Android: Add support for Voice Typing for most languages (#8309)

2025-02-19 20:00:20 +02:00 · 2023-06-13 18:06:54 +01:00 · 2023-06-13 18:06:54 +01:00 · ce7da882ee
commit ce7da882ee
parent 92c68882e2
26 changed files with 228 additions and 299 deletions
--- a/.yarn/patches/react-native-vosk-npm-0.1.12-76b1caaae8.patch
+++ b/.yarn/patches/react-native-vosk-npm-0.1.12-76b1caaae8.patch
@ -36,7 +36,7 @@ index 6afcbbf0cc8ca2d69dd78077d61e59a90b2136bb..9f8d72b4ec5b2b3d290975d6a255917c
 def kotlin_version = getExtOrDefault('kotlinVersion')
 
 diff --git a/android/src/main/java/com/reactnativevosk/VoskModule.kt b/android/src/main/java/com/reactnativevosk/VoskModule.kt
-index 0e2b6595b1b2cf1ee01c6c64239c4b0ea37fce19..f3da440bc2863a59db6d2d1691c54d8d4870cb3f 100644
+index 0e2b6595b1b2cf1ee01c6c64239c4b0ea37fce19..5a8539b9cce8951967640dba755e29a4e3ff404a 100644
 --- a/android/src/main/java/com/reactnativevosk/VoskModule.kt
 +++ b/android/src/main/java/com/reactnativevosk/VoskModule.kt
@@ -19,13 +19,25 @@ class VoskModule(reactContext: ReactApplicationContext) : ReactContextBaseJavaMo
@ -66,7 +66,25 @@ index 0e2b6595b1b2cf1ee01c6c64239c4b0ea37fce19..f3da440bc2863a59db6d2d1691c54d8d
       sendEvent("onResult", text)
     }
   }
-@@ -153,6 +165,25 @@ class VoskModule(reactContext: ReactApplicationContext) : ReactContextBaseJavaMo
+@@ -93,12 +105,11 @@ class VoskModule(reactContext: ReactApplicationContext) : ReactContextBaseJavaMo
+   @ReactMethod
+   fun loadModel(path: String, promise: Promise) {
+     cleanModel();
+-    StorageService.unpack(context, path, "models",
+-      { model: Model? ->
+-        this.model = model
+-        promise.resolve("Model successfully loaded")
+-      }
+-    ) { e: IOException ->
+    
+    try {
+      this.model = Model(path);
+      promise.resolve("Model successfully loaded")
+    } catch (e: IOException) {
+       this.model = null
+       promise.reject(e)
+     }
+@@ -153,6 +164,25 @@ class VoskModule(reactContext: ReactApplicationContext) : ReactContextBaseJavaMo
     cleanRecognizer();
   }
 
--- a/packages/app-mobile/android/app/src/main/assets/model-fr-fr/README
+++ b/packages/app-mobile/android/app/src/main/assets/model-fr-fr/README
@ -1,7 +0,0 @@
-French small model for Vosk
-
-WER
-
-%WER 23.95 [ 37203 / 155330, 5373 ins, 4427 del, 27403 sub ] exp/chain_a/tdnn/decode_test_cv/wer_12_0.0
-%WER 19.30 [ 2975 / 15412, 683 ins, 672 del, 1620 sub ] exp/chain_a/tdnn/decode_test_mtedx/wer_10_0.0
-%WER 27.25 [ 20208 / 74145, 2647 ins, 5852 del, 11709 sub ] exp/chain_a/tdnn/decode_test_podcast_reseg/wer_10_0.0
--- a/packages/app-mobile/android/app/src/main/assets/model-fr-fr/am/final.mdl
+++ b/packages/app-mobile/android/app/src/main/assets/model-fr-fr/am/final.mdl
--- a/packages/app-mobile/android/app/src/main/assets/model-fr-fr/conf/mfcc.conf
+++ b/packages/app-mobile/android/app/src/main/assets/model-fr-fr/conf/mfcc.conf
@ -1,8 +0,0 @@
--use-energy=false
--sample-frequency=16000
--num-mel-bins=40
--num-ceps=40
--low-freq=40
--high-freq=-200
--allow-upsample=true
--allow-downsample=true
--- a/packages/app-mobile/android/app/src/main/assets/model-fr-fr/conf/model.conf
+++ b/packages/app-mobile/android/app/src/main/assets/model-fr-fr/conf/model.conf
@ -1,10 +0,0 @@
--min-active=200
--max-active=7000
--beam=13.0
--lattice-beam=4.0
--acoustic-scale=1.0
--frame-subsampling-factor=3
--endpoint.silence-phones=1:2:3:4:5:6:7:8:9:10
--endpoint.rule2.min-trailing-silence=0.5
--endpoint.rule3.min-trailing-silence=1.0
--endpoint.rule4.min-trailing-silence=2.0
--- a/packages/app-mobile/android/app/src/main/assets/model-fr-fr/graph/Gr.fst
+++ b/packages/app-mobile/android/app/src/main/assets/model-fr-fr/graph/Gr.fst
--- a/packages/app-mobile/android/app/src/main/assets/model-fr-fr/graph/HCLr.fst
+++ b/packages/app-mobile/android/app/src/main/assets/model-fr-fr/graph/HCLr.fst
--- a/packages/app-mobile/android/app/src/main/assets/model-fr-fr/graph/disambig_tid.int
+++ b/packages/app-mobile/android/app/src/main/assets/model-fr-fr/graph/disambig_tid.int
@ -1,76 +0,0 @@
-9365
-9366
-9367
-9368
-9369
-9370
-9371
-9372
-9373
-9374
-9375
-9376
-9377
-9378
-9379
-9380
-9381
-9382
-9383
-9384
-9385
-9386
-9387
-9388
-9389
-9390
-9391
-9392
-9393
-9394
-9395
-9396
-9397
-9398
-9399
-9400
-9401
-9402
-9403
-9404
-9405
-9406
-9407
-9408
-9409
-9410
-9411
-9412
-9413
-9414
-9415
-9416
-9417
-9418
-9419
-9420
-9421
-9422
-9423
-9424
-9425
-9426
-9427
-9428
-9429
-9430
-9431
-9432
-9433
-9434
-9435
-9436
-9437
-9438
-9439
-9440
--- a/packages/app-mobile/android/app/src/main/assets/model-fr-fr/graph/phones/word_boundary.int
+++ b/packages/app-mobile/android/app/src/main/assets/model-fr-fr/graph/phones/word_boundary.int
@ -1,154 +0,0 @@
-1 nonword
-2 begin
-3 end
-4 internal
-5 singleton
-6 nonword
-7 begin
-8 end
-9 internal
-10 singleton
-11 begin
-12 end
-13 internal
-14 singleton
-15 begin
-16 end
-17 internal
-18 singleton
-19 begin
-20 end
-21 internal
-22 singleton
-23 begin
-24 end
-25 internal
-26 singleton
-27 begin
-28 end
-29 internal
-30 singleton
-31 begin
-32 end
-33 internal
-34 singleton
-35 begin
-36 end
-37 internal
-38 singleton
-39 begin
-40 end
-41 internal
-42 singleton
-43 begin
-44 end
-45 internal
-46 singleton
-47 begin
-48 end
-49 internal
-50 singleton
-51 begin
-52 end
-53 internal
-54 singleton
-55 begin
-56 end
-57 internal
-58 singleton
-59 begin
-60 end
-61 internal
-62 singleton
-63 begin
-64 end
-65 internal
-66 singleton
-67 begin
-68 end
-69 internal
-70 singleton
-71 begin
-72 end
-73 internal
-74 singleton
-75 begin
-76 end
-77 internal
-78 singleton
-79 begin
-80 end
-81 internal
-82 singleton
-83 begin
-84 end
-85 internal
-86 singleton
-87 begin
-88 end
-89 internal
-90 singleton
-91 begin
-92 end
-93 internal
-94 singleton
-95 begin
-96 end
-97 internal
-98 singleton
-99 begin
-100 end
-101 internal
-102 singleton
-103 begin
-104 end
-105 internal
-106 singleton
-107 begin
-108 end
-109 internal
-110 singleton
-111 begin
-112 end
-113 internal
-114 singleton
-115 begin
-116 end
-117 internal
-118 singleton
-119 begin
-120 end
-121 internal
-122 singleton
-123 begin
-124 end
-125 internal
-126 singleton
-127 begin
-128 end
-129 internal
-130 singleton
-131 begin
-132 end
-133 internal
-134 singleton
-135 begin
-136 end
-137 internal
-138 singleton
-139 begin
-140 end
-141 internal
-142 singleton
-143 begin
-144 end
-145 internal
-146 singleton
-147 begin
-148 end
-149 internal
-150 singleton
-151 begin
-152 end
-153 internal
-154 singleton
--- a/packages/app-mobile/android/app/src/main/assets/model-fr-fr/ivector/final.dubm
+++ b/packages/app-mobile/android/app/src/main/assets/model-fr-fr/ivector/final.dubm
--- a/packages/app-mobile/android/app/src/main/assets/model-fr-fr/ivector/final.ie
+++ b/packages/app-mobile/android/app/src/main/assets/model-fr-fr/ivector/final.ie
--- a/packages/app-mobile/android/app/src/main/assets/model-fr-fr/ivector/final.mat
+++ b/packages/app-mobile/android/app/src/main/assets/model-fr-fr/ivector/final.mat
--- a/packages/app-mobile/android/app/src/main/assets/model-fr-fr/ivector/global_cmvn.stats
+++ b/packages/app-mobile/android/app/src/main/assets/model-fr-fr/ivector/global_cmvn.stats
@ -1,3 +0,0 @@
- [
-  1.022245e+11 -6.33291e+09 -2.480997e+09 8.290258e+09 -9.084483e+09 -8.092173e+09 -1.4735e+10 -7.041795e+09 -1.171205e+10 -2.976464e+08 -1.009425e+10 -6765179 -7.821326e+09 1.449499e+09 -6.413975e+09 -5.303802e+08 -4.998635e+09 9.521598e+07 -3.073041e+09 1.56756e+08 -1.287956e+09 1.738752e+08 -2.382392e+08 -2.716675e+07 4.404485e+08 -1.913359e+08 7.780919e+08 -4.006922e+08 7.895809e+08 -5.401082e+08 5.17605e+08 -6.227134e+08 6.58271e+08 -6.204593e+07 5.187754e+08 -4.497048e+08 4.219366e+07 -2.78742e+08 -1.797385e+07 -3.604475e+07 1.053647e+09 
-  1.040194e+13 6.245521e+11 4.223293e+11 6.831219e+11 6.078478e+11 6.3425e+11 7.943839e+11 6.013323e+11 6.781652e+11 5.272091e+11 5.810814e+11 4.353831e+11 4.473305e+11 3.42063e+11 3.083377e+11 2.14257e+11 1.892057e+11 1.163827e+11 8.367058e+10 4.203224e+10 2.297476e+10 7.596307e+09 1.099877e+09 2.886651e+08 3.797438e+09 9.372847e+09 1.629059e+10 2.196351e+10 2.747149e+10 3.072878e+10 3.238528e+10 3.330232e+10 3.407238e+10 3.230687e+10 2.676914e+10 2.252055e+10 1.914305e+10 1.565974e+10 1.224627e+10 8.415393e+09 0 ]
--- a/packages/app-mobile/android/app/src/main/assets/model-fr-fr/ivector/online_cmvn.conf
+++ b/packages/app-mobile/android/app/src/main/assets/model-fr-fr/ivector/online_cmvn.conf
--- a/packages/app-mobile/android/app/src/main/assets/model-fr-fr/ivector/splice.conf
+++ b/packages/app-mobile/android/app/src/main/assets/model-fr-fr/ivector/splice.conf
@ -1,2 +0,0 @@
--left-context=3
--right-context=3
--- a/packages/app-mobile/android/app/src/main/assets/model-fr-fr/uuid
+++ b/packages/app-mobile/android/app/src/main/assets/model-fr-fr/uuid
@ -1 +0,0 @@
-1b7180e6-e500-4818-adc8-a41fe97a84ce
--- a/packages/app-mobile/components/screens/Note.tsx
+++ b/packages/app-mobile/components/screens/Note.tsx
@ -46,6 +46,7 @@ import { NoteEntity } from '@joplin/lib/services/database/types';
 import Logger from '@joplin/lib/Logger';
 import VoiceTypingDialog from '../voiceTyping/VoiceTypingDialog';
 import { voskEnabled } from '../../services/voiceTyping/vosk';
+import { isSupportedLanguage } from '../../services/voiceTyping/vosk.android';
 const urlUtils = require('@joplin/lib/urlUtils');

 // import Vosk from 'react-native-vosk';
@ -1000,7 +1001,7 @@ class NoteScreenComponent extends BaseScreenComponent {
 		});

 		// Voice typing is enabled only for French language and on Android for now
-		if (voskEnabled && shim.mobilePlatform() === 'android' && currentLocale() === 'fr_FR') {
+		if (voskEnabled && shim.mobilePlatform() === 'android' && isSupportedLanguage(currentLocale())) {
 			output.push({
 				title: _('Voice typing...'),
 				onPress: () => {
@ -1307,7 +1308,7 @@ class NoteScreenComponent extends BaseScreenComponent {

 		const renderVoiceTypingDialog = () => {
 			if (!this.state.voiceTypingDialogShown) return null;
-			return <VoiceTypingDialog onText={this.voiceTypingDialog_onText} onDismiss={this.voiceTypingDialog_onDismiss}/>;
+			return <VoiceTypingDialog locale={currentLocale()} onText={this.voiceTypingDialog_onText} onDismiss={this.voiceTypingDialog_onDismiss}/>;
 		};

 		return (
--- a/packages/app-mobile/components/voiceTyping/VoiceTypingDialog.tsx
+++ b/packages/app-mobile/components/voiceTyping/VoiceTypingDialog.tsx
@ -1,12 +1,14 @@
 import * as React from 'react';
 import { useState, useEffect, useCallback } from 'react';
-import { Banner, ActivityIndicator } from 'react-native-paper';
-import { _ } from '@joplin/lib/locale';
+import { Banner, ActivityIndicator, Modal } from 'react-native-paper';
+import { _, languageName } from '@joplin/lib/locale';
 import useAsyncEffect, { AsyncEffectEvent } from '@joplin/lib/hooks/useAsyncEffect';
 import { getVosk, Recorder, startRecording, Vosk } from '../../services/voiceTyping/vosk';
 import { IconSource } from 'react-native-paper/lib/typescript/src/components/Icon';
+import { modelIsDownloaded } from '../../services/voiceTyping/vosk.android';

 interface Props {
+	locale: string;
 	onDismiss: ()=> void;
 	onText: (text: string)=> void;
 }
@ -16,30 +18,39 @@ enum RecorderState {
 	Recording = 2,
 	Processing = 3,
 	Error = 4,
+	Downloading = 5,
 }

-const useVosk = (): [Error | null, Vosk|null] => {
+const useVosk = (locale: string): [Error | null, boolean, Vosk|null] => {
 	const [vosk, setVosk] = useState<Vosk>(null);
 	const [error, setError] = useState<Error>(null);
+	const [mustDownloadModel, setMustDownloadModel] = useState<boolean | null>(null);

 	useAsyncEffect(async (event: AsyncEffectEvent) => {
+		if (mustDownloadModel === null) return;
+
 		try {
-			const v = await getVosk();
+			const v = await getVosk(locale);
 			if (event.cancelled) return;
 			setVosk(v);
 		} catch (error) {
 			setError(error);
+		} finally {
+			setMustDownloadModel(false);
 		}
-	}, []);
+	}, [locale, mustDownloadModel]);

-	return [error, vosk];
+	useAsyncEffect(async (_event: AsyncEffectEvent) => {
+		setMustDownloadModel(!(await modelIsDownloaded(locale)));
+	}, [locale]);
+
+	return [error, mustDownloadModel, vosk];
 };

 export default (props: Props) => {
 	const [recorder, setRecorder] = useState<Recorder>(null);
 	const [recorderState, setRecorderState] = useState<RecorderState>(RecorderState.Loading);
-
-	const [voskError, vosk] = useVosk();
+	const [voskError, mustDownloadModel, vosk] = useVosk(props.locale);

 	useEffect(() => {
 		if (voskError) {
@ -49,6 +60,12 @@ export default (props: Props) => {
 		}
 	}, [vosk, voskError]);

+	useEffect(() => {
+		if (mustDownloadModel) {
+			setRecorderState(RecorderState.Downloading);
+		}
+	}, [mustDownloadModel]);
+
 	useEffect(() => {
 		if (recorderState === RecorderState.Recording) {
 			setRecorder(startRecording(vosk, {
@ -60,7 +77,7 @@ export default (props: Props) => {
 	}, [recorderState, vosk, props.onText]);

 	const onDismiss = useCallback(() => {
-		recorder.cleanup();
+		if (recorder) recorder.cleanup();
 		props.onDismiss();
 	}, [recorder, props.onDismiss]);

@ -69,6 +86,7 @@ export default (props: Props) => {
 			[RecorderState.Loading]: () => _('Loading...'),
 			[RecorderState.Recording]: () => _('Please record your voice...'),
 			[RecorderState.Processing]: () => _('Converting speech to text...'),
+			[RecorderState.Downloading]: () => _('Downloading %s language files...', languageName(props.locale)),
 			[RecorderState.Error]: () => _('Error: %s', voskError.message),
 		};

@ -80,6 +98,7 @@ export default (props: Props) => {
 			[RecorderState.Loading]: ({ size }: { size: number }) => <ActivityIndicator animating={true} style={{ width: size, height: size }} />,
 			[RecorderState.Recording]: 'microphone',
 			[RecorderState.Processing]: 'microphone',
+			[RecorderState.Downloading]: ({ size }: { size: number }) => <ActivityIndicator animating={true} style={{ width: size, height: size }} />,
 			[RecorderState.Error]: 'alert-circle-outline',
 		};

@ -87,16 +106,18 @@ export default (props: Props) => {
 	};

 	return (
-		<Banner
-			visible={true}
-			icon={renderIcon()}
-			actions={[
-				{
-					label: _('Done'),
-					onPress: onDismiss,
-				},
-			]}>
-			{`${_('Voice typing...')}\n${renderContent()}`}
-		</Banner>
+		<Modal visible={true} style={{ display: 'flex', flexDirection: 'column', justifyContent: 'flex-end' }}>
+			<Banner
+				visible={true}
+				icon={renderIcon()}
+				actions={[
+					{
+						label: _('Done'),
+						onPress: onDismiss,
+					},
+				]}>
+				{`${_('Voice typing...')}\n${renderContent()}`}
+			</Banner>
+		</Modal>
 	);
 };
--- a/packages/app-mobile/ios/Podfile.lock
+++ b/packages/app-mobile/ios/Podfile.lock
@ -483,7 +483,15 @@ PODS:
    - React-Core
  - RNVectorIcons (9.2.0):
    - React-Core
+  - RNZipArchive (6.0.9):
+    - React-Core
+    - RNZipArchive/Core (= 6.0.9)
+    - SSZipArchive (~> 2.2)
+  - RNZipArchive/Core (6.0.9):
+    - React-Core
+    - SSZipArchive (~> 2.2)
  - SocketRocket (0.6.0)
+  - SSZipArchive (2.4.3)
  - Yoga (1.14.0)
  - YogaKit (1.18.1):
    - Yoga (~> 1.14)
@ -577,6 +585,7 @@ DEPENDENCIES:
  - RNSecureRandom (from `../node_modules/react-native-securerandom`)
  - RNShare (from `../node_modules/react-native-share`)
  - RNVectorIcons (from `../node_modules/react-native-vector-icons`)
+  - RNZipArchive (from `../node_modules/react-native-zip-archive`)
  - Yoga (from `../node_modules/react-native/ReactCommon/yoga`)

 SPEC REPOS:
@ -595,6 +604,7 @@ SPEC REPOS:
    - libevent
    - OpenSSL-Universal
    - SocketRocket
+    - SSZipArchive
    - YogaKit

 EXTERNAL SOURCES:
@ -724,6 +734,8 @@ EXTERNAL SOURCES:
    :path: "../node_modules/react-native-share"
  RNVectorIcons:
    :path: "../node_modules/react-native-vector-icons"
+  RNZipArchive:
+    :path: "../node_modules/react-native-zip-archive"
  Yoga:
    :path: "../node_modules/react-native/ReactCommon/yoga"

@ -804,7 +816,9 @@ SPEC CHECKSUMS:
  RNSecureRandom: 07efbdf2cd99efe13497433668e54acd7df49fef
  RNShare: d82e10f6b7677f4b0048c23709bd04098d5aee6c
  RNVectorIcons: fcc2f6cb32f5735b586e66d14103a74ce6ad61f8
+  RNZipArchive: 68a0c6db4b1c103f846f1559622050df254a3ade
  SocketRocket: fccef3f9c5cedea1353a9ef6ada904fde10d6608
+  SSZipArchive: fe6a26b2a54d5a0890f2567b5cc6de5caa600aef
  Yoga: e7ea9e590e27460d28911403b894722354d73479
  YogaKit: f782866e155069a2cca2517aafea43200b01fd5a

--- a/packages/app-mobile/package.json
+++ b/packages/app-mobile/package.json
@ -69,6 +69,7 @@
    "react-native-version-info": "1.1.1",
    "react-native-vosk": "0.1.12",
    "react-native-webview": "12.4.0",
+    "react-native-zip-archive": "6.0.9",
    "react-redux": "7.2.9",
    "redux": "4.2.1",
    "rn-fetch-blob": "0.12.0",
--- a/packages/app-mobile/services/voiceTyping/vosk.android.ts
+++ b/packages/app-mobile/services/voiceTyping/vosk.android.ts
@ -1,17 +1,25 @@
+import { languageCodeOnly } from '@joplin/lib/locale';
 import Logger from '@joplin/lib/Logger';
+import shim from '@joplin/lib/shim';
 import Vosk from 'react-native-vosk';
+import { unzip } from 'react-native-zip-archive';
+import RNFetchBlob from 'rn-fetch-blob';
+const md5 = require('md5');
+
 const logger = Logger.create('voiceTyping/vosk');

 enum State {
 	Idle = 0,
 	Recording,
+	Completing,
 }

 interface StartOptions {
 	onResult: (text: string)=> void;
 }

-let vosk_: Vosk|null = null;
+let vosk_: Record<string, Vosk> = {};
+
 let state_: State = State.Idle;

 export const voskEnabled = true;
@ -23,12 +31,113 @@ export interface Recorder {
 	cleanup: ()=> void;
 }

-export const getVosk = async () => {
-	if (vosk_) return vosk_;
-	vosk_ = new Vosk();
-	const result = await vosk_.loadModel('model-fr-fr');
+const supportedLanguages = {
+	'en': 'https://alphacephei.com/vosk/models/vosk-model-small-en-us-0.15.zip',
+	'cn': 'https://alphacephei.com/vosk/models/vosk-model-small-cn-0.22.zip',
+	'ru': 'https://alphacephei.com/vosk/models/vosk-model-small-ru-0.22.zip',
+	'fr': 'https://alphacephei.com/vosk/models/vosk-model-small-fr-0.22.zip',
+	'de': 'https://alphacephei.com/vosk/models/vosk-model-small-de-0.15.zip',
+	'es': 'https://alphacephei.com/vosk/models/vosk-model-small-es-0.42.zip',
+	'pt': 'https://alphacephei.com/vosk/models/vosk-model-small-pt-0.3.zip',
+	'tr': 'https://alphacephei.com/vosk/models/vosk-model-small-tr-0.3.zip',
+	'vn': 'https://alphacephei.com/vosk/models/vosk-model-small-vn-0.4.zip',
+	'it': 'https://alphacephei.com/vosk/models/vosk-model-small-it-0.22.zip',
+	'nl': 'https://alphacephei.com/vosk/models/vosk-model-small-nl-0.22.zip',
+	'uk': 'https://alphacephei.com/vosk/models/vosk-model-small-uk-v3-small.zip',
+	'ja': 'https://alphacephei.com/vosk/models/vosk-model-small-ja-0.22.zip',
+	'hi': 'https://alphacephei.com/vosk/models/vosk-model-small-hi-0.22.zip',
+	'cs': 'https://alphacephei.com/vosk/models/vosk-model-small-cs-0.4-rhasspy.zip',
+	'pl': 'https://alphacephei.com/vosk/models/vosk-model-small-pl-0.22.zip',
+	'uz': 'https://alphacephei.com/vosk/models/vosk-model-small-uz-0.22.zip',
+	'ko': 'https://alphacephei.com/vosk/models/vosk-model-small-ko-0.22.zip',
+};
+
+export const isSupportedLanguage = (locale: string) => {
+	const l = languageCodeOnly(locale).toLowerCase();
+	return Object.keys(supportedLanguages).includes(l);
+};
+
+// Where all the models files for all the languages are
+const getModelRootDir = () => {
+	return `${RNFetchBlob.fs.dirs.DocumentDir}/vosk-models`;
+};
+
+// Where we unzip a model after downloading it
+const getUnzipDir = (locale: string) => {
+	return `${getModelRootDir()}/${locale}`;
+};
+
+// Where the model for a particular language is
+const getModelDir = (locale: string) => {
+	return `${getUnzipDir(locale)}/model`;
+};
+
+const languageModelUrl = (locale: string) => {
+	const l = languageCodeOnly(locale).toLowerCase();
+	if (!(l in supportedLanguages)) throw new Error(`No language file for: ${locale}`);
+	return (supportedLanguages as any)[l];
+};
+
+export const modelIsDownloaded = async (locale: string) => {
+	const uuidFile = `${getModelDir(locale)}/uuid`;
+	return shim.fsDriver().exists(uuidFile);
+};
+
+export const getVosk = async (locale: string) => {
+	if (vosk_[locale]) return vosk_[locale];
+
+	const vosk = new Vosk();
+	const modelDir = await downloadModel(locale);
+	logger.info(`Loading model from ${modelDir}`);
+	await shim.fsDriver().readDirStats(modelDir);
+	const result = await vosk.loadModel(modelDir);
 	logger.info('getVosk:', result);
-	return vosk_;
+
+	vosk_ = { [locale]: vosk };
+
+	return vosk;
+};
+
+const downloadModel = async (locale: string) => {
+	const modelUrl = languageModelUrl(locale);
+	const unzipDir = getUnzipDir(locale);
+	const zipFilePath = `${unzipDir}.zip`;
+	const modelDir = getModelDir(locale);
+	const uuidFile = `${modelDir}/uuid`;
+
+	if (await modelIsDownloaded(locale)) {
+		logger.info(`Model for ${locale} already exists at ${modelDir}`);
+		return modelDir;
+	}
+
+	await shim.fsDriver().remove(unzipDir);
+
+	logger.info(`Downloading model from: ${modelUrl}`);
+
+	await shim.fetchBlob(languageModelUrl(locale), {
+		path: zipFilePath,
+	});
+
+	logger.info(`Unzipping ${zipFilePath} => ${unzipDir}`);
+
+	await unzip(zipFilePath, unzipDir);
+
+	const dirs = await shim.fsDriver().readDirStats(unzipDir);
+	if (dirs.length !== 1) {
+		logger.error('Expected 1 directory but got', dirs);
+		throw new Error(`Expected 1 directory, but got ${dirs.length}`);
+	}
+
+	const fullUnzipPath = `${unzipDir}/${dirs[0].path}`;
+
+	logger.info(`Moving ${fullUnzipPath} =>  ${modelDir}`);
+	await shim.fsDriver().rename(fullUnzipPath, modelDir);
+
+	await shim.fsDriver().writeFile(uuidFile, md5(modelUrl));
+
+	await shim.fsDriver().remove(zipFilePath);
+
+	return modelDir;
 };

 export const startRecording = (vosk: Vosk, options: StartOptions): Recorder => {
@ -110,8 +219,9 @@ export const startRecording = (vosk: Vosk, options: StartOptions): Recorder => {
 			});
 		},
 		cleanup: () => {
-			if (state_ !== State.Idle) {
+			if (state_ === State.Recording) {
 				logger.info('Cancelling...');
+				state_ = State.Completing;
 				vosk.stopOnly();
 				completeRecording('', null);
 			}
--- a/packages/app-mobile/services/voiceTyping/vosk.ios.ts
+++ b/packages/app-mobile/services/voiceTyping/vosk.ios.ts
@ -13,7 +13,15 @@ export interface Recorder {
 	cleanup: ()=> void;
 }

-export const getVosk = async () => {
+export const isSupportedLanguage = (_locale: string) => {
+	return false;
+};
+
+export const modelIsDownloaded = async (_locale: string) => {
+	return false;
+};
+
+export const getVosk = async (_locale: string) => {
 	return {} as any;
 };

--- a/packages/app-mobile/utils/fs-driver-rn.ts
+++ b/packages/app-mobile/utils/fs-driver-rn.ts
@ -116,6 +116,13 @@ export default class FsDriverRN extends FsDriverBase {
 		return RNFS.moveFile(source, dest);
 	}

+	public async rename(source: string, dest: string) {
+		if (isScopedUri(source) || isScopedUri(dest)) {
+			await RNSAF.rename(source, dest);
+		}
+		return RNFS.moveFile(source, dest);
+	}
+
 	public async exists(path: string) {
 		if (isScopedUri(path)) {
 			return RNSAF.exists(path);
--- a/packages/lib/locale.ts
+++ b/packages/lib/locale.ts
@ -508,7 +508,8 @@ function languageNameInEnglish(languageCode: string) {
 	return codeToLanguageE_[languageCode] ? codeToLanguageE_[languageCode] : '';
 }

-function languageName(languageCode: string, defaultToEnglish: boolean = true) {
+function languageName(canonicalName: string, defaultToEnglish: boolean = true) {
+	const languageCode = languageCodeOnly(canonicalName);
 	if (codeToLanguage_[languageCode]) return codeToLanguage_[languageCode];
 	if (defaultToEnglish) return languageNameInEnglish(languageCode);
 	return '';
@ -603,4 +604,4 @@ const stringByLocale = (locale: string, s: string, ...args: any[]): string => {
 	}
 };

-export { _, _n, supportedLocales, currentLocale, localesFromLanguageCode, languageCodeOnly, countryDisplayName, localeStrings, setLocale, supportedLocalesToLanguages, defaultLocale, closestSupportedLocale, languageCode, countryCodeOnly };
+export { _, _n, supportedLocales, languageName, currentLocale, localesFromLanguageCode, languageCodeOnly, countryDisplayName, localeStrings, setLocale, supportedLocalesToLanguages, defaultLocale, closestSupportedLocale, languageCode, countryCodeOnly };
--- a/packages/tools/release-android.ts
+++ b/packages/tools/release-android.ts
@ -118,8 +118,6 @@ async function createRelease(projectName: string, name: string, tagName: string,
 			content = content.replace(/\s+"react-native-vosk": ".*",/, '');
 			return content;
 		});
-
-		await patcher.removeFile(`${rnDir}/android/app/src/main/assets/model-fr-fr`);
 	}

 	if (name === 'vosk') {
--- a/yarn.lock
+++ b/yarn.lock
@ -4425,6 +4425,7 @@ __metadata:
    react-native-version-info: 1.1.1
    react-native-vosk: 0.1.12
    react-native-webview: 12.4.0
+    react-native-zip-archive: 6.0.9
    react-redux: 7.2.9
    redux: 4.2.1
    rn-fetch-blob: 0.12.0
@ -27380,11 +27381,11 @@ __metadata:

 "react-native-vosk@patch:react-native-vosk@npm%3A0.1.12#./.yarn/patches/react-native-vosk-npm-0.1.12-76b1caaae8.patch::locator=root%40workspace%3A.":
  version: 0.1.12
-  resolution: "react-native-vosk@patch:react-native-vosk@npm%3A0.1.12#./.yarn/patches/react-native-vosk-npm-0.1.12-76b1caaae8.patch::version=0.1.12&hash=b82215&locator=root%40workspace%3A."
+  resolution: "react-native-vosk@patch:react-native-vosk@npm%3A0.1.12#./.yarn/patches/react-native-vosk-npm-0.1.12-76b1caaae8.patch::version=0.1.12&hash=5deb2c&locator=root%40workspace%3A."
  peerDependencies:
    react: "*"
    react-native: "*"
-  checksum: 9451534de711f28274e117087a2c27266ee60ffc46723136841669c02345fd5644333a204c98c18d82e2c4221bb65906ac20529fd63fdcd5439c8aad86710e3b
+  checksum: ca31cf6345f422b778c18bc2365f5114ab2c3e20214e684bbb77220e6476cd877bb5c256d88070b2235f01714e8cd8e5bee999b854894b066d31834b9b17dca4
  languageName: node
  linkType: hard

@ -27401,6 +27402,16 @@ __metadata:
  languageName: node
  linkType: hard

+"react-native-zip-archive@npm:6.0.9":
+  version: 6.0.9
+  resolution: "react-native-zip-archive@npm:6.0.9"
+  peerDependencies:
+    react: ">=16.8.6"
+    react-native: ">=0.60.0"
+  checksum: 96c83864a596fcd8a82f55a01fa5e97822c9e80c7cf55de6b011a27cc436fbf58371207cb61c63b19dbd22b852ee62b95935358b4b9a73686695133dffdcfe27
+  languageName: node
+  linkType: hard
+
 "react-native@npm:0.70.6":
  version: 0.70.6
  resolution: "react-native@npm:0.70.6"