Use shared eSpeak frontend for Kokoro English

CMakeLists.txt

@@ -586,16 +586,13 @@ audiocpp_add_model(demucs
 audiocpp_add_model(kokoro_tts
     SOURCES
         src/models/kokoro_tts/assets.cpp
-        src/models/kokoro_tts/package.cpp
         src/models/kokoro_tts/g2p_multilingual.cpp
-        src/models/kokoro_tts/loader.cpp
         src/models/kokoro_tts/frontend.cpp
-        src/models/kokoro_tts/g2p_en.cpp
         src/models/kokoro_tts/predictor.cpp
         src/models/kokoro_tts/decoder.cpp
         src/models/kokoro_tts/plbert.cpp
         src/models/kokoro_tts/session.cpp
-    INCLUDES engine/models/kokoro_tts/loader.h
+    INCLUDES engine/models/kokoro_tts/session.h
     LOADERS engine::models::kokoro_tts::make_kokoro_tts_loader
 )
 if(MSVC)

docs/models/kokoro_tts.md

@@ -0,0 +1,136 @@
+# Kokoro 82M
+
+Kokoro 82M is a compact multilingual text-to-speech model exposed as
+`--family kokoro_tts`. audio.cpp packages the model as standalone GGUF files
+with all 54 upstream voice packs.
+
+## Install
+
+```bash
+python3 tools/model_manager_v2.py install kokoro_82m_q8_0
+```
+
+The default package installs:
+
+```text
+models/Kokoro-82M-GGUF/kokoro-82m-q8_0.gguf
+```
+
+BF16 is also available:
+
+```bash
+python3 tools/model_manager_v2.py install kokoro_82m_bf16
+```
+
+## Quick Start
+
+```bash
+audiocpp_cli --task tts --family kokoro_tts \
+  --model models/Kokoro-82M-GGUF/kokoro-82m-q8_0.gguf \
+  --backend cpu \
+  --language en-us \
+  --voice-id af_heart \
+  --text "Hello from Kokoro." \
+  --out out.wav
+```
+
+Chinese example:
+
+```bash
+audiocpp_cli --task tts --family kokoro_tts \
+  --model models/Kokoro-82M-GGUF/kokoro-82m-q8_0.gguf \
+  --backend cpu \
+  --language zh \
+  --voice-id zf_xiaobei \
+  --text "你好，这是中文语音测试。" \
+  --out out_zh.wav
+```
+
+## Model
+
+| Field | Value |
+|---|---|
+| Family | `kokoro_tts` |
+| Task | `tts` |
+| Modes | `offline` |
+| Default package | `kokoro_82m_q8_0` |
+| Other package | `kokoro_82m_bf16` |
+| Model file | `models/Kokoro-82M-GGUF/kokoro-82m-q8_0.gguf` |
+
+## Voices and Languages
+
+Use `--voice-id <id>` to select one of the packaged voices. The voice prefix
+selects the language family:
+
+| Prefix | Language |
+|---|---|
+| `af`, `am` | American English |
+| `bf`, `bm` | British English |
+| `ef`, `em` | Spanish |
+| `ff` | French |
+| `hf`, `hm` | Hindi |
+| `if`, `im` | Italian |
+| `jf`, `jm` | Japanese |
+| `pf`, `pm` | Brazilian Portuguese |
+| `zf`, `zm` | Mandarin Chinese |
+
+The request language must match the selected voice. For example, use
+`--language zh` with `zf_*` or `zm_*` voices.
+
+## Runtime Resources
+
+The release GGUF includes model weights, config, vocabulary, all voice packs,
+and the generated `g2p/ja.json` and `g2p/zh.json` tables.
+
+English, Spanish, French, Hindi, Italian, and Portuguese use the shared eSpeak
+runtime. Install or package eSpeak data as described in
+[`docs/espeak_phonemizer.md`](../espeak_phonemizer.md).
+
+Chinese works from the release GGUF because `g2p/zh.json` is bundled. Avoid
+mixed Latin words inside Chinese text unless they are known to map to Kokoro's
+vocabulary.
+
+Japanese also needs MeCab and UniDic. The small release GGUF does not include
+UniDic because it is large. Export a local full multilingual GGUF with
+`--embed-multilingual-resources` if you need Japanese to work from a bundled
+package.
+
+## Options
+
+| Option | Values | Default | Meaning |
+|---|---|---:|---|
+| `--language <code>` / `--request-option language=<code>` | language code | voice prefix | Text frontend language. |
+| `--voice-id <id>` | packaged voice id | `af_heart` | Built-in voice pack. |
+| `--seed <n>` / `--request-option seed=<n>` | integer | random | Decoder noise seed. |
+| `--text-chunk-size <n>` / `--request-option text_chunk_size=<n>` | integer chars | `240` | Long-form chunk size. |
+| `--session-option kokoro_tts.weight_type=<type>` | `native`, `f32`, `f16`, `bf16`, `q8_0` | `native` | Matmul weight storage type. |
+| `--session-option kokoro_tts.conv_weight_type=<type>` | `native`, `f32`, `f16` | `native` | Convolution weight storage type. |
+
+For GGUF packages, leave weight options at `native` unless you are testing a
+conversion or storage policy.
+
+## Conversion
+
+Convert from the official `hexgrad/Kokoro-82M` source checkout:
+
+```bash
+python tools/prepare_kokoro_gguf.py \
+  --source /path/to/Kokoro-82M \
+  --output-dir /path/to/Kokoro-82M-GGUF \
+  --type both \
+  --overwrite
+```
+
+For a fully bundled local multilingual package with eSpeak data and UniDic:
+
+```bash
+python tools/prepare_kokoro_gguf.py \
+  --source /path/to/Kokoro-82M \
+  --output-dir /path/to/Kokoro-82M-GGUF \
+  --type both \
+  --embed-multilingual-resources \
+  --overwrite
+```
+
+The detailed validation notes live in
+[`tests/kokoro_tts/MULTILINGUAL_GGUF.md`](../../tests/kokoro_tts/MULTILINGUAL_GGUF.md).

docs/tts.md

@@ -28,6 +28,7 @@
 | IndexTTS2 | `index_tts2` | `tts` | [IndexTTS](models/index_tts.md) |
 | IndexTTS2.5 | `index_tts2` (variant `2.5`) | `tts` | [IndexTTS](models/index_tts.md) |
 | Irodori-TTS | `irodori_tts` | `tts`, `vdes` | [Irodori-TTS](#irodori-tts) |
+| Kokoro 82M | `kokoro_tts` | `tts` | [Kokoro 82M](models/kokoro_tts.md) |
 | GLM-TTS | `glm_tts` | `tts`, `clon` | [GLM-TTS](#glm-tts) |
 | Inflect Micro v2 | `inflect_v2` | `tts` | [Inflect v2](#inflect-v2) |
 | OuteTTS | `outetts` | `tts`, `clon` | [OuteTTS](#outetts) |

include/engine/models/kokoro_tts/assets.h

@@ -1,6 +1,7 @@
 #pragma once
 
 #include "engine/framework/assets/tensor_source.h"
+#include "engine/framework/assets/resource_bundle.h"
 #include "engine/framework/core/backend_weight_store.h"
 #include "engine/framework/core/module.h"
 #include "engine/framework/io/json.h"
@@ -243,9 +244,6 @@ struct KokoroWeights {
 
     std::shared_ptr<engine::core::BackendWeightStore> store;
 };
-namespace g2p_en {
-class EnglishG2P;
-}
 }
 
 namespace engine::models::kokoro_tts {
@@ -259,20 +257,17 @@ struct KokoroVoicePack {
 };
 
 struct KokoroAssets {
-    std::shared_ptr<struct KokoroPackage> package;
     std::shared_ptr<class MultilingualG2P> multilingual_g2p;
     std::filesystem::path model_root;
+    engine::assets::ResourceBundle resources;
     engine::io::json::Value config;
     std::shared_ptr<const engine::assets::TensorSource> model_weights;
     int64_t context_length = 512;
     std::unordered_map<std::string, int32_t> vocab;
     std::unordered_map<std::string, KokoroVoicePack> voices;
-    std::filesystem::path english_lexicon_dir;
-    std::shared_ptr<const kokoro_ggml::g2p_en::EnglishG2P> english_g2p_us;
-    std::shared_ptr<const kokoro_ggml::g2p_en::EnglishG2P> english_g2p_gb;
 };
 
-std::shared_ptr<const KokoroAssets> load_kokoro_assets(const std::filesystem::path & model_root);
+std::shared_ptr<const KokoroAssets> load_kokoro_assets(const std::filesystem::path & model_path);
 
 std::shared_ptr<const kokoro_ggml::KokoroWeights> load_kokoro_backend_weights(
     const KokoroAssets & assets,

include/engine/models/kokoro_tts/frontend.h

@@ -31,12 +31,6 @@ KokoroFrontendSessionState resolve_kokoro_frontend_session_state(
     const std::optional<runtime::VoiceCondition> & voice,
     const KokoroAssets & assets);
 
-void validate_kokoro_frontend_session_state(
-    const runtime::Transcript & text,
-    const std::optional<runtime::VoiceCondition> & voice,
-    const KokoroFrontendSessionState & state,
-    const KokoroAssets & assets);
-
 KokoroSynthesisInput build_kokoro_synthesis_input(
     const runtime::Transcript & text,
     const KokoroFrontendSessionState & state,