EAAE,CAAC,OAAO,CAAC,IAAI,CAAkB,CAAC,
|
|
1
|
+
{"version":3,"file":"host.js","sourceRoot":"","sources":["../../../s/driver/fns/host.ts"],"names":[],"mappings":"AACA,OAAO,EAAC,OAAO,EAAC,MAAM,eAAe,CAAA;AAIrC,MAAM,CAAC,MAAM,eAAe,GAAG,CAAC,OAAgB,EAAE,EAAE,CAAC,CACpD,OAAO,CAAC,IAAI,CAAkB,MAAM,CAAC,EAAE,CAAC,CAAC;IACxC,KAAK,CAAC,KAAK;QACV,OAAO,CAAC,KAAK,EAAE,CAAA;IAChB,CAAC;CACD,CAAC,CAAC,CACH,CAAA"}
|
|
@@ -27,7 +27,7 @@ export interface EncoderInput {
|
|
|
27
27
|
audio: AudioEncodingConfig;
|
|
28
28
|
};
|
|
29
29
|
}
|
|
30
|
-
export type DecoderSource =
|
|
30
|
+
export type DecoderSource = Blob | string | URL;
|
|
31
31
|
export interface DecoderInput {
|
|
32
32
|
source: DecoderSource;
|
|
33
33
|
onFrame?: (frame: VideoFrame) => Promise<VideoFrame>;
|
package/x/driver/fns/work.js
CHANGED
|
@@ -1,16 +1,15 @@
|
|
|
1
1
|
import { Comrade } from "@e280/comrade";
|
|
2
|
-
import { Input, ALL_FORMATS, VideoSampleSink, Output, Mp4OutputFormat, VideoSampleSource, VideoSample, AudioSampleSink, AudioSampleSource, AudioSample, StreamTarget, BlobSource, UrlSource } from "mediabunny";
|
|
3
2
|
import { autoDetectRenderer, Container, Sprite, Text, Texture, DOMAdapter, WebWorkerAdapter } from "pixi.js";
|
|
3
|
+
import { Input, ALL_FORMATS, VideoSampleSink, Output, Mp4OutputFormat, VideoSampleSource, VideoSample, AudioSampleSink, AudioSampleSource, AudioSample, StreamTarget, BlobSource, UrlSource } from "mediabunny";
|
|
4
4
|
DOMAdapter.set(WebWorkerAdapter);
|
|
5
|
-
export const setupDriverWork = Comrade.work(
|
|
5
|
+
export const setupDriverWork = (Comrade.work(shell => ({
|
|
6
6
|
async hello() {
|
|
7
|
-
await host.world();
|
|
7
|
+
await shell.host.world();
|
|
8
8
|
},
|
|
9
9
|
async decode({ source, video, audio }) {
|
|
10
10
|
const loadSource = async () => {
|
|
11
|
-
if (source instanceof
|
|
12
|
-
|
|
13
|
-
return new BlobSource(file);
|
|
11
|
+
if (source instanceof Blob) {
|
|
12
|
+
return new BlobSource(source);
|
|
14
13
|
}
|
|
15
14
|
else {
|
|
16
15
|
return new UrlSource(source);
|
|
@@ -110,10 +109,11 @@ export const setupDriverWork = Comrade.work(({ host }, rig) => ({
|
|
|
110
109
|
for (const disposable of disposables) {
|
|
111
110
|
disposable.destroy(true);
|
|
112
111
|
}
|
|
113
|
-
|
|
112
|
+
shell.transfer = [frame];
|
|
114
113
|
return frame;
|
|
115
114
|
}
|
|
116
|
-
}));
|
|
115
|
+
})));
|
|
116
|
+
// TODO suspicious global, probably bad
|
|
117
117
|
let pixi = null;
|
|
118
118
|
async function renderPIXI(width, height) {
|
|
119
119
|
if (pixi)
|
package/x/driver/fns/work.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"work.js","sourceRoot":"","sources":["../../../s/driver/fns/work.ts"],"names":[],"mappings":"AAAA,OAAO,EAAC,OAAO,EAAC,MAAM,eAAe,CAAA;AACrC,OAAO,
|
|
1
|
+
{"version":3,"file":"work.js","sourceRoot":"","sources":["../../../s/driver/fns/work.ts"],"names":[],"mappings":"AAAA,OAAO,EAAC,OAAO,EAAC,MAAM,eAAe,CAAA;AACrC,OAAO,EAAC,kBAAkB,EAAE,SAAS,EAAY,MAAM,EAAE,IAAI,EAAE,OAAO,EAAE,UAAU,EAAE,gBAAgB,EAAC,MAAM,SAAS,CAAA;AACpH,OAAO,EAAC,KAAK,EAAE,WAAW,EAAE,eAAe,EAAE,MAAM,EAAE,eAAe,EAAE,iBAAiB,EAAE,WAAW,EAAE,eAAe,EAAE,iBAAiB,EAAE,WAAW,EAAE,YAAY,EAAE,UAAU,EAAE,SAAS,EAAC,MAAM,YAAY,CAAA;AAI7M,UAAU,CAAC,GAAG,CAAC,gBAAgB,CAAC,CAAA;AAEhC,MAAM,CAAC,MAAM,eAAe,GAAG,CAC9B,OAAO,CAAC,IAAI,CAAkB,KAAK,CAAC,EAAE,CAAC,CAAC;IACvC,KAAK,CAAC,KAAK;QACV,MAAM,KAAK,CAAC,IAAI,CAAC,KAAK,EAAE,CAAA;IACzB,CAAC;IAED,KAAK,CAAC,MAAM,CAAC,EAAC,MAAM,EAAE,KAAK,EAAE,KAAK,EAAC;QAClC,MAAM,UAAU,GAAG,KAAK,IAAI,EAAE;YAC7B,IAAG,MAAM,YAAY,IAAI,EAAE,CAAC;gBAC3B,OAAO,IAAI,UAAU,CAAC,MAAM,CAAC,CAAA;YAC9B,CAAC;iBAAM,CAAC;gBACP,OAAO,IAAI,SAAS,CAAC,MAAM,CAAC,CAAA;YAC7B,CAAC;QACF,CAAC,CAAA;QACD,MAAM,KAAK,GAAG,IAAI,KAAK,CAAC;YACvB,MAAM,EAAE,MAAM,UAAU,EAAE;YAC1B,OAAO,EAAE,WAAW;SACpB,CAAC,CAAA;QAEF,MAAM,CAAC,UAAU,EAAE,UAAU,CAAC,GAAG,MAAM,OAAO,CAAC,GAAG,CAAC;YAClD,KAAK,CAAC,oBAAoB,EAAE;YAC5B,KAAK,CAAC,oBAAoB,EAAE;SAC5B,CAAC,CAAA;QAEF,MAAM,cAAc,GAAG,MAAM,UAAU,EAAE,SAAS,EAAE,CAAA;QACpD,MAAM,cAAc,GAAG,MAAM,UAAU,EAAE,SAAS,EAAE,CAAA;QAEpD,MAAM,WAAW,GAAG,KAAK,CAAC,SAAS,EAAE,CAAA;QACrC,MAAM,WAAW,GAAG,KAAK,CAAC,SAAS,EAAE,CAAA;QAErC,MAAM,OAAO,CAAC,GAAG,CAAC;YACjB,CAAC,KAAK,IAAI,EAAE;gBACX,IAAI,cAAc,IAAI,UAAU,EAAE,CAAC;oBAClC,MAAM,IAAI,GAAG,IAAI,eAAe,CAAC,UAAU,CAAC,CAAA;oBAC5C,IAAI,KAAK,EAAE,MAAM,MAAM,IAAI,IAAI,CAAC,OAAO,EAAE,EAAE,CAAC;wBAC3C,MAAM,KAAK,GAAG,MAAM,CAAC,YAAY,EAAE,CAAA;wBACnC,MAAM,WAAW,CAAC,KAAK,CAAC,KAAK,CAAC,CAAA;wBAC9B,MAAM,CAAC,KAAK,EAAE,CAAA;wBACd,KAAK,CAAC,KAAK,EAAE,CAAA;oBACd,CAAC;oBACD,MAAM,WAAW,CAAC,KAAK,EAAE,CAAA;gBAC1B,CAAC;YACF,CAAC,CAAC,EAAE;YACJ,CAAC,KAAK,IAAI,EAAE;gBACX,IAAI,cAAc,IAAI,UAAU,EAAE,CAAC;oBAClC,MAAM,IAAI,GAAG,IAAI,eAAe,CAAC,UAAU,CAAC,CAAA;oBAC5C,IAAI,KAAK,EAAE,MAAM,MAAM,IAAI,IAAI,CAAC,OAAO,EAAE,EAAE,CAAC;wBAC3C,MAAM,KAAK,GAAG,MAAM,CAAC,WAAW,EAAE,CAAA;wBAClC,MAAM,WAAW,CAAC,KAAK,CAAC,KAAK,CAAC,CAAA;wBAC9B,MAAM,CAAC,KAAK,EAAE,CAAA;wBACd,KAAK,CAAC,KAAK,EAAE,CAAA;oBACd,CAAC;oBACD,MAAM,WAAW,CAAC,KAAK,EAAE,CAAA;gBAC1B,CAAC;YACF,CAAC,CAAC,EAAE;SACJ,CAAC,CAAA;IACH,CAAC;IAED,KAAK,CAAC,MAAM,CAAC,EAAC,SAAS,EAAE,MAAM,EAAE,MAAM,EAAC;QACvC,MAAM,MAAM,GAAG,IAAI,MAAM,CAAC;YACzB,MAAM,EAAE,IAAI,eAAe,EAAE;YAC7B,MAAM,EAAE,IAAI,YAAY,CAAC,MAAM,EAAE,EAAC,OAAO,EAAE,IAAI,EAAC,CAAC;SACjD,CAAC,CAAA;QACF,MAAM,WAAW,GAAG,IAAI,iBAAiB,CAAC,MAAM,CAAC,KAAK,CAAC,CAAA;QACvD,MAAM,CAAC,aAAa,CAAC,WAAW,CAAC,CAAA;QACjC,oFAAoF;QACpF,uEAAuE;QACvE,MAAM,WAAW,GAAG,IAAI,iBAAiB,CAAC,MAAM,CAAC,KAAK,CAAC,CAAA;QACvD,MAAM,CAAC,aAAa,CAAC,WAAW,CAAC,CAAA;QAEjC,MAAM,MAAM,CAAC,KAAK,EAAE,CAAA;QAEpB,MAAM,WAAW,GAAG,SAAS,CAAC,KAAK,CAAC,SAAS,EAAE,CAAA;QAC/C,MAAM,WAAW,GAAG,SAAS,CAAC,KAAK,CAAC,SAAS,EAAE,CAAA;QAE/C,MAAM,OAAO,CAAC,GAAG,CAAC;YACjB,CAAC,KAAK,IAAI,EAAE;gBACX,OAAO,IAAI,EAAE,CAAC;oBACb,MAAM,EAAC,IAAI,EAAE,KAAK,EAAC,GAAG,MAAM,WAAW,CAAC,IAAI,EAAE,CAAA;oBAC9C,IAAI,IAAI;wBAAE,MAAK;oBACf,MAAM,MAAM,GAAG,IAAI,WAAW,CAAC,KAAK,CAAC,CAAA;oBACrC,MAAM,WAAW,CAAC,GAAG,CAAC,MAAM,CAAC,CAAA;oBAC7B,MAAM,CAAC,KAAK,EAAE,CAAA;gBACf,CAAC;YACF,CAAC,CAAC,EAAE;YACJ,CAAC,KAAK,IAAI,EAAE;gBACX,OAAO,IAAI,EAAE,CAAC;oBACb,MAAM,EAAC,IAAI,EAAE,KAAK,EAAC,GAAG,MAAM,WAAW,CAAC,IAAI,EAAE,CAAA;oBAC9C,IAAI,IAAI;wBAAE,MAAK;oBACf,MAAM,MAAM,GAAG,IAAI,WAAW,CAAC,KAAK,CAAC,CAAA;oBACrC,MAAM,WAAW,CAAC,GAAG,CAAC,MAAM,CAAC,CAAA;oBAC7B,MAAM,CAAC,KAAK,EAAE,CAAA;oBACd,KAAK,CAAC,KAAK,EAAE,CAAA;gBACd,CAAC;YACF,CAAC,CAAC,EAAE;SACJ,CAAC,CAAA;QAEF,MAAM,MAAM,CAAC,QAAQ,EAAE,CAAA;IACxB,CAAC;IAED,KAAK,CAAC,SAAS,CAAC,WAAW;QAC1B,MAAM,EAAC,KAAK,EAAE,QAAQ,EAAC,GAAG,MAAM,UAAU,CAAC,IAAI,EAAE,IAAI,CAAC,CAAA;QACtD,KAAK,CAAC,cAAc,EAAE,CAAA;QAEtB,MAAM,EAAC,SAAS,EAAE,WAAW,EAAC,GAAG,MAAM,WAAW,CAAC,WAAW,EAAE,KAAK,CAAC,CAAA;QACtE,QAAQ,CAAC,MAAM,CAAC,KAAK,CAAC,CAAA;QAEtB,gGAAgG;QAChG,yCAAyC;QACzC,MAAM,KAAK,GAAG,IAAI,UAAU,CAAC,QAAQ,CAAC,MAAM,EAAE;YAC7C,SAAS,EAAE,SAAS,EAAE,SAAS;YAC/B,QAAQ,EAAE,SAAS,EAAE,QAAQ,IAAI,SAAS;SAC1C,CAAC,CAAA;QAEF,SAAS,EAAE,KAAK,EAAE,CAAA;QAClB,QAAQ,CAAC,KAAK,EAAE,CAAA;QAEhB,KAAK,MAAM,UAAU,IAAI,WAAW,EAAE,CAAC;YACtC,UAAU,CAAC,OAAO,CAAC,IAAI,CAAC,CAAA;QACzB,CAAC;QAED,KAAK,CAAC,QAAQ,GAAG,CAAC,KAAK,CAAC,CAAA;QACxB,OAAO,KAAK,CAAA;IACb,CAAC;CACD,CAAC,CAAC,CACH,CAAA;AAED,uCAAuC;AACvC,IAAI,IAAI,GAGG,IAAI,CAAA;AAEf,KAAK,UAAU,UAAU,CAAC,KAAa,EAAE,MAAc;IACtD,IAAI,IAAI;QACP,OAAO,IAAI,CAAA;IAEZ,MAAM,QAAQ,GAAG,MAAM,kBAAkB,CAAC;QACzC,KAAK;QACL,MAAM;QACN,UAAU,EAAE,OAAO,EAAE,iDAAiD;QACtE,UAAU,EAAE,OAAO;QACnB,kBAAkB,EAAE,CAAC;KACrB,CAAC,CAAA;IAEF,MAAM,KAAK,GAAG,IAAI,SAAS,EAAE,CAAA;IAC7B,IAAI,GAAG,EAAC,QAAQ,EAAE,KAAK,EAAC,CAAA;IAExB,OAAO,IAAI,CAAA;AACZ,CAAC;AAID,KAAK,UAAU,WAAW,CACzB,KAA0B,EAC1B,MAAiB,EACjB,cAAkC,EAAE;IAEpC,IAAI,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC,EAAE,CAAC;QAC1B,IAAI,SAAiC,CAAA;QACrC,KAAK,MAAM,KAAK,IAAI,KAAK,EAAE,CAAC;YAC3B,MAAM,MAAM,GAAG,MAAM,WAAW,CAAC,KAAK,EAAE,MAAM,EAAE,WAAW,CAAC,CAAA;YAC5D,SAAS,KAAK,MAAM,CAAC,SAAS,CAAA;QAC/B,CAAC;QACD,OAAO,EAAC,SAAS,EAAE,WAAW,EAAC,CAAA;IAChC,CAAC;IAED,IAAI,CAAC,iBAAiB,CAAC,KAAK,CAAC,EAAE,CAAC;QAC/B,OAAO,CAAC,IAAI,CAAC,eAAe,EAAE,KAAK,CAAC,CAAA;QACpC,OAAO,EAAC,WAAW,EAAC,CAAA;IACrB,CAAC;IAED,QAAQ,KAAK,CAAC,IAAI,EAAE,CAAC;QACpB,KAAK,MAAM;YACV,OAAO,eAAe,CAAC,KAAK,EAAE,MAAM,EAAE,WAAW,CAAC,CAAA;QACnD,KAAK,OAAO;YACX,OAAO,gBAAgB,CAAC,KAAK,EAAE,MAAM,EAAE,WAAW,CAAC,CAAA;QACpD;YACC,OAAO,CAAC,IAAI,CAAC,oBAAoB,EAAG,KAAa,CAAC,IAAI,CAAC,CAAA;YACvD,OAAO,EAAC,WAAW,EAAC,CAAA;IACtB,CAAC;AACF,CAAC;AAED,SAAS,iBAAiB,CAAC,KAAU;IACpC,OAAO,CAAC,CAAC,KAAK,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,OAAO,KAAK,CAAC,IAAI,KAAK,QAAQ,CAAA;AAC9E,CAAC;AAED,SAAS,eAAe,CACvB,KAAqC,EACrC,MAAiB,EACjB,WAA+B;IAE/B,MAAM,IAAI,GAAG,IAAI,IAAI,CAAC;QACrB,IAAI,EAAE,KAAK,CAAC,OAAO;QACnB,KAAK,EAAE;YACN,UAAU,EAAE,YAAY;YACxB,QAAQ,EAAE,KAAK,CAAC,QAAQ,IAAI,EAAE;YAC9B,IAAI,EAAE,KAAK,CAAC,KAAK,IAAI,OAAO;SAC5B;KACD,CAAC,CAAA;IACF,cAAc,CAAC,IAAI,EAAE,KAAK,CAAC,CAAA;IAC3B,MAAM,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAA;IACrB,WAAW,CAAC,IAAI,CAAC,IAAI,CAAC,CAAA;IACtB,OAAO,EAAC,WAAW,EAAC,CAAA;AACrB,CAAC;AAED,SAAS,gBAAgB,CACxB,KAAsC,EACtC,MAAiB,EACjB,WAA+B;IAE/B,MAAM,OAAO,GAAG,OAAO,CAAC,IAAI,CAAC,KAAK,CAAC,KAAK,CAAC,CAAA;IACzC,MAAM,MAAM,GAAG,IAAI,MAAM,CAAC,OAAO,CAAC,CAAA;IAClC,cAAc,CAAC,MAAM,EAAE,KAAK,CAAC,CAAA;IAC7B,MAAM,CAAC,QAAQ,CAAC,MAAM,CAAC,CAAA;IACvB,WAAW,CAAC,IAAI,CAAC,MAAM,EAAE,OAAO,CAAC,CAAA;IACjC,OAAO,EAAC,SAAS,EAAE,KAAK,CAAC,KAAK,EAAE,WAAW,EAAC,CAAA;AAC7C,CAAC;AAED,SAAS,cAAc,CAAC,MAAqB,EAAE,IAAe,EAAE;IAC/D,IAAG,CAAC,CAAC,CAAC;QAAE,MAAM,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAA;IACtB,IAAG,CAAC,CAAC,CAAC;QAAE,MAAM,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAA;IACtB,IAAG,CAAC,CAAC,KAAK;QAAE,MAAM,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,KAAK,CAAC,CAAA;IACrC,IAAG,CAAC,CAAC,OAAO;QAAE,MAAM,CAAC,KAAK,GAAG,CAAC,CAAC,OAAO,CAAA;IACtC,IAAG,CAAC,CAAC,MAAM,IAAI,QAAQ,IAAI,MAAM;QAAE,MAAM,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,MAAM,CAAC,CAAA;AAC/D,CAAC"}
|
|
@@ -1,2 +1,3 @@
|
|
|
1
|
+
import { BlobSource, UrlSource } from "mediabunny";
|
|
1
2
|
import { DecoderSource } from "../fns/schematic.js";
|
|
2
|
-
export declare function loadDecoderSource(source: DecoderSource): Promise<
|
|
3
|
+
export declare function loadDecoderSource(source: DecoderSource): Promise<UrlSource | BlobSource>;
|
|
@@ -1,9 +1,8 @@
|
|
|
1
1
|
import { BlobSource, UrlSource } from "mediabunny";
|
|
2
2
|
// only streamable sources
|
|
3
3
|
export async function loadDecoderSource(source) {
|
|
4
|
-
if (source instanceof
|
|
5
|
-
|
|
6
|
-
return new BlobSource(file);
|
|
4
|
+
if (source instanceof Blob) {
|
|
5
|
+
return new BlobSource(source);
|
|
7
6
|
}
|
|
8
7
|
else {
|
|
9
8
|
return new UrlSource(source);
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"load-decoder-source.js","sourceRoot":"","sources":["../../../s/driver/utils/load-decoder-source.ts"],"names":[],"mappings":"AAAA,OAAO,EAAC,UAAU,EAAE,SAAS,EAAC,MAAM,YAAY,CAAA;AAGhD,0BAA0B;AAC1B,MAAM,CAAC,KAAK,UAAU,iBAAiB,CAAC,MAAqB;IAC5D,IAAG,MAAM,YAAY,
|
|
1
|
+
{"version":3,"file":"load-decoder-source.js","sourceRoot":"","sources":["../../../s/driver/utils/load-decoder-source.ts"],"names":[],"mappings":"AAAA,OAAO,EAAC,UAAU,EAAE,SAAS,EAAC,MAAM,YAAY,CAAA;AAGhD,0BAA0B;AAC1B,MAAM,CAAC,KAAK,UAAU,iBAAiB,CAAC,MAAqB;IAC5D,IAAG,MAAM,YAAY,IAAI,EAAE,CAAC;QAC3B,OAAO,IAAI,UAAU,CAAC,MAAM,CAAC,CAAA;IAC9B,CAAC;SAAM,CAAC;QACP,OAAO,IAAI,SAAS,CAAC,MAAM,CAAC,CAAA;IAC7B,CAAC;AACF,CAAC"}
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"default-spec.js","sourceRoot":"","sources":["../../../../s/features/speech/transcribe/default-spec.ts"],"names":[],"mappings":"AAGA,MAAM,CAAC,MAAM,sBAAsB,GAAG,GAAoB,EAAE,CAAC,CAAC;IAC7D,KAAK,EAAE,yCAAyC;IAChD,KAAK,EAAE,IAAI;IACX,MAAM,EAAE,MAAM;IACd,WAAW,EAAE,EAAE;IACf,YAAY,EAAE,CAAC;CACf,CAAC,CAAA"}
|
|
@@ -0,0 +1,13 @@
|
|
|
1
|
+
import { pipeline } from "@huggingface/transformers";
|
|
2
|
+
export async function loadPipe(options) {
|
|
3
|
+
const { spec, onLoading } = options;
|
|
4
|
+
const pipe = await pipeline("automatic-speech-recognition", spec.model, {
|
|
5
|
+
device: spec.device,
|
|
6
|
+
dtype: spec.dtype,
|
|
7
|
+
progress_callback: (data) => {
|
|
8
|
+
onLoading({ total: data.total, progress: data.progress });
|
|
9
|
+
},
|
|
10
|
+
});
|
|
11
|
+
return pipe;
|
|
12
|
+
}
|
|
13
|
+
//# sourceMappingURL=load-pipe.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"load-pipe.js","sourceRoot":"","sources":["../../../../../s/features/speech/transcribe/parts/load-pipe.ts"],"names":[],"mappings":"AACA,OAAO,EAAC,QAAQ,EAAC,MAAM,2BAA2B,CAAA;AAIlD,MAAM,CAAC,KAAK,UAAU,QAAQ,CAAC,OAA+B;IAC7D,MAAM,EAAC,IAAI,EAAE,SAAS,EAAC,GAAG,OAAO,CAAA;IAEjC,MAAM,IAAI,GAAG,MAAM,QAAQ,CAAC,8BAA8B,EAAE,IAAI,CAAC,KAAK,EAAE;QACvE,MAAM,EAAE,IAAI,CAAC,MAAM;QACnB,KAAK,EAAE,IAAI,CAAC,KAAK;QACjB,iBAAiB,EAAE,CAAC,IAAS,EAAE,EAAE;YAChC,SAAS,CAAC,EAAC,KAAK,EAAE,IAAI,CAAC,KAAK,EAAE,QAAQ,EAAE,IAAI,CAAC,QAAQ,EAAC,CAAC,CAAA;QACxD,CAAC;KACD,CAAC,CAAA;IAEF,OAAO,IAAI,CAAA;AACZ,CAAC"}
|
|
@@ -0,0 +1,21 @@
|
|
|
1
|
+
export async function prepAudio(driver, source) {
|
|
2
|
+
const arrayBuffer = await source.arrayBuffer();
|
|
3
|
+
const audioCTX = new AudioContext({ sampleRate: 16000 });
|
|
4
|
+
const audioData = await audioCTX.decodeAudioData(arrayBuffer);
|
|
5
|
+
let audio;
|
|
6
|
+
if (audioData.numberOfChannels === 2) {
|
|
7
|
+
const SCALING_FACTOR = Math.sqrt(2);
|
|
8
|
+
const left = audioData.getChannelData(0);
|
|
9
|
+
const right = audioData.getChannelData(1);
|
|
10
|
+
audio = new Float32Array(left.length);
|
|
11
|
+
for (let i = 0; i < audioData.length; ++i) {
|
|
12
|
+
audio[i] = (SCALING_FACTOR * (left[i] + right[i])) / 2;
|
|
13
|
+
}
|
|
14
|
+
}
|
|
15
|
+
else {
|
|
16
|
+
audio = audioData.getChannelData(0);
|
|
17
|
+
}
|
|
18
|
+
const duration = await driver.getAudioDuration(source);
|
|
19
|
+
return { audio, duration };
|
|
20
|
+
}
|
|
21
|
+
//# sourceMappingURL=prep-audio.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"prep-audio.js","sourceRoot":"","sources":["../../../../../s/features/speech/transcribe/parts/prep-audio.ts"],"names":[],"mappings":"AAGA,MAAM,CAAC,KAAK,UAAU,SAAS,CAAC,MAAc,EAAE,MAAY;IAC3D,MAAM,WAAW,GAAG,MAAM,MAAM,CAAC,WAAW,EAAE,CAAA;IAC9C,MAAM,QAAQ,GAAG,IAAI,YAAY,CAAC,EAAC,UAAU,EAAE,KAAK,EAAC,CAAC,CAAA;IACtD,MAAM,SAAS,GAAG,MAAM,QAAQ,CAAC,eAAe,CAAC,WAAW,CAAC,CAAA;IAC7D,IAAI,KAAmB,CAAA;IACvB,IAAI,SAAS,CAAC,gBAAgB,KAAK,CAAC,EAAE,CAAC;QACtC,MAAM,cAAc,GAAG,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC,CAAA;QACnC,MAAM,IAAI,GAAG,SAAS,CAAC,cAAc,CAAC,CAAC,CAAC,CAAA;QACxC,MAAM,KAAK,GAAG,SAAS,CAAC,cAAc,CAAC,CAAC,CAAC,CAAA;QACzC,KAAK,GAAG,IAAI,YAAY,CAAC,IAAI,CAAC,MAAM,CAAC,CAAA;QACrC,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,SAAS,CAAC,MAAM,EAAE,EAAE,CAAC,EAAE,CAAC;YAC3C,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,cAAc,GAAG,CAAC,IAAI,CAAC,CAAC,CAAC,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAA;QACvD,CAAC;IACF,CAAC;SAAM,CAAC;QACP,KAAK,GAAG,SAAS,CAAC,cAAc,CAAC,CAAC,CAAC,CAAA;IACpC,CAAC;IACD,MAAM,QAAQ,GAAG,MAAM,MAAM,CAAC,gBAAgB,CAAC,MAAM,CAAC,CAAA;IACtD,OAAO,EAAC,KAAK,EAAE,QAAQ,EAAC,CAAA;AACzB,CAAC"}
|
|
@@ -0,0 +1,56 @@
|
|
|
1
|
+
import { WhisperTextStreamer } from "@huggingface/transformers";
|
|
2
|
+
export async function transcribe(options) {
|
|
3
|
+
const { pipe, spec, request, callbacks } = options;
|
|
4
|
+
if (!pipe.processor.feature_extractor)
|
|
5
|
+
throw new Error("no feature_extractor");
|
|
6
|
+
const timePrecision = (pipe.processor.feature_extractor?.config.chunk_length /
|
|
7
|
+
// @ts-ignore
|
|
8
|
+
pipe.model.config.max_source_positions);
|
|
9
|
+
let chunkCount = 0;
|
|
10
|
+
let startTime = null;
|
|
11
|
+
let tokenCount = 0;
|
|
12
|
+
let tokensPerSecond = 0;
|
|
13
|
+
const chunkDuration = spec.chunkLength - spec.strideLength;
|
|
14
|
+
const calculateProgress = () => {
|
|
15
|
+
const audioProgressSeconds = chunkCount * chunkDuration;
|
|
16
|
+
return Math.min(audioProgressSeconds / request.duration, 1);
|
|
17
|
+
};
|
|
18
|
+
// TODO type error on pipe.tokenizer
|
|
19
|
+
const tokenizer = pipe.tokenizer;
|
|
20
|
+
const streamer = new WhisperTextStreamer(tokenizer, {
|
|
21
|
+
time_precision: timePrecision,
|
|
22
|
+
token_callback_function: () => {
|
|
23
|
+
startTime ??= performance.now();
|
|
24
|
+
if (++tokenCount > 1) {
|
|
25
|
+
tokensPerSecond = (tokenCount / (performance.now() - startTime)) * 1000;
|
|
26
|
+
}
|
|
27
|
+
},
|
|
28
|
+
callback_function: (textChunk) => {
|
|
29
|
+
// TODO
|
|
30
|
+
callbacks.onTranscription(textChunk);
|
|
31
|
+
callbacks.onReport({ tokensPerSecond, progress: calculateProgress() });
|
|
32
|
+
},
|
|
33
|
+
on_finalize: () => {
|
|
34
|
+
startTime = null;
|
|
35
|
+
tokenCount = 0;
|
|
36
|
+
chunkCount++;
|
|
37
|
+
callbacks.onReport({ tokensPerSecond, progress: calculateProgress() });
|
|
38
|
+
},
|
|
39
|
+
});
|
|
40
|
+
const result = await pipe(new Float32Array(request.audio), {
|
|
41
|
+
top_k: 0,
|
|
42
|
+
do_sample: false,
|
|
43
|
+
chunk_length_s: spec.chunkLength,
|
|
44
|
+
stride_length_s: spec.strideLength,
|
|
45
|
+
language: request.language,
|
|
46
|
+
task: "transcribe",
|
|
47
|
+
return_timestamps: "word", // if using "word" the on_chunk_start & end is not called thus we cant retrieve timestamps, only after whole thing finishes
|
|
48
|
+
force_full_sequences: false,
|
|
49
|
+
streamer,
|
|
50
|
+
});
|
|
51
|
+
return {
|
|
52
|
+
text: result.text,
|
|
53
|
+
chunks: result.chunks
|
|
54
|
+
};
|
|
55
|
+
}
|
|
56
|
+
//# sourceMappingURL=transcribe.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"transcribe.js","sourceRoot":"","sources":["../../../../../s/features/speech/transcribe/parts/transcribe.ts"],"names":[],"mappings":"AACA,OAAO,EAAC,mBAAmB,EAAC,MAAM,2BAA2B,CAAA;AAG7D,MAAM,CAAC,KAAK,UAAU,UAAU,CAAC,OAA0B;IAC1D,MAAM,EAAC,IAAI,EAAE,IAAI,EAAE,OAAO,EAAE,SAAS,EAAC,GAAG,OAAO,CAAA;IAEhD,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,iBAAiB;QACpC,MAAM,IAAI,KAAK,CAAC,sBAAsB,CAAC,CAAA;IAExC,MAAM,aAAa,GAAG,CACrB,IAAI,CAAC,SAAS,CAAC,iBAAiB,EAAE,MAAM,CAAC,YAAY;QACrD,aAAa;QACb,IAAI,CAAC,KAAK,CAAC,MAAM,CAAC,oBAAoB,CACtC,CAAA;IAED,IAAI,UAAU,GAAG,CAAC,CAAA;IAClB,IAAI,SAAS,GAAkB,IAAI,CAAA;IACnC,IAAI,UAAU,GAAG,CAAC,CAAA;IAClB,IAAI,eAAe,GAAG,CAAC,CAAA;IAEvB,MAAM,aAAa,GAAG,IAAI,CAAC,WAAW,GAAG,IAAI,CAAC,YAAY,CAAA;IAE1D,MAAM,iBAAiB,GAAG,GAAG,EAAE;QAC9B,MAAM,oBAAoB,GAAG,UAAU,GAAG,aAAa,CAAA;QACvD,OAAO,IAAI,CAAC,GAAG,CAAC,oBAAoB,GAAG,OAAO,CAAC,QAAQ,EAAE,CAAC,CAAC,CAAA;IAC5D,CAAC,CAAA;IAED,oCAAoC;IACpC,MAAM,SAAS,GAAG,IAAI,CAAC,SAAgB,CAAA;IAEvC,MAAM,QAAQ,GAAG,IAAI,mBAAmB,CAAC,SAAS,EAAE;QACnD,cAAc,EAAE,aAAa;QAC7B,uBAAuB,EAAE,GAAG,EAAE;YAC7B,SAAS,KAAK,WAAW,CAAC,GAAG,EAAE,CAAA;YAC/B,IAAI,EAAE,UAAU,GAAG,CAAC,EAAE,CAAC;gBACtB,eAAe,GAAG,CAAC,UAAU,GAAG,CAAC,WAAW,CAAC,GAAG,EAAE,GAAG,SAAS,CAAC,CAAC,GAAG,IAAI,CAAA;YACxE,CAAC;QACF,CAAC;QACD,iBAAiB,EAAE,CAAC,SAAc,EAAE,EAAE;YACrC,OAAO;YACP,SAAS,CAAC,eAAe,CAAC,SAAS,CAAC,CAAA;YACpC,SAAS,CAAC,QAAQ,CAAC,EAAC,eAAe,EAAE,QAAQ,EAAE,iBAAiB,EAAE,EAAC,CAAC,CAAA;QACrE,CAAC;QACD,WAAW,EAAE,GAAG,EAAE;YACjB,SAAS,GAAG,IAAI,CAAA;YAChB,UAAU,GAAG,CAAC,CAAA;YACd,UAAU,EAAE,CAAA;YACZ,SAAS,CAAC,QAAQ,CAAC,EAAC,eAAe,EAAE,QAAQ,EAAE,iBAAiB,EAAE,EAAC,CAAC,CAAA;QACrE,CAAC;KACD,CAAC,CAAA;IAEF,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,IAAI,YAAY,CAAC,OAAO,CAAC,KAAK,CAAC,EAAE;QAC1D,KAAK,EAAE,CAAC;QACR,SAAS,EAAE,KAAK;QAChB,cAAc,EAAE,IAAI,CAAC,WAAW;QAChC,eAAe,EAAE,IAAI,CAAC,YAAY;QAClC,QAAQ,EAAE,OAAO,CAAC,QAAQ;QAC1B,IAAI,EAAE,YAAY;QAClB,iBAAiB,EAAE,MAAM,EAAE,2HAA2H;QACtJ,oBAAoB,EAAE,KAAK;QAC3B,QAAQ;KACR,CAAC,CAAA;IAEF,OAAO;QACN,IAAI,EAAE,MAAM,CAAC,IAAI;QACjB,MAAM,EAAE,MAAM,CAAC,MAAM;KACrB,CAAA;AACF,CAAC"}
|
|
@@ -0,0 +1,5 @@
|
|
|
1
|
+
import { TranscriberOptions, TranscriptionOptions } from "./types.js";
|
|
2
|
+
export declare function makeTranscriber({ driver, spec, workerUrl, onLoading }: TranscriberOptions): Promise<{
|
|
3
|
+
transcribe: (info: TranscriptionOptions) => Promise<import("./types.js").Transcription>;
|
|
4
|
+
dispose: void;
|
|
5
|
+
}>;
|
|
@@ -0,0 +1,33 @@
|
|
|
1
|
+
import { Comrade } from "@e280/comrade";
|
|
2
|
+
import { coalesce, queue, sub } from "@e280/stz";
|
|
3
|
+
import { prepAudio } from "./parts/prep-audio.js";
|
|
4
|
+
export async function makeTranscriber({ driver, spec, workerUrl, onLoading }) {
|
|
5
|
+
const onReport = sub();
|
|
6
|
+
const onTranscription = sub();
|
|
7
|
+
const thread = await Comrade.thread({
|
|
8
|
+
label: "OmnitoolSpeechTranscriber",
|
|
9
|
+
workerUrl,
|
|
10
|
+
setupHost: () => ({
|
|
11
|
+
loading: async (loading) => onLoading(loading),
|
|
12
|
+
deliverReport: async (report) => onReport.pub(report),
|
|
13
|
+
deliverTranscription: async (transcription) => onTranscription.pub(transcription),
|
|
14
|
+
}),
|
|
15
|
+
});
|
|
16
|
+
await thread.work.prepare(spec);
|
|
17
|
+
return {
|
|
18
|
+
transcribe: queue(async (info) => {
|
|
19
|
+
const { source, language } = info;
|
|
20
|
+
const { audio, duration } = await prepAudio(driver, source);
|
|
21
|
+
const detachCallbacks = coalesce(onReport(info.onReport), onTranscription(info.onTranscription));
|
|
22
|
+
const result = await thread.work.transcribe({
|
|
23
|
+
duration,
|
|
24
|
+
language,
|
|
25
|
+
audio: audio.buffer,
|
|
26
|
+
});
|
|
27
|
+
detachCallbacks();
|
|
28
|
+
return result;
|
|
29
|
+
}),
|
|
30
|
+
dispose: thread.terminate()
|
|
31
|
+
};
|
|
32
|
+
}
|
|
33
|
+
//# sourceMappingURL=transcriber.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"transcriber.js","sourceRoot":"","sources":["../../../../s/features/speech/transcribe/transcriber.ts"],"names":[],"mappings":"AACA,OAAO,EAAC,OAAO,EAAC,MAAM,eAAe,CAAA;AACrC,OAAO,EAAC,QAAQ,EAAE,KAAK,EAAE,GAAG,EAAC,MAAM,WAAW,CAAA;AAE9C,OAAO,EAAC,SAAS,EAAC,MAAM,uBAAuB,CAAA;AAG/C,MAAM,CAAC,KAAK,UAAU,eAAe,CAAC,EAAC,MAAM,EAAE,IAAI,EAAE,SAAS,EAAE,SAAS,EAAqB;IAC7F,MAAM,QAAQ,GAAG,GAAG,EAAiC,CAAA;IACrD,MAAM,eAAe,GAAG,GAAG,EAA2B,CAAA;IAEtD,MAAM,MAAM,GAAG,MAAM,OAAO,CAAC,MAAM,CAAuB;QACzD,KAAK,EAAE,2BAA2B;QAClC,SAAS;QACT,SAAS,EAAE,GAAG,EAAE,CAAC,CAAC;YACjB,OAAO,EAAE,KAAK,EAAC,OAAO,EAAC,EAAE,CAAC,SAAS,CAAC,OAAO,CAAC;YAC5C,aAAa,EAAE,KAAK,EAAC,MAAM,EAAC,EAAE,CAAC,QAAQ,CAAC,GAAG,CAAC,MAAM,CAAC;YACnD,oBAAoB,EAAE,KAAK,EAAC,aAAa,EAAC,EAAE,CAAC,eAAe,CAAC,GAAG,CAAC,aAAa,CAAC;SAC/E,CAAC;KACF,CAAC,CAAA;IAEF,MAAM,MAAM,CAAC,IAAI,CAAC,OAAO,CAAC,IAAI,CAAC,CAAA;IAE/B,OAAO;QACN,UAAU,EAAE,KAAK,CAAC,KAAK,EAAC,IAA0B,EAAE,EAAE;YACrD,MAAM,EAAC,MAAM,EAAE,QAAQ,EAAC,GAAG,IAAI,CAAA;YAC/B,MAAM,EAAC,KAAK,EAAE,QAAQ,EAAC,GAAG,MAAM,SAAS,CAAC,MAAM,EAAE,MAAM,CAAC,CAAA;YAEzD,MAAM,eAAe,GAAG,QAAQ,CAC/B,QAAQ,CAAC,IAAI,CAAC,QAAQ,CAAC,EACvB,eAAe,CAAC,IAAI,CAAC,eAAe,CAAC,CACrC,CAAA;YAED,MAAM,MAAM,GAAG,MAAM,MAAM,CAAC,IAAI,CAAC,UAAU,CAAC;gBAC3C,QAAQ;gBACR,QAAQ;gBACR,KAAK,EAAE,KAAK,CAAC,MAAM;aACnB,CAAC,CAAA;YAEF,eAAe,EAAE,CAAA;YACjB,OAAO,MAAM,CAAA;QACd,CAAC,CAAC;QACF,OAAO,EAAE,MAAM,CAAC,SAAS,EAAE;KAC3B,CAAA;AACF,CAAC"}
|
|
@@ -0,0 +1,66 @@
|
|
|
1
|
+
import { AsSchematic } from "@e280/comrade";
|
|
2
|
+
import { DataType, DeviceType, Pipeline } from "@huggingface/transformers";
|
|
3
|
+
import { Driver } from "../../../driver/driver.js";
|
|
4
|
+
export type TranscriberSchematic = AsSchematic<{
|
|
5
|
+
work: {
|
|
6
|
+
prepare(spec: TranscriberSpec): Promise<void>;
|
|
7
|
+
transcribe(request: TranscriptionRequest): Promise<Transcription>;
|
|
8
|
+
};
|
|
9
|
+
host: {
|
|
10
|
+
loading(load: Loading): Promise<void>;
|
|
11
|
+
deliverReport(report: TranscriptionReport): Promise<void>;
|
|
12
|
+
deliverTranscription(transcription: string): Promise<void>;
|
|
13
|
+
};
|
|
14
|
+
}>;
|
|
15
|
+
export type Loading = {
|
|
16
|
+
total: number;
|
|
17
|
+
progress: number;
|
|
18
|
+
};
|
|
19
|
+
export type TranscribeOptions = {
|
|
20
|
+
pipe: Pipeline;
|
|
21
|
+
spec: TranscriberSpec;
|
|
22
|
+
request: TranscriptionRequest;
|
|
23
|
+
callbacks: TranscriptionCallbacks;
|
|
24
|
+
};
|
|
25
|
+
export type TranscriberPipeOptions = {
|
|
26
|
+
spec: TranscriberSpec;
|
|
27
|
+
onLoading: (loading: Loading) => void;
|
|
28
|
+
};
|
|
29
|
+
export type SpeechTime = [start: number, end: number];
|
|
30
|
+
export type Transcription = {
|
|
31
|
+
text: string;
|
|
32
|
+
chunks: {
|
|
33
|
+
text: string;
|
|
34
|
+
timestamp: SpeechTime;
|
|
35
|
+
}[];
|
|
36
|
+
};
|
|
37
|
+
export type TranscriberSpec = {
|
|
38
|
+
model: string;
|
|
39
|
+
dtype: DataType;
|
|
40
|
+
device: DeviceType;
|
|
41
|
+
chunkLength: number;
|
|
42
|
+
strideLength: number;
|
|
43
|
+
};
|
|
44
|
+
export type TranscriptionOptions = {
|
|
45
|
+
source: Blob;
|
|
46
|
+
language: string | null;
|
|
47
|
+
} & TranscriptionCallbacks;
|
|
48
|
+
export type TranscriptionRequest = {
|
|
49
|
+
audio: ArrayBufferLike;
|
|
50
|
+
language: string | null;
|
|
51
|
+
duration: number;
|
|
52
|
+
};
|
|
53
|
+
export type TranscriptionReport = {
|
|
54
|
+
progress: number;
|
|
55
|
+
tokensPerSecond: number;
|
|
56
|
+
};
|
|
57
|
+
export type TranscriptionCallbacks = {
|
|
58
|
+
onReport: (report: TranscriptionReport) => void;
|
|
59
|
+
onTranscription: (transcription: string) => void;
|
|
60
|
+
};
|
|
61
|
+
export type TranscriberOptions = {
|
|
62
|
+
driver: Driver;
|
|
63
|
+
spec: TranscriberSpec;
|
|
64
|
+
workerUrl: URL | string;
|
|
65
|
+
onLoading: (loading: Loading) => void;
|
|
66
|
+
};
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"types.js","sourceRoot":"","sources":["../../../../s/features/speech/transcribe/types.ts"],"names":[],"mappings":""}
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
export {};
|
|
@@ -0,0 +1,33 @@
|
|
|
1
|
+
import { defer, once } from "@e280/stz";
|
|
2
|
+
import { Comrade } from "@e280/comrade";
|
|
3
|
+
import { loadPipe } from "./parts/load-pipe.js";
|
|
4
|
+
import { transcribe } from "./parts/transcribe.js";
|
|
5
|
+
const deferred = defer();
|
|
6
|
+
const makePrepare = (host) => once(async (spec) => {
|
|
7
|
+
deferred.resolve({
|
|
8
|
+
spec,
|
|
9
|
+
pipe: await loadPipe({
|
|
10
|
+
spec,
|
|
11
|
+
onLoading: loading => host.loading(loading),
|
|
12
|
+
}),
|
|
13
|
+
});
|
|
14
|
+
});
|
|
15
|
+
await Comrade.worker(shell => {
|
|
16
|
+
const prepare = makePrepare(shell.host);
|
|
17
|
+
return {
|
|
18
|
+
prepare,
|
|
19
|
+
async transcribe(request) {
|
|
20
|
+
const { pipe, spec } = await deferred.promise;
|
|
21
|
+
return transcribe({
|
|
22
|
+
pipe,
|
|
23
|
+
spec,
|
|
24
|
+
request,
|
|
25
|
+
callbacks: {
|
|
26
|
+
onReport: report => shell.host.deliverReport(report),
|
|
27
|
+
onTranscription: transcription => shell.host.deliverTranscription(transcription),
|
|
28
|
+
},
|
|
29
|
+
});
|
|
30
|
+
}
|
|
31
|
+
};
|
|
32
|
+
});
|
|
33
|
+
//# sourceMappingURL=worker.bundle.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"worker.bundle.js","sourceRoot":"","sources":["../../../../s/features/speech/transcribe/worker.bundle.ts"],"names":[],"mappings":"AACA,OAAO,EAAC,KAAK,EAAE,IAAI,EAAC,MAAM,WAAW,CAAA;AACrC,OAAO,EAAC,OAAO,EAAO,MAAM,eAAe,CAAA;AAG3C,OAAO,EAAC,QAAQ,EAAC,MAAM,sBAAsB,CAAA;AAC7C,OAAO,EAAC,UAAU,EAAC,MAAM,uBAAuB,CAAA;AAGhD,MAAM,QAAQ,GAAG,KAAK,EAA2C,CAAA;AAEjE,MAAM,WAAW,GAAG,CAAC,IAAgC,EAAE,EAAE,CAAC,IAAI,CAAC,KAAK,EAAC,IAAqB,EAAE,EAAE;IAC7F,QAAQ,CAAC,OAAO,CAAC;QAChB,IAAI;QACJ,IAAI,EAAE,MAAM,QAAQ,CAAC;YACpB,IAAI;YACJ,SAAS,EAAE,OAAO,CAAC,EAAE,CAAC,IAAI,CAAC,OAAO,CAAC,OAAO,CAAC;SAC3C,CAAC;KACF,CAAC,CAAA;AACH,CAAC,CAAC,CAAA;AAEF,MAAM,OAAO,CAAC,MAAM,CAAuB,KAAK,CAAC,EAAE;IAClD,MAAM,OAAO,GAAG,WAAW,CAAC,KAAK,CAAC,IAAI,CAAC,CAAA;IACvC,OAAO;QACN,OAAO;QACP,KAAK,CAAC,UAAU,CAAC,OAAO;YACvB,MAAM,EAAC,IAAI,EAAE,IAAI,EAAC,GAAG,MAAM,QAAQ,CAAC,OAAO,CAAA;YAC3C,OAAO,UAAU,CAAC;gBACjB,IAAI;gBACJ,IAAI;gBACJ,OAAO;gBACP,SAAS,EAAE;oBACV,QAAQ,EAAE,MAAM,CAAC,EAAE,CAAC,KAAK,CAAC,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC;oBACpD,eAAe,EAAE,aAAa,CAAC,EAAE,CAAC,KAAK,CAAC,IAAI,CAAC,oBAAoB,CAAC,aAAa,CAAC;iBAChF;aACD,CAAC,CAAA;QACH,CAAC;KACD,CAAA;AACF,CAAC,CAAC,CAAA"}
|