Upload TTS APKs to huggingface (#400)

.github/workflows/apk-tts.yaml

@@ -0,0 +1,103 @@
+name: apk-tts
+
+on:
+  push:
+    branches:
+      - apk-tts
+      - apk
+    tags:
+      - '*'
+
+  workflow_dispatch:
+
+concurrency:
+  group: apk-tts-${{ github.ref }}
+  cancel-in-progress: true
+
+permissions:
+  contents: write
+
+jobs:
+  apk_tts:
+    if: github.repository_owner == 'csukuangfj' || github.repository_owner == 'k2-fsa'
+    runs-on: ${{ matrix.os }}
+    name: apk for tts ${{ matrix.index }}/${{ matrix.total }}
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [ubuntu-latest]
+        total: ["12"]
+        index: ["0", "1", "2", "3", "4", "5", "6", "7", "8", "9", "10", "11"]
+
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      - name: Display NDK HOME
+        shell: bash
+        run: |
+          echo "ANDROID_NDK_LATEST_HOME: ${ANDROID_NDK_LATEST_HOME}"
+          ls -lh ${ANDROID_NDK_LATEST_HOME}
+
+      - name: Install Python dependencies
+        shell: bash
+        run: |
+          python3 -m pip install --upgrade pip jinja2
+
+      - name: Generate build script
+        shell: bash
+        run: |
+          cd scripts/apk
+
+          total=${{ matrix.total }}
+          index=${{ matrix.index }}
+
+          ./generate-tts-apk-script.py --total $total --index $index
+
+          chmod +x build-apk-tts.sh
+          mv -v ./build-apk-tts.sh ../..
+
+      - name: build APK
+        shell: bash
+        run: |
+          export ANDROID_NDK=$ANDROID_NDK_LATEST_HOME
+          ./build-apk-tts.sh
+
+      - name: Display APK
+        shell: bash
+        run: |
+          ls -lh ./apks/
+          du -h -d1 .
+
+      # - uses: actions/upload-artifact@v3
+      #   with:
+      #     name: tts-apk
+      #     path: ./apks/*.apk
+
+      - name: Publish to huggingface
+        env:
+          HF_TOKEN: ${{ secrets.HF_TOKEN }}
+        uses: nick-fields/retry@v2
+        with:
+          max_attempts: 20
+          timeout_seconds: 200
+          shell: bash
+          command: |
+            git config --global user.email "csukuangfj@gmail.com"
+            git config --global user.name "Fangjun Kuang"
+
+            rm -rf huggingface
+            export GIT_LFS_SKIP_SMUDGE=1
+
+            git clone https://huggingface.co/csukuangfj/sherpa-onnx-apk huggingface
+            cd huggingface
+            git pull
+
+            mkdir -p tts
+            cp -v ../apks/*.apk ./tts/
+            git status
+            git lfs track "*.apk"
+            git add .
+            git commit -m "add more apks"
+            git push https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/sherpa-onnx-apk main

.github/workflows/apk.yaml

@@ -38,7 +38,6 @@ jobs:
         shell: bash
         run: |
           export ANDROID_NDK=$ANDROID_NDK_LATEST_HOME
-          ./build-apk-tts.sh
           ./build-apk-vad.sh
           ./build-apk-two-pass.sh
           ./build-apk.sh

android/SherpaOnnxTts/app/src/main/java/com/k2fsa/sherpa/onnx/MainActivity.kt

@@ -84,6 +84,8 @@ class MainActivity : AppCompatActivity() {
         val ok = audio.samples.size > 0 && audio.save(filename)
         if (ok) {
             play.isEnabled = true
+            // Play automatically after generation
+            onClickPlay()
         }
     }
 
@@ -97,10 +99,24 @@ class MainActivity : AppCompatActivity() {
     }
 
     fun initTts() {
-        // 0 - vits-vctk (multi-speaker, English)
+        var modelDir :String?
-        // 1 - vits-zh-aishell3 (multi-speaker, Chinese)
+        var modelName :String?
-        val type = 0
+
-        val config = getOfflineTtsConfig(type = type, debug = true)!!
+        // The purpose of such a design is to make the CI test easier
+        // Please see
+        // https://github.com/k2-fsa/sherpa-onnx/blob/master/scripts/apk/generate-tts-apk-script.py
+        modelDir = null
+        modelName = null
+
+        // Example 1:
+        // modelDir = "vits-vctk"
+        // modelName = "vits-vctk.onnx"
+
+        // Example 2:
+        // modelDir = "vits-piper-en_US-lessac-medium"
+        // modelName = "en_US-lessac-medium.onnx"
+
+        val config = getOfflineTtsConfig(modelDir = modelDir!!, modelName = modelName!!)!!
         tts = OfflineTts(assetManager = application.assets, config = config)
     }
 }

android/SherpaOnnxTts/app/src/main/java/com/k2fsa/sherpa/onnx/Tts.kt

@@ -116,45 +116,17 @@ class OfflineTts(
 // please refer to
 // https://k2-fsa.github.io/sherpa/onnx/tts/pretrained_models/index.html
 // to download models
-//
+fun getOfflineTtsConfig(modelDir: String, modelName: String): OfflineTtsConfig? {
-// You can change the type as you wish
+    return OfflineTtsConfig(
-fun getOfflineTtsConfig(type: Int, debug: Boolean = false): OfflineTtsConfig? {
+        model = OfflineTtsModelConfig(
-    when (type) {
+            vits = OfflineTtsVitsModelConfig(
-        0 -> {
+                model = "$modelDir/$modelName",
-            val modelDir = "vits-vctk"
+                lexicon = "$modelDir/lexicon.txt",
-            return OfflineTtsConfig(
+                tokens = "$modelDir/tokens.txt"
-                model = OfflineTtsModelConfig(
+            ),
-                    vits = OfflineTtsVitsModelConfig(
+            numThreads = 2,
-                        model = "$modelDir/vits-vctk.onnx",
+            debug = false,
-                        lexicon = "$modelDir/lexicon.txt",
+            provider = "cpu",
-                        tokens = "$modelDir/tokens.txt"
+        )
-                    ),
+    )
-                    numThreads = 2,
-                    debug = debug,
-                    provider = "cpu",
-                )
-            )
-        }
-
-        1 -> {
-            val modelDir = "vits-zh-aishell3"
-            return OfflineTtsConfig(
-                model = OfflineTtsModelConfig(
-                    vits = OfflineTtsVitsModelConfig(
-                        model = "$modelDir/vits-aishell3.onnx",
-                        lexicon = "$modelDir/lexicon.txt",
-                        tokens = "$modelDir/tokens.txt"
-                    ),
-                    numThreads = 2,
-                    debug = debug,
-                    provider = "cpu",
-                )
-            )
-        }
-    }
-
-    println("Unsupported type $type")
-
-    return null
-
 }

scripts/apk/.gitignore

@@ -0,0 +1 @@
+build-apk-tts.sh

scripts/apk/README.md

@@ -0,0 +1,3 @@
+# Introduction
+
+This folder contains scripts for building Android APKs.

scripts/apk/build-apk-tts.sh.in

@@ -1,4 +1,6 @@
 #!/usr/bin/env bash
+#
+# Auto generated! Please DO NOT EDIT!
 
 # Please set the environment variable ANDROID_NDK
 # before running this script
@@ -27,61 +29,31 @@ log "====================x86-64===================="
 log "====================x86===================="
 ./build-android-x86.sh
 
-
 mkdir -p apks
 
-# Download the model
+{% for tts_model in tts_model_list %}
 pushd ./android/SherpaOnnxTts/app/src/main/assets/
-mkdir vits-vctk
+model_dir={{ tts_model.model_dir }}
+model_name={{ tts_model.model_name }}
+lang={{ tts_model.lang }}
+
+mkdir $model_dir
+cd $model_dir
+wget -qq https://huggingface.co/csukuangfj/$model_dir/resolve/main/$model_name
+wget -qq https://huggingface.co/csukuangfj/$model_dir/resolve/main/lexicon.txt
+wget -qq https://huggingface.co/csukuangfj/$model_dir/resolve/main/tokens.txt
+wget -qq https://huggingface.co/csukuangfj/$model_dir/resolve/main/MODEL_CARD 2>/dev/null || true
 
-cd vits-vctk
-wget -qq https://huggingface.co/csukuangfj/vits-vctk/resolve/main/vits-vctk.onnx
-wget -qq https://huggingface.co/csukuangfj/vits-vctk/resolve/main/lexicon.txt
-wget -qq https://huggingface.co/csukuangfj/vits-vctk/resolve/main/tokens.txt
 popd
-
+# Now we are at the project root directory
-for arch in arm64-v8a armeabi-v7a x86_64 x86; do
-  log "------------------------------------------------------------"
-  log "build tts apk for $arch"
-  log "------------------------------------------------------------"
-  src_arch=$arch
-  if [ $arch == "armeabi-v7a" ]; then
-    src_arch=armv7-eabi
-  elif [ $arch == "x86_64" ]; then
-    src_arch=x86-64
-  fi
-
-  ls -lh ./build-android-$src_arch/install/lib/*.so
-
-  cp -v ./build-android-$src_arch/install/lib/*.so ./android/SherpaOnnxTts/app/src/main/jniLibs/$arch/
-
-  pushd ./android/SherpaOnnxTts
-  ./gradlew build
-  popd
-
-  mv android/SherpaOnnxTts/app/build/outputs/apk/debug/app-debug.apk ./apks/sherpa-onnx-${SHERPA_ONNX_VERSION}-$arch-en-tts-multi-speaker-vctk.apk
-  ls -lh apks
-  rm -v ./android/SherpaOnnxTts/app/src/main/jniLibs/$arch/*.so
-done
-
-rm -rf ./android/SherpaOnnxTts/app/src/main/assets/vits-vctk
 
 git checkout .
 pushd android/SherpaOnnxTts/app/src/main/java/com/k2fsa/sherpa/onnx
-sed -i.bak s/"type = 0"/"type = 1"/ ./MainActivity.kt
+sed -i.bak s/"modelDir = null"/"modelDir = \"$model_dir\""/ ./MainActivity.kt
+sed -i.bak s/"modelName = null"/"modelName = \"$model_name\""/ ./MainActivity.kt
 git diff
 popd
 
-pushd ./android/SherpaOnnxTts/app/src/main/assets/
-mkdir vits-zh-aishell3
-cd vits-zh-aishell3
-
-wget -qq https://huggingface.co/csukuangfj/vits-zh-aishell3/resolve/main/vits-aishell3.onnx
-wget -qq https://huggingface.co/csukuangfj/vits-zh-aishell3/resolve/main/lexicon.txt
-wget -qq  https://huggingface.co/csukuangfj/vits-zh-aishell3/resolve/main/tokens.txt
-
-popd
-
 for arch in arm64-v8a armeabi-v7a x86_64 x86; do
   log "------------------------------------------------------------"
   log "build tts apk for $arch"
@@ -101,12 +73,13 @@ for arch in arm64-v8a armeabi-v7a x86_64 x86; do
   ./gradlew build
   popd
 
-  mv android/SherpaOnnxTts/app/build/outputs/apk/debug/app-debug.apk ./apks/sherpa-onnx-${SHERPA_ONNX_VERSION}-$arch-zh-tts-multi-speaker-aishell3.apk
+  mv android/SherpaOnnxTts/app/build/outputs/apk/debug/app-debug.apk ./apks/sherpa-onnx-${SHERPA_ONNX_VERSION}-$arch-$lang-tts-$model_dir.apk
   ls -lh apks
   rm -v ./android/SherpaOnnxTts/app/src/main/jniLibs/$arch/*.so
 done
 
-rm -rf ./android/SherpaOnnxTts/app/src/main/assets/vits-vctk
+rm -rf ./android/SherpaOnnxTts/app/src/main/assets/$model_dir
+{% endfor %}
 
 git checkout .
 

scripts/apk/generate-tts-apk-script.py

@@ -0,0 +1,127 @@
+#!/usr/bin/env python3
+
+from dataclasses import dataclass
+
+import jinja2
+from typing import List
+import argparse
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--total",
+        type=int,
+        default=1,
+        help="Number of runners",
+    )
+    parser.add_argument(
+        "--index",
+        type=int,
+        default=0,
+        help="Index of the current runner",
+    )
+    return parser.parse_args()
+
+
+@dataclass
+class TtsModel:
+    model_dir: str
+    model_name: str
+    lang: str  # en, zh, fr, de, etc.
+
+
+def get_all_models() -> List[TtsModel]:
+    return [
+        TtsModel(
+            model_dir="vits-zh-aishell3", model_name="vits-aishell3.onnx", lang="zh"
+        ),
+        # English (US)
+        # fmt: off
+        TtsModel(model_dir="vits-vctk", model_name="vits-vctk.onnx", lang="en"),
+        TtsModel(model_dir="vits-ljs", model_name="vits-ljs.onnx", lang="en"),
+        TtsModel(model_dir="vits-piper-en_US-amy-low", model_name="en_US-amy-low.onnx", lang="en",),
+        TtsModel(model_dir="vits-piper-en_US-amy-medium", model_name="en_US-amy-medium.onnx", lang="en",),
+        TtsModel(model_dir="vits-piper-en_US-arctic-medium", model_name="en_US-arctic-medium.onnx", lang="en",),
+        TtsModel(model_dir="vits-piper-en_US-danny-low", model_name="en_US-danny-low.onnx", lang="en",),
+        TtsModel(model_dir="vits-piper-en_US-hfc_male-medium", model_name="en_US-hfc_male-medium.onnx", lang="en",),
+        TtsModel(model_dir="vits-piper-en_US-joe-medium", model_name="en_US-joe-medium.onnx", lang="en",),
+        TtsModel(model_dir="vits-piper-en_US-kathleen-low", model_name="en_US-kathleen-low.onnx", lang="en",),
+        TtsModel(model_dir="vits-piper-en_US-kusal-medium", model_name="en_US-kusal-medium.onnx", lang="en",),
+        TtsModel(model_dir="vits-piper-en_US-l2arctic-medium", model_name="en_US-l2arctic-medium.onnx", lang="en",),
+        TtsModel(model_dir="vits-piper-en_US-lessac-low", model_name="en_US-lessac-low.onnx", lang="en",),
+        TtsModel(model_dir="vits-piper-en_US-lessac-medium", model_name="en_US-lessac-medium.onnx", lang="en",),
+        TtsModel(model_dir="vits-piper-en_US-lessac-high", model_name="en_US-lessac-high.onnx", lang="en",),
+        TtsModel(model_dir="vits-piper-en_US-libritts-high", model_name="en_US-libritts-high.onnx", lang="en",),
+        TtsModel(model_dir="vits-piper-en_US-libritts_r-medium", model_name="en_US-libritts_r-medium.onnx", lang="en",),
+        TtsModel(model_dir="vits-piper-en_US-ryan-low", model_name="en_US-ryan-low.onnx", lang="en",),
+        TtsModel(model_dir="vits-piper-en_US-ryan-medium", model_name="en_US-ryan-medium.onnx", lang="en",),
+        TtsModel(model_dir="vits-piper-en_US-ryan-high", model_name="en_US-ryan-high.onnx", lang="en",),
+        # English (GB)
+        TtsModel(model_dir="vits-piper-en_GB-alan-low", model_name="en_GB-alan-low.onnx",lang="en",),
+        TtsModel(model_dir="vits-piper-en_GB-alan-medium", model_name="en_GB-alan-medium.onnx",lang="en",),
+        TtsModel(model_dir="vits-piper-en_GB-alba-medium", model_name="en_GB-alba-medium.onnx",lang="en",),
+        TtsModel(model_dir="vits-piper-en_GB-jenny_dioco-medium", model_name="en_GB-jenny_dioco-medium.onnx",lang="en",),
+        TtsModel(model_dir="vits-piper-en_GB-northern_english_male-medium", model_name="en_GB-northern_english_male-medium.onnx",lang="en",),
+        TtsModel(model_dir="vits-piper-en_GB-semaine-medium", model_name="en_GB-semaine-medium.onnx",lang="en",),
+        TtsModel(model_dir="vits-piper-en_GB-southern_english_female-low", model_name="en_GB-southern_english_female-low.onnx",lang="en",),
+        TtsModel(model_dir="vits-piper-en_GB-vctk-medium", model_name="en_GB-vctk-medium.onnx",lang="en",),
+        # German (DE)
+        TtsModel(model_dir="vits-piper-de_DE-eva_k-x_low", model_name="de_DE-eva_k-x_low.onnx",lang="de",),
+        TtsModel(model_dir="vits-piper-de_DE-karlsson-low", model_name="de_DE-karlsson-low.onnx",lang="de",),
+        TtsModel(model_dir="vits-piper-de_DE-kerstin-low", model_name="de_DE-kerstin-low.onnx",lang="de",),
+        TtsModel(model_dir="vits-piper-de_DE-pavoque-low", model_name="de_DE-pavoque-low.onnx",lang="de",),
+        TtsModel(model_dir="vits-piper-de_DE-ramona-low", model_name="de_DE-ramona-low.onnx",lang="de",),
+        TtsModel(model_dir="vits-piper-de_DE-thorsten-low", model_name="de_DE-thorsten-low.onnx",lang="de",),
+        TtsModel(model_dir="vits-piper-de_DE-thorsten-medium", model_name="de_DE-thorsten-medium.onnx",lang="de",),
+        TtsModel(model_dir="vits-piper-de_DE-thorsten-high", model_name="de_DE-thorsten-high.onnx",lang="de",),
+        TtsModel(model_dir="vits-piper-de_DE-thorsten_emotional-medium", model_name="de_DE-thorsten_emotional-medium.onnx",lang="de",),
+        # French (FR)
+        TtsModel(model_dir="vits-piper-fr_FR-upmc-medium", model_name="fr_FR-upmc-medium.onnx",lang="fr",),
+        TtsModel(model_dir="vits-piper-fr_FR-siwis-low", model_name="fr_FR-siwis-low.onnx",lang="fr",),
+        TtsModel(model_dir="vits-piper-fr_FR-siwis-medium", model_name="fr_FR-siwis-medium.onnx",lang="fr",),
+
+        # Spanish (ES)
+        TtsModel(model_dir="vits-piper-es_ES-carlfm-x_low", model_name="es_ES-carlfm-x_low.onnx",lang="es",),
+        TtsModel(model_dir="vits-piper-es_ES-davefx-medium", model_name="es_ES-davefx-medium.onnx",lang="es",),
+        TtsModel(model_dir="vits-piper-es_ES-mls_10246-low", model_name="es_ES-mls_10246-low.onnx",lang="es",),
+        TtsModel(model_dir="vits-piper-es_ES-mls_9972-low", model_name="es_ES-mls_9972-low.onnx",lang="es",),
+        TtsModel(model_dir="vits-piper-es_ES-sharvard-medium", model_name="es_ES-sharvard-medium.onnx",lang="es",),
+
+        # Spanish (MX)
+        TtsModel(model_dir="vits-piper-es_MX-ald-medium", model_name="es_MX-ald-medium.onnx",lang="es",),
+        # fmt: on
+    ]
+
+
+def main():
+    args = get_args()
+    index = args.index
+    total = args.total
+    assert 0 <= index < total, (index, total)
+    environment = jinja2.Environment()
+    with open("./build-apk-tts.sh.in") as f:
+        s = f.read()
+    template = environment.from_string(s)
+    d = dict()
+    all_model_list = get_all_models()
+    num_models = len(all_model_list)
+
+    num_per_runner = num_models // total
+    if num_per_runner <= 0:
+        raise ValueError(f"num_models: {num_models}, num_runners: {total}")
+
+    start = index * num_per_runner
+    end = start + num_per_runner
+    if index == args.total - 1:
+        end = num_models
+
+    print(f"{index}/{total}: {start}-{end}/{num_models}")
+    d["tts_model_list"] = all_model_list[start:end]
+    s = template.render(**d)
+    with open("./build-apk-tts.sh", "w") as f:
+        print(s, file=f)
+
+
+if __name__ == "__main__":
+    main()