Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

feat: add tdrzEnable option #264

Merged
merged 5 commits into from
Nov 2, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 14 additions & 1 deletion android/src/main/java/com/rnwhisper/WhisperContext.java
Original file line number Diff line number Diff line change
Expand Up @@ -53,6 +53,7 @@ public class WhisperContext {
private boolean isCapturing = false;
private boolean isStoppedByAction = false;
private boolean isTranscribing = false;
private boolean isTdrzEnable = false;
private Thread rootFullHandler = null;
private Thread fullHandler = null;

Expand All @@ -73,6 +74,7 @@ private void rewind() {
isCapturing = false;
isStoppedByAction = false;
isTranscribing = false;
isTdrzEnable = false;
rootFullHandler = null;
fullHandler = null;
}
Expand Down Expand Up @@ -113,6 +115,8 @@ public int startRealtimeTranscribe(int jobId, ReadableMap options) {
double realtimeAudioMinSec = options.hasKey("realtimeAudioMinSec") ? options.getDouble("realtimeAudioMinSec") : 0;
final double audioMinSec = realtimeAudioMinSec > 0.5 && realtimeAudioMinSec <= audioSliceSec ? realtimeAudioMinSec : 1;

this.isTdrzEnable = options.hasKey("tdrzEnable") && options.getBoolean("tdrzEnable");

createRealtimeTranscribeJob(jobId, context, options);

sliceNSamples = new ArrayList<Integer>();
Expand Down Expand Up @@ -333,8 +337,9 @@ public WritableMap transcribeInputStream(int jobId, InputStream inputStream, Rea
throw new Exception("Context is already in capturing or transcribing");
}
rewind();

this.jobId = jobId;
this.isTdrzEnable = options.hasKey("tdrzEnable") && options.getBoolean("tdrzEnable");

isTranscribing = true;
float[] audioData = AudioUtils.decodeWaveFile(inputStream);

Expand Down Expand Up @@ -368,8 +373,15 @@ private WritableMap getTextSegments(int start, int count) {

WritableMap data = Arguments.createMap();
WritableArray segments = Arguments.createArray();

for (int i = 0; i < count; i++) {
String text = getTextSegment(context, i);

// If tdrzEnable is enabled and speaker turn is detected
if (this.isTdrzEnable && getTextSegmentSpeakerTurnNext(context, i)) {
text += " [SPEAKER_TURN]";
}

builder.append(text);

WritableMap segment = Arguments.createMap();
Expand Down Expand Up @@ -499,6 +511,7 @@ protected static native int fullWithNewJob(
protected static native String getTextSegment(long context, int index);
protected static native int getTextSegmentT0(long context, int index);
protected static native int getTextSegmentT1(long context, int index);
protected static native boolean getTextSegmentSpeakerTurnNext(long context, int index);

protected static native void createRealtimeTranscribeJob(
int job_id,
Expand Down
10 changes: 10 additions & 0 deletions android/src/main/jni.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -208,6 +208,7 @@ struct whisper_full_params createFullParams(JNIEnv *env, jobject options) {
params.translate = readablemap::getBool(env, options, "translate", false);
params.speed_up = readablemap::getBool(env, options, "speedUp", false);
params.token_timestamps = readablemap::getBool(env, options, "tokenTimestamps", false);
params.tdrz_enable = readablemap::getBool(env, options, "tdrzEnable", false);
params.offset_ms = 0;
params.no_context = true;
params.single_segment = false;
Expand Down Expand Up @@ -493,4 +494,13 @@ Java_com_rnwhisper_WhisperContext_freeContext(
whisper_free(context);
}

JNIEXPORT jboolean JNICALL
Java_com_rnwhisper_WhisperContext_getTextSegmentSpeakerTurnNext(
JNIEnv *env, jobject thiz, jlong context_ptr, jint index) {
UNUSED(env);
UNUSED(thiz);
struct whisper_context *context = reinterpret_cast<struct whisper_context *>(context_ptr);
return whisper_full_get_segment_speaker_turn_next(context, index);
}

} // extern "C"
37 changes: 19 additions & 18 deletions docs/API/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -58,7 +58,7 @@ whisper.rn

#### Defined in

[index.ts:76](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L76)
[index.ts:76](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L76)

___

Expand All @@ -80,7 +80,7 @@ ___

#### Defined in

[index.ts:441](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L441)
[index.ts:441](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L441)

___

Expand All @@ -90,7 +90,7 @@ ___

#### Defined in

[index.ts:59](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L59)
[index.ts:59](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L59)

___

Expand All @@ -108,7 +108,7 @@ ___

#### Defined in

[index.ts:52](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L52)
[index.ts:52](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L52)

___

Expand All @@ -127,7 +127,7 @@ ___

#### Defined in

[index.ts:45](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L45)
[index.ts:45](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L45)

___

Expand All @@ -149,6 +149,7 @@ ___
| `offset?` | `number` | Time offset in milliseconds |
| `prompt?` | `string` | Initial Prompt |
| `speedUp?` | `boolean` | Speed up audio by x2 (reduced accuracy) |
| `tdrzEnable?` | `boolean` | Enable tinydiarize (requires a tdrz model) |
| `temperature?` | `number` | Tnitial decoding temperature |
| `temperatureInc?` | `number` | - |
| `tokenTimestamps?` | `boolean` | Enable token-level timestamps |
Expand All @@ -157,7 +158,7 @@ ___

#### Defined in

[NativeRNWhisper.ts:5](https://github.com/mybigday/whisper.rn/blob/85066fc/src/NativeRNWhisper.ts#L5)
[NativeRNWhisper.ts:5](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/NativeRNWhisper.ts#L5)

___

Expand All @@ -175,7 +176,7 @@ ___

#### Defined in

[index.ts:70](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L70)
[index.ts:70](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L70)

___

Expand All @@ -200,7 +201,7 @@ ___

#### Defined in

[index.ts:138](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L138)
[index.ts:138](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L138)

___

Expand All @@ -218,7 +219,7 @@ ___

#### Defined in

[index.ts:171](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L171)
[index.ts:171](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L171)

___

Expand All @@ -242,7 +243,7 @@ ___

#### Defined in

[index.ts:158](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L158)
[index.ts:158](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L158)

___

Expand All @@ -252,7 +253,7 @@ ___

#### Defined in

[index.ts:84](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L84)
[index.ts:84](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L84)

___

Expand All @@ -270,7 +271,7 @@ ___

#### Defined in

[NativeRNWhisper.ts:37](https://github.com/mybigday/whisper.rn/blob/85066fc/src/NativeRNWhisper.ts#L37)
[NativeRNWhisper.ts:39](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/NativeRNWhisper.ts#L39)

## Variables

Expand All @@ -295,7 +296,7 @@ AudioSession Utility, iOS only.

#### Defined in

[AudioSessionIos.ts:50](https://github.com/mybigday/whisper.rn/blob/85066fc/src/AudioSessionIos.ts#L50)
[AudioSessionIos.ts:50](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/AudioSessionIos.ts#L50)

___

Expand All @@ -307,7 +308,7 @@ Is allow fallback to CPU if load CoreML model failed

#### Defined in

[index.ts:543](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L543)
[index.ts:543](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L543)

___

Expand All @@ -319,7 +320,7 @@ Is use CoreML models on iOS

#### Defined in

[index.ts:540](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L540)
[index.ts:540](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L540)

___

Expand All @@ -331,7 +332,7 @@ Current version of whisper.cpp

#### Defined in

[index.ts:535](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L535)
[index.ts:535](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L535)

## Functions

Expand All @@ -351,7 +352,7 @@ Current version of whisper.cpp

#### Defined in

[index.ts:467](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L467)
[index.ts:467](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L467)

___

Expand All @@ -365,4 +366,4 @@ ___

#### Defined in

[index.ts:530](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L530)
[index.ts:530](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L530)
14 changes: 7 additions & 7 deletions docs/API/classes/WhisperContext.md
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,7 @@

#### Defined in

[index.ts:195](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L195)
[index.ts:195](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L195)

## Properties

Expand All @@ -44,7 +44,7 @@

#### Defined in

[index.ts:191](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L191)
[index.ts:191](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L191)

___

Expand All @@ -54,7 +54,7 @@ ___

#### Defined in

[index.ts:189](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L189)
[index.ts:189](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L189)

___

Expand All @@ -64,7 +64,7 @@ ___

#### Defined in

[index.ts:193](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L193)
[index.ts:193](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L193)

## Methods

Expand All @@ -78,7 +78,7 @@ ___

#### Defined in

[index.ts:436](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L436)
[index.ts:436](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L436)

___

Expand Down Expand Up @@ -106,7 +106,7 @@ Transcribe audio file

#### Defined in

[index.ts:206](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L206)
[index.ts:206](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L206)

___

Expand All @@ -128,4 +128,4 @@ Transcribe the microphone audio stream, the microphone user permission is requir

#### Defined in

[index.ts:302](https://github.com/mybigday/whisper.rn/blob/85066fc/src/index.ts#L302)
[index.ts:302](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/index.ts#L302)
12 changes: 6 additions & 6 deletions docs/API/enums/AudioSessionCategoryIos.md
Original file line number Diff line number Diff line change
Expand Up @@ -25,7 +25,7 @@ https://developer.apple.com/documentation/avfaudio/avaudiosessioncategory?langua

#### Defined in

[AudioSessionIos.ts:8](https://github.com/mybigday/whisper.rn/blob/85066fc/src/AudioSessionIos.ts#L8)
[AudioSessionIos.ts:8](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/AudioSessionIos.ts#L8)

___

Expand All @@ -35,7 +35,7 @@ ___

#### Defined in

[AudioSessionIos.ts:13](https://github.com/mybigday/whisper.rn/blob/85066fc/src/AudioSessionIos.ts#L13)
[AudioSessionIos.ts:13](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/AudioSessionIos.ts#L13)

___

Expand All @@ -45,7 +45,7 @@ ___

#### Defined in

[AudioSessionIos.ts:12](https://github.com/mybigday/whisper.rn/blob/85066fc/src/AudioSessionIos.ts#L12)
[AudioSessionIos.ts:12](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/AudioSessionIos.ts#L12)

___

Expand All @@ -55,7 +55,7 @@ ___

#### Defined in

[AudioSessionIos.ts:10](https://github.com/mybigday/whisper.rn/blob/85066fc/src/AudioSessionIos.ts#L10)
[AudioSessionIos.ts:10](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/AudioSessionIos.ts#L10)

___

Expand All @@ -65,7 +65,7 @@ ___

#### Defined in

[AudioSessionIos.ts:11](https://github.com/mybigday/whisper.rn/blob/85066fc/src/AudioSessionIos.ts#L11)
[AudioSessionIos.ts:11](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/AudioSessionIos.ts#L11)

___

Expand All @@ -75,4 +75,4 @@ ___

#### Defined in

[AudioSessionIos.ts:9](https://github.com/mybigday/whisper.rn/blob/85066fc/src/AudioSessionIos.ts#L9)
[AudioSessionIos.ts:9](https://github.com/mybigday/whisper.rn/blob/8f61e46/src/AudioSessionIos.ts#L9)
Loading
Loading