feat: 完成 faster-whisper 转写节点

wov_node_whisper/__init__.py

@@ -0,0 +1 @@
+"""WOV faster-whisper ASR node."""

wov_node_whisper/__main__.py

@@ -0,0 +1,76 @@
+from __future__ import annotations
+
+import json
+import os
+from pathlib import Path
+
+from wov_sdk.models import InvokeRequest, InvokeResponse, NodeManifest
+from wov_sdk.server import run_node
+
+
+def format_timestamp(seconds: float) -> str:
+    total_ms = int(seconds * 1000)
+    hours, remainder = divmod(total_ms, 3600000)
+    minutes, remainder = divmod(remainder, 60000)
+    secs, millis = divmod(remainder, 1000)
+    return f"{hours:02d}:{minutes:02d}:{secs:02d},{millis:03d}"
+
+
+def invoke(request: InvokeRequest) -> InvokeResponse:
+    audio_uri = request.inputs.get("audio_uri")
+    if not audio_uri:
+        return InvokeResponse(status="failed", error="audio_uri is required")
+
+    audio_path = Path(audio_uri)
+    if not audio_path.is_file():
+        return InvokeResponse(status="failed", error="audio file not found")
+
+    try:
+        from faster_whisper import WhisperModel
+
+        model_path = str(
+            request.params.get("model_path")
+            or os.getenv("WHISPER_MODEL_PATH", "large-v3")
+        )
+        device = str(request.params.get("device") or os.getenv("WHISPER_DEVICE", "auto"))
+        compute_type = str(request.params.get("compute_type") or "float16")
+        model = WhisperModel(
+            model_path,
+            device=device,
+            compute_type=compute_type,
+        )
+        segments, _info = model.transcribe(
+            str(audio_path),
+            language=str(request.params.get("language", "ja")),
+            beam_size=int(request.params.get("beam_size", 1)),
+            vad_filter=bool(request.params.get("vad_filter", True)),
+        )
+
+        output_dir = Path(request.output_dir)
+        output_dir.mkdir(parents=True, exist_ok=True)
+        output_path = output_dir / "transcript.srt"
+        lines: list[str] = []
+        for index, segment in enumerate(segments, start=1):
+            lines.extend(
+                [
+                    str(index),
+                    f"{format_timestamp(segment.start)} --> {format_timestamp(segment.end)}",
+                    segment.text.strip(),
+                    "",
+                ]
+            )
+        output_path.write_text("\n".join(lines), encoding="utf-8")
+        return InvokeResponse(status="completed", outputs={"srt_uri": str(output_path)})
+    except Exception as exc:  # noqa: BLE001
+        return InvokeResponse(status="failed", error=str(exc))
+
+
+def main() -> None:
+    manifest_path = Path(__file__).resolve().parent.parent / "node.manifest.json"
+    with open(manifest_path, "r", encoding="utf-8") as f:
+        manifest = NodeManifest.from_dict(json.load(f))
+    run_node(manifest, invoke)
+
+
+if __name__ == "__main__":
+    main()