feat: add new TTS examples for Kokoro, Vits, and Matcha models

.gitignore

@@ -26,3 +26,5 @@ sherpa-onnx-kws-*
 jniLibs/
 build/
 kokoro-en-*/
+matcha-*
+/

crates/sherpa-rs/Cargo.toml

@@ -37,9 +37,19 @@ cuda = ["sherpa-rs-sys/cuda"]
 directml = ["sherpa-rs-sys/directml"]
 
 [[example]]
-name = "tts"
+name = "tts_kokoro"
 required-features = ["tts"]
-path = "../../examples/tts.rs"
+path = "../../examples/tts_kokoro.rs"
+
+[[example]]
+name = "tts_vits"
+required-features = ["tts"]
+path = "../../examples/tts_vits.rs"
+
+[[example]]
+name = "tts_matcha"
+required-features = ["tts"]
+path = "../../examples/tts_matcha.rs"
 
 [[example]]
 name = "audio_tag"

crates/sherpa-rs/src/lib.rs

@@ -18,19 +18,21 @@ pub mod tts;
 #[cfg(feature = "sys")]
 pub use sherpa_rs_sys;
 
-use eyre::{bail, Result};
+use eyre::{ bail, Result };
 
 pub fn get_default_provider() -> String {
-    if cfg!(feature = "cuda") {
-        "cuda"
-    } else if cfg!(target_os = "macos") {
-        "coreml"
-    } else if cfg!(feature = "directml") {
-        "directml"
-    } else {
-        "cpu"
-    }
-    .into()
+    "cpu".into()
+    // Other providers has many issues with different models!!
+    // if cfg!(feature = "cuda") {
+    //     "cuda"
+    // } else if cfg!(target_os = "macos") {
+    //     "coreml"
+    // } else if cfg!(feature = "directml") {
+    //     "directml"
+    // } else {
+    //     "cpu"
+    // }
+    // .into()
 }
 
 pub fn read_audio_file(path: &str) -> Result<(Vec<f32>, u32)> {
@@ -45,8 +47,44 @@ pub fn read_audio_file(path: &str) -> Result<(Vec<f32>, u32)> {
     // Collect samples into a Vec<f32>
     let samples: Vec<f32> = reader
         .samples::<i16>()
-        .map(|s| s.unwrap() as f32 / i16::MAX as f32)
+        .map(|s| (s.unwrap() as f32) / (i16::MAX as f32))
         .collect();
 
     Ok((samples, sample_rate))
 }
+
+pub fn write_audio_file(path: &str, samples: &[f32], sample_rate: u32) -> Result<()> {
+    // Create a WAV file writer
+    let spec = hound::WavSpec {
+        channels: 1,
+        sample_rate,
+        bits_per_sample: 16,
+        sample_format: hound::SampleFormat::Int,
+    };
+
+    let mut writer = hound::WavWriter::create(path, spec)?;
+
+    // Convert samples from f32 to i16 and write them to the WAV file
+    for &sample in samples {
+        let scaled_sample = (sample * (i16::MAX as f32)).clamp(
+            i16::MIN as f32,
+            i16::MAX as f32
+        ) as i16;
+        writer.write_sample(scaled_sample)?;
+    }
+
+    writer.finalize()?;
+    Ok(())
+}
+
+pub struct OnnxConfig {
+    pub provider: String,
+    pub debug: bool,
+    pub num_threads: i32,
+}
+
+impl Default for OnnxConfig {
+    fn default() -> Self {
+        Self { provider: get_default_provider(), debug: false, num_threads: 1 }
+    }
+}

crates/sherpa-rs/src/tts/kokoro.rs

@@ -0,0 +1,64 @@
+use std::{ mem, ptr::null };
+
+use eyre::Result;
+use sherpa_rs_sys;
+use crate::{ utils::RawCStr, OnnxConfig };
+
+use super::TtsAudio;
+
+pub struct KokoroTts {
+    tts: *const sherpa_rs_sys::SherpaOnnxOfflineTts,
+}
+
+#[derive(Default)]
+pub struct KokoroTtsConfig {
+    pub model: String,
+    pub voices: String,
+    pub tokens: String,
+    pub data_dir: String,
+    pub length_scale: f32,
+    pub onnx_config: OnnxConfig,
+}
+
+impl KokoroTts {
+    pub fn new(config: KokoroTtsConfig) -> Self {
+        let tts = unsafe {
+            let model = RawCStr::new(&config.model);
+            let voices = RawCStr::new(&config.voices);
+            let tokens = RawCStr::new(&config.tokens);
+            let data_dir = RawCStr::new(&config.data_dir);
+
+            let provider = RawCStr::new(&config.onnx_config.provider);
+
+            let model_config = sherpa_rs_sys::SherpaOnnxOfflineTtsModelConfig {
+                vits: mem::zeroed::<_>(),
+                num_threads: config.onnx_config.num_threads,
+                debug: config.onnx_config.debug.into(),
+                provider: provider.as_ptr(),
+                matcha: mem::zeroed::<_>(),
+                kokoro: sherpa_rs_sys::SherpaOnnxOfflineTtsKokoroModelConfig {
+                    model: model.as_ptr(),
+                    voices: voices.as_ptr(),
+                    tokens: tokens.as_ptr(),
+                    data_dir: data_dir.as_ptr(),
+                    length_scale: config.length_scale,
+                },
+            };
+            let config = sherpa_rs_sys::SherpaOnnxOfflineTtsConfig {
+                max_num_sentences: 0,
+                model: model_config,
+                rule_fars: null(),
+                rule_fsts: null(),
+            };
+            sherpa_rs_sys::SherpaOnnxCreateOfflineTts(&config)
+        };
+
+        Self {
+            tts,
+        }
+    }
+
+    pub fn create(&mut self, text: &str, sid: i32, speed: f32) -> Result<TtsAudio> {
+        unsafe { super::create(self.tts, text, sid, speed) }
+    }
+}