fix

yujonglee · yujonglee · commit 9a0776379870 · 2025-06-20T15:37:54.000+09:00
diff --git a/Cargo.lock b/Cargo.lock
diff --git a/crates/audio-utils/src/lib.rs b/crates/audio-utils/src/lib.rs
@@ -30,16 +30,15 @@ pub trait AudioFormatExt: AsyncSource {
 pub fn i16_to_f32_samples(samples: &[i16]) -> Vec<f32> {
     samples
         .iter()
-        .map(|&sample| sample as f32 / std::i16::MAX as f32)
+        .map(|&sample| sample as f32 / 32768.0)
         .collect()
 }
 
 pub fn f32_to_i16_samples(samples: &[f32]) -> Vec<i16> {
     samples
         .iter()
         .map(|&sample| {
-            let scaled =
-                (sample * std::i16::MAX as f32).clamp(std::i16::MIN as f32, std::i16::MAX as f32);
+            let scaled = (sample * 32768.0).clamp(-32768.0, 32768.0);
             scaled as i16
         })
         .collect()
diff --git a/crates/pyannote/Cargo.toml b/crates/pyannote/Cargo.toml
@@ -14,7 +14,7 @@ reqwest = { workspace = true, features = ["json"], optional = true }
 url = { workspace = true, optional = true }
 
 hypr-onnx = { workspace = true, optional = true }
-knf-rs = { git = "https://github.com/thewh1teagle/pyannote-rs", rev = "e3abad6", package = "knf-rs", optional = true }
+knf-rs = { version = "0.2.9", optional = true }
 
 anyhow = { workspace = true }
 thiserror = { workspace = true }
diff --git a/plugins/local-stt/Cargo.toml b/plugins/local-stt/Cargo.toml
@@ -29,7 +29,7 @@ hypr-chunker = { workspace = true }
 hypr-db-user = { workspace = true }
 hypr-file = { workspace = true }
 hypr-listener-interface = { workspace = true }
-hypr-pyannote = { workspace = true, features = ["local"] }
+hypr-pyannote = { workspace = true, features = [] }
 hypr-whisper = { workspace = true, features = ["local"] }
 hypr-ws-utils = { workspace = true }
 
diff --git a/plugins/local-stt/src/ext.rs b/plugins/local-stt/src/ext.rs
@@ -176,31 +176,34 @@ impl<R: Runtime, T: Manager<R>> LocalSttPluginExt<R> for T {
             .dynamic_prompt("")
             .build();
 
-        let mut segmenter = hypr_pyannote::local::segmentation::Segmenter::new(16000).unwrap();
+        // TODO
+        // https://github.com/thewh1teagle/pyannote-rs/issues/13
+
+        // let mut segmenter = hypr_pyannote::local::segmentation::Segmenter::new(16000).unwrap();
         let segments = segmenter.process(&samples, 16000).unwrap();
 
         let mut words = Vec::new();
 
-        for segment in segments {
-            let audio_f32 = hypr_audio_utils::i16_to_f32_samples(&segment.samples);
-
-            let whisper_segments = model.transcribe(&audio_f32).unwrap();
-
-            for whisper_segment in whisper_segments {
-                let start_sec: f64 = segment.start + (whisper_segment.start() as f64);
-                let end_sec: f64 = segment.start + (whisper_segment.end() as f64);
-                let start_ms = (start_sec * 1000.0) as u64;
-                let end_ms = (end_sec * 1000.0) as u64;
-
-                words.push(Word {
-                    text: whisper_segment.text().to_string(),
-                    speaker: None,
-                    confidence: Some(whisper_segment.confidence()),
-                    start_ms: Some(start_ms),
-                    end_ms: Some(end_ms),
-                });
-            }
-        }
+        // for segment in segments {
+        //     let audio_f32 = hypr_audio_utils::i16_to_f32_samples(&segment.samples);
+
+        //     let whisper_segments = model.transcribe(&audio_f32).unwrap();
+
+        //     for whisper_segment in whisper_segments {
+        //         let start_sec: f64 = segment.start + (whisper_segment.start() as f64);
+        //         let end_sec: f64 = segment.start + (whisper_segment.end() as f64);
+        //         let start_ms = (start_sec * 1000.0) as u64;
+        //         let end_ms = (end_sec * 1000.0) as u64;
+
+        //         words.push(Word {
+        //             text: whisper_segment.text().to_string(),
+        //             speaker: None,
+        //             confidence: Some(whisper_segment.confidence()),
+        //             start_ms: Some(start_ms),
+        //             end_ms: Some(end_ms),
+        //         });
+        //     }
+        // }
 
         Ok(words)
     }
