added --task parameter to allow translation to english from other languages

tests/test_faster_whisper_translate.py

@@ -0,0 +1,105 @@
+"""Tests for wyoming-faster-whisper"""
+
+import asyncio
+import re
+import sys
+import wave
+from asyncio.subprocess import PIPE
+from pathlib import Path
+
+import pytest
+from wyoming.asr import Transcribe, Transcript
+from wyoming.audio import AudioStart, AudioStop, wav_to_chunks
+from wyoming.event import async_read_event, async_write_event
+from wyoming.info import Describe, Info
+
+_DIR = Path(__file__).parent
+_PROGRAM_DIR = _DIR.parent
+_LOCAL_DIR = _PROGRAM_DIR / "local"
+_SAMPLES_PER_CHUNK = 1024
+
+# Need to give time for the model to download
+_START_TIMEOUT = 60
+_TRANSCRIBE_TIMEOUT = 60
+
+
+@pytest.mark.asyncio
+async def test_faster_whisper() -> None:
+    proc = await asyncio.create_subprocess_exec(
+        sys.executable,
+        "-m",
+        "wyoming_faster_whisper",
+        "--uri",
+        "stdio://",
+        "--model",
+        "base-int8",
+        "--data-dir",
+        str(_LOCAL_DIR),
+        "--task",
+        "translate",
+        "--language",
+        "fr",
+        stdin=PIPE,
+        stdout=PIPE,
+    )
+    assert proc.stdin is not None
+    assert proc.stdout is not None
+
+    # Check info
+    await async_write_event(Describe().event(), proc.stdin)
+    while True:
+        event = await asyncio.wait_for(
+            async_read_event(proc.stdout), timeout=_START_TIMEOUT
+        )
+        assert event is not None
+
+        if not Info.is_type(event.type):
+            continue
+
+        info = Info.from_event(event)
+        assert len(info.asr) == 1, "Expected one asr service"
+        asr = info.asr[0]
+        assert len(asr.models) > 0, "Expected at least one model"
+        assert any(
+            m.name == "base-int8" for m in asr.models
+        ), "Expected base-int8 model"
+        break
+
+    # We want to use the whisper model
+    await async_write_event(Transcribe(name="base-int8").event(), proc.stdin)
+
+    # Test known WAV
+    with wave.open(str(_DIR / "whats_your_name_french.wav"), "rb") as example_wav:
+        await async_write_event(
+            AudioStart(
+                rate=example_wav.getframerate(),
+                width=example_wav.getsampwidth(),
+                channels=example_wav.getnchannels(),
+            ).event(),
+            proc.stdin,
+        )
+        for chunk in wav_to_chunks(example_wav, _SAMPLES_PER_CHUNK):
+            await async_write_event(chunk.event(), proc.stdin)
+
+        await async_write_event(AudioStop().event(), proc.stdin)
+
+    while True:
+        event = await asyncio.wait_for(
+            async_read_event(proc.stdout), timeout=_TRANSCRIBE_TIMEOUT
+        )
+        assert event is not None
+
+        if not Transcript.is_type(event.type):
+            continue
+
+        transcript = Transcript.from_event(event)
+        text = transcript.text.lower().strip()
+        text = re.sub(r"[^a-z ]", "", text)
+        assert text == "how do you call yourself"
+        break
+
+    # Need to close stdin for graceful termination
+    proc.stdin.close()
+    _, stderr = await proc.communicate()
+
+    assert proc.returncode == 0, stderr.decode()

wyoming_faster_whisper/__main__.py

@@ -59,6 +59,12 @@ async def main() -> None:
         "--initial-prompt",
         help="Optional text to provide as a prompt for the first window",
     )
+    parser.add_argument(
+        "--task",
+        default="transcribe",
+        help="Whether to transcribe or translate (default: transcribe)",
+        choices=["transcribe", "translate"],
+    )
     #
     parser.add_argument("--debug", action="store_true", help="Log DEBUG messages")
     parser.add_argument(

wyoming_faster_whisper/handler.py

@@ -38,6 +38,7 @@ def __init__(
         self.model_lock = model_lock
         self.initial_prompt = initial_prompt
         self._language = self.cli_args.language
+        self._task = self.cli_args.task
         self._wav_dir = tempfile.TemporaryDirectory()
         self._wav_path = os.path.join(self._wav_dir.name, "speech.wav")
         self._wav_file: Optional[wave.Wave_write] = None
@@ -71,6 +72,7 @@ async def handle_event(self, event: Event) -> bool:
                     beam_size=self.cli_args.beam_size,
                     language=self._language,
                     initial_prompt=self.initial_prompt,
+                    task=self._task,
                 )
 
             text = " ".join(segment.text for segment in segments)