Add React Native Web support

AGENTS.md

@@ -21,6 +21,7 @@ Main flow:
 - `src/index.ts`: public exports.
 - `src/KittenTTS.ts`: main SDK class and lifecycle.
 - `src/KittenTTSConfig.ts`: user config and defaults.
+- `src/*.web.ts`: React Native Web entrypoints and platform-specific browser implementations.
 - `src/KittenTTSError.ts`: SDK error codes and helpers.
 - `src/KittenModel.ts`: model names, download URLs, sizes, speed priors.
 - `src/KittenVoice.ts`: voice enum and display helpers.
@@ -29,6 +30,7 @@ Main flow:
 - `src/engine/TTSEngine.ts`: text-to-token-to-ONNX inference.
 - `src/phonemizer/CEPhonemizer.ts`: JS/Emscripten phonemizer adapter.
 - `src/audio/AudioOutput.ts`: optional playback helpers.
+- `src/storage/AssetStorage.ts`: web/Node asset cache abstraction used by the web platform files.
 - `vendor/cephonemizer/`: vendored C++ phonemizer source.
 - `scripts/build-cephonemizer.js`: builds generated phonemizer runtime.
 - `scripts/patch-onnxruntime-react-native.js`: postinstall ONNX Runtime compatibility patches.

CHANGELOG.md

@@ -5,6 +5,8 @@
 - Added Swift-parity word timing metadata via `KittenTTSResult.wordTimings`.
 - Added `KittenTTS.generateStreaming()` for sentence-by-sentence generation.
 - Added `tts.play(result)` so apps can inspect timings before playback.
+- Added React Native Web support through browser-specific ONNX Runtime Web,
+  Cache API asset storage, CE phonemizer, and audio playback implementations.
 
 ## 0.8.0
 

README.md

@@ -5,9 +5,9 @@
 </p>
 
 <p align="center">
-  On-device text-to-speech for React Native.
+  On-device text-to-speech for React Native and React Native Web.
   <br />
-  Generate speech on iOS and Android without sending text to a cloud TTS API.
+  Generate speech on iOS, Android, and web without sending text to a cloud TTS API.
 </p>
 
 <p align="center">
@@ -21,9 +21,15 @@
 
 > Developer preview. APIs may change between releases.
 
-> Expo Go will not work. KittenTTS uses native modules
-> (`onnxruntime-react-native` and `react-native-fs`), so Expo apps need a
-> development build or a prebuilt native project.
+> Expo Go will not work for native iOS/Android. KittenTTS uses native modules
+> (`onnxruntime-react-native` and `react-native-fs`) on mobile, so Expo apps
+> need a development build or a prebuilt native project. Web builds use
+> `onnxruntime-web` and browser storage instead.
+
+> React Native Web loads a pinned ONNX Runtime Web script and WASM assets from
+> jsDelivr by default. For production apps that need CDN independence or stricter
+> supply-chain controls, self-host those ONNX Runtime assets and set
+> `ortWasmPath`.
 
 ## See It In Action
 
@@ -36,6 +42,14 @@
   <strong>Device: iOS</strong> · Expo example &nbsp;&nbsp;&nbsp; <strong>Device: Android</strong> · Word timings
 </p>
 
+<p align="center">
+  <img src="assets/web-example.gif" alt="KittenTTS React Native Web example running in a browser" width="90%" />
+</p>
+
+<p align="center">
+  <strong>Web</strong> · Browser example
+</p>
+
 ---
 
 ## What Is KittenTTS React Native?
@@ -60,6 +74,7 @@ No cloud. No API key. No text leaving the device for speech generation.
 | --- | --- | --- |
 | React Native iOS | Developer preview | [Getting started](docs/getting-started.md) |
 | React Native Android | Developer preview | [Getting started](docs/getting-started.md) |
+| React Native Web | Developer preview | [Getting started](docs/getting-started.md#web) |
 | Expo development build | Supported | [Expo setup](docs/getting-started.md#expo-development-build) |
 | Expo Go | Not supported | [Why not?](docs/troubleshooting.md#expo-go-fails) |
 
@@ -109,6 +124,21 @@ const tts = await KittenTTS.create({
 await tts.speak('This voice is generated on the device.');
 ```
 
+Play audio in a web build:
+
+```tsx
+import {
+  KittenTTS,
+  createBrowserAudioPlayer,
+} from '@kittentts/react-native';
+
+const tts = await KittenTTS.create({
+  player: createBrowserAudioPlayer(),
+});
+
+await tts.speak('This voice is generated in the browser.');
+```
+
 [Full getting started guide →](docs/getting-started.md)
 
 ---
@@ -153,7 +183,7 @@ If the app opens in Expo Go, stop it and run `npx expo run:ios` or
 
 ## Features
 
-- [On-device TTS inference](docs/getting-started.md) on iOS and Android.
+- [On-device TTS inference](docs/getting-started.md) on iOS, Android, and web.
 - [Model download and cache](docs/reference/api.md#cache-methods) with progress callbacks.
 - [Bundled offline assets](docs/guides/offline-assets.md) for apps that cannot depend on a first-run download.
 - [Expo development builds](docs/getting-started.md#expo-development-build); Expo Go is [not supported](docs/troubleshooting.md#expo-go-fails).

docs/getting-started.md

@@ -10,6 +10,7 @@ instance, and generate speech.
 | React Native | `>= 0.72` |
 | iOS | `15.1+` |
 | Android | API `24+` |
+| Web | modern browser with WebAssembly support |
 | Node.js | `20+` recommended for examples |
 
 Expo Go will not work. KittenTTS depends on native modules:
@@ -18,6 +19,8 @@ Expo Go will not work. KittenTTS depends on native modules:
 - `react-native-fs`
 
 Use a bare React Native app, an Expo development build, or a prebuilt Expo app.
+React Native Web builds use `onnxruntime-web` and do not require those native
+modules at runtime.
 
 ## Install
 
@@ -57,6 +60,35 @@ npm install react-native-sound
 cd ios && pod install && cd ..
 ```
 
+## Web
+
+React Native Web builds resolve the package's browser entrypoint. The web
+runtime uses `onnxruntime-web`, Cache API storage for downloaded model files,
+and the same JavaScript CE phonemizer.
+
+```tsx
+import {
+  KittenTTS,
+  createBrowserAudioPlayer,
+} from '@kittentts/react-native';
+
+const tts = await KittenTTS.create({
+  player: createBrowserAudioPlayer(),
+});
+
+await tts.speak('Hello from KittenTTS on web.');
+await tts.dispose();
+```
+
+The browser path also supports `generate()`, `wordTimings`, `wavData()`, and
+`wavBase64()`. Pass `ortWasmPath` if your app needs to self-host ONNX Runtime
+WASM assets instead of using the SDK defaults.
+
+By default, browser builds load the pinned ONNX Runtime Web script and WASM
+assets from jsDelivr. That keeps the SDK simple to drop into React Native Web,
+but production apps that require tighter supply-chain control or CDN outage
+isolation should self-host those files and set `ortWasmPath` to that directory.
+
 ## Generate Audio
 
 Use `generate()` when you want audio data back without playing it immediately.
@@ -117,6 +149,8 @@ await tts.dispose();
 
 The first `KittenTTS.create()` downloads the selected model, `voices.npz`, and
 phonemizer files. Later calls reuse the device cache.
+On web, the cache is stored through the browser Cache API when available and
+falls back to memory storage.
 
 Default model cache:
 

docs/guides/playback.md

@@ -52,6 +52,23 @@ const tts = await KittenTTS.create({
 await tts.speak('This plays through react-native-sound.');
 ```
 
+## Browser Audio
+
+React Native Web builds can use the browser audio helper:
+
+```tsx
+import {
+  KittenTTS,
+  createBrowserAudioPlayer,
+} from '@kittentts/react-native';
+
+const tts = await KittenTTS.create({
+  player: createBrowserAudioPlayer(),
+});
+
+await tts.speak('This plays through an HTML audio element.');
+```
+
 ## Generate First, Then Play
 
 This is useful when the UI needs metadata from the generated result before