Add JavaScript (WebAssembly) API for Omnilingual ASR CTC models

.github/scripts/test-nodejs-npm.sh

@@ -9,6 +9,14 @@ git status
 ls -lh
 ls -lh node_modules
 
+curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-omnilingual-asr-1600-languages-300M-ctc-int8-2025-11-12.tar.bz2
+tar xvf sherpa-onnx-omnilingual-asr-1600-languages-300M-ctc-int8-2025-11-12.tar.bz2
+rm sherpa-onnx-omnilingual-asr-1600-languages-300M-ctc-int8-2025-11-12.tar.bz2
+
+node ./test-offline-omnilingual-asr-ctc.js
+
+rm -rf sherpa-onnx-omnilingual-asr-1600-languages-300M-ctc-int8-2025-11-12
+
 curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-wenetspeech-yue-u2pp-conformer-ctc-zh-en-cantonese-int8-2025-09-10.tar.bz2
 tar xvf sherpa-onnx-wenetspeech-yue-u2pp-conformer-ctc-zh-en-cantonese-int8-2025-09-10.tar.bz2
 rm sherpa-onnx-wenetspeech-yue-u2pp-conformer-ctc-zh-en-cantonese-int8-2025-09-10.tar.bz2

nodejs-examples/README.md

@@ -203,10 +203,26 @@ rm sherpa-onnx-zipformer-ctc-zh-int8-2025-07-03.tar.bz2
 node ./test-offline-zipformer-ctc.js
 ```
 
+## ./test-offline-omnilingual-asr-ctc.js
+
+[./test-offline-omnilingual-asr-ctc.js](./test-offline-omnilingual-asr-ctc.js) demonstrates
+how to decode a file with a Omnilingual ASR CTC model. In the code we use
+[sherpa-onnx-omnilingual-asr-1600-languages-300M-ctc-int8-2025-11-12.tar.bz2](https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-omnilingual-asr-1600-languages-300M-ctc-int8-2025-11-12.tar.bz2).
+
+You can use the following command to run it:
+
+```bash
+wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-omnilingual-asr-1600-languages-300M-ctc-int8-2025-11-12.tar.bz2
+tar xvf sherpa-onnx-omnilingual-asr-1600-languages-300M-ctc-int8-2025-11-12.tar.bz2
+rm sherpa-onnx-omnilingual-asr-1600-languages-300M-ctc-int8-2025-11-12.tar.bz2
+
+node ./test-offline-omnilingual-asr-ctc.js
+```
+
 ## ./test-offline-wenet-ctc.js
 
 [./test-offline-wenet-ctc.js](./test-offline-wenet-ctc.js) demonstrates
-how to decode a file with a Wenet CTC model. In the code we use
+how to decode a file with a WeNet CTC model. In the code we use
 [sherpa-onnx-wenetspeech-yue-u2pp-conformer-ctc-zh-en-cantonese-int8-2025-09-10.tar.bz2](https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-wenetspeech-yue-u2pp-conformer-ctc-zh-en-cantonese-int8-2025-09-10.tar.bz2).
 
 You can use the following command to run it:

nodejs-examples/test-offline-omnilingual-asr-ctc.js

@@ -0,0 +1,37 @@
+// Copyright (c)  2025  Xiaomi Corporation (authors: Fangjun Kuang)
+//
+const fs = require('fs');
+const {Readable} = require('stream');
+const wav = require('wav');
+
+const sherpa_onnx = require('sherpa-onnx');
+
+function createOfflineRecognizer() {
+  let config = {
+    modelConfig: {
+      omnilingual: {
+        model:
+            './sherpa-onnx-omnilingual-asr-1600-languages-300M-ctc-int8-2025-11-12/model.int8.onnx',
+      },
+      tokens:
+          './sherpa-onnx-omnilingual-asr-1600-languages-300M-ctc-int8-2025-11-12/tokens.txt',
+    }
+  };
+
+  return sherpa_onnx.createOfflineRecognizer(config);
+}
+
+const recognizer = createOfflineRecognizer();
+const stream = recognizer.createStream();
+
+const waveFilename =
+    './sherpa-onnx-omnilingual-asr-1600-languages-300M-ctc-int8-2025-11-12/test_wavs/en.wav';
+const wave = sherpa_onnx.readWave(waveFilename);
+stream.acceptWaveform(wave.sampleRate, wave.samples);
+
+recognizer.decode(stream);
+const text = recognizer.getResult(stream).text;
+console.log(text);
+
+stream.free();
+recognizer.free();

wasm/asr/sherpa-onnx-asr.js

@@ -55,6 +55,10 @@ function freeConfig(config, Module) {
     freeConfig(config.wenetCtc, Module)
   }
 
+  if ('omnilingual' in config) {
+    freeConfig(config.omnilingual, Module)
+  }
+
   if ('moonshine' in config) {
     freeConfig(config.moonshine, Module)
   }
@@ -755,6 +759,23 @@ function initSherpaOnnxOfflineWenetCtcModelConfig(config, Module) {
   }
 }
 
+function initSherpaOnnxOfflineOmnilingualAsrCtcModelConfig(config, Module) {
+  const n = Module.lengthBytesUTF8(config.model || '') + 1;
+
+  const buffer = Module._malloc(n);
+
+  const len = 1 * 4;  // 1 pointer
+  const ptr = Module._malloc(len);
+
+  Module.stringToUTF8(config.model || '', buffer, n);
+
+  Module.setValue(ptr, buffer, 'i8*');
+
+  return {
+    buffer: buffer, ptr: ptr, len: len,
+  }
+}
+
 function initSherpaOnnxOfflineWhisperModelConfig(config, Module) {
   const encoderLen = Module.lengthBytesUTF8(config.encoder || '') + 1;
   const decoderLen = Module.lengthBytesUTF8(config.decoder || '') + 1;
@@ -1025,6 +1046,12 @@ function initSherpaOnnxOfflineModelConfig(config, Module) {
     };
   }
 
+  if (!('omnilingual' in config)) {
+    config.omnilingual = {
+      model: '',
+    };
+  }
+
   if (!('whisper' in config)) {
     config.whisper = {
       encoder: '',
@@ -1109,9 +1136,13 @@ function initSherpaOnnxOfflineModelConfig(config, Module) {
   const wenetCtc =
       initSherpaOnnxOfflineWenetCtcModelConfig(config.wenetCtc, Module);
 
+  const omnilingual = initSherpaOnnxOfflineOmnilingualAsrCtcModelConfig(
+      config.omnilingual, Module);
+
   const len = transducer.len + paraformer.len + nemoCtc.len + whisper.len +
       tdnn.len + 8 * 4 + senseVoice.len + moonshine.len + fireRedAsr.len +
-      dolphin.len + zipformerCtc.len + canary.len + wenetCtc.len;
+      dolphin.len + zipformerCtc.len + canary.len + wenetCtc.len +
+      omnilingual.len;
 
   const ptr = Module._malloc(len);
 
@@ -1222,12 +1253,15 @@ function initSherpaOnnxOfflineModelConfig(config, Module) {
   Module._CopyHeap(wenetCtc.ptr, wenetCtc.len, ptr + offset);
   offset += wenetCtc.len;
 
+  Module._CopyHeap(omnilingual.ptr, omnilingual.len, ptr + offset);
+  offset += omnilingual.len;
+
   return {
     buffer: buffer, ptr: ptr, len: len, transducer: transducer,
         paraformer: paraformer, nemoCtc: nemoCtc, whisper: whisper, tdnn: tdnn,
         senseVoice: senseVoice, moonshine: moonshine, fireRedAsr: fireRedAsr,
         dolphin: dolphin, zipformerCtc: zipformerCtc, canary: canary,
-        wenetCtc: wenetCtc,
+        wenetCtc: wenetCtc, omnilingual: omnilingual
   }
 }
 

wasm/nodejs/sherpa-onnx-wasm-nodejs.cc

@@ -15,6 +15,7 @@ static_assert(sizeof(SherpaOnnxOfflineParaformerModelConfig) == 4, "");
 
 static_assert(sizeof(SherpaOnnxOfflineZipformerCtcModelConfig) == 4, "");
 static_assert(sizeof(SherpaOnnxOfflineWenetCtcModelConfig) == 4, "");
+static_assert(sizeof(SherpaOnnxOfflineOmnilingualAsrCtcModelConfig) == 4, "");
 static_assert(sizeof(SherpaOnnxOfflineDolphinModelConfig) == 4, "");
 static_assert(sizeof(SherpaOnnxOfflineNemoEncDecCtcModelConfig) == 4, "");
 static_assert(sizeof(SherpaOnnxOfflineWhisperModelConfig) == 5 * 4, "");
@@ -37,7 +38,8 @@ static_assert(sizeof(SherpaOnnxOfflineModelConfig) ==
                       sizeof(SherpaOnnxOfflineDolphinModelConfig) +
                       sizeof(SherpaOnnxOfflineZipformerCtcModelConfig) +
                       sizeof(SherpaOnnxOfflineCanaryModelConfig) +
-                      sizeof(SherpaOnnxOfflineWenetCtcModelConfig),
+                      sizeof(SherpaOnnxOfflineWenetCtcModelConfig) +
+                      sizeof(SherpaOnnxOfflineOmnilingualAsrCtcModelConfig),
 
               "");
 static_assert(sizeof(SherpaOnnxFeatureConfig) == 2 * 4, "");
@@ -86,6 +88,7 @@ void PrintOfflineRecognizerConfig(SherpaOnnxOfflineRecognizerConfig *config) {
   auto zipformer_ctc = &model_config->zipformer_ctc;
   auto canary = &model_config->canary;
   auto wenet_ctc = &model_config->wenet_ctc;
+  auto omnilingual = &model_config->omnilingual;
 
   fprintf(stdout, "----------offline transducer model config----------\n");
   fprintf(stdout, "encoder: %s\n", transducer->encoder);
@@ -139,6 +142,9 @@ void PrintOfflineRecognizerConfig(SherpaOnnxOfflineRecognizerConfig *config) {
   fprintf(stdout, "----------offline wenet ctc model config----------\n");
   fprintf(stdout, "model: %s\n", wenet_ctc->model);
 
+  fprintf(stdout, "----------offline Omnilingual ASR model config----------\n");
+  fprintf(stdout, "model: %s\n", omnilingual->model);
+
   fprintf(stdout, "tokens: %s\n", model_config->tokens);
   fprintf(stdout, "num_threads: %d\n", model_config->num_threads);
   fprintf(stdout, "provider: %s\n", model_config->provider);