whispering/whisper_streaming/websocket_client.py

124 lines
3 KiB
Python
Raw Normal View History

2022-09-24 11:46:10 +00:00
#!/usr/bin/env python3
import argparse
import asyncio
from logging import DEBUG, INFO, basicConfig, getLogger
from typing import Optional, Union
import sounddevice as sd
import websockets
from whisper.audio import N_FRAMES, SAMPLE_RATE
from whisper_streaming.schema import ParsedChunk
logger = getLogger(__name__)
def sd_callback(indata, frames, time, status):
if status:
logger.warning(status)
loop.call_soon_threadsafe(q.put_nowait, indata.ravel().tobytes())
async def transcribe_from_mic_and_send(
*,
sd_device: Optional[Union[int, str]],
num_block: int,
host: str,
port: int,
) -> None:
uri = f"ws://{host}:{port}"
with sd.InputStream(
samplerate=SAMPLE_RATE,
blocksize=N_FRAMES * num_block,
device=sd_device,
dtype="float32",
channels=1,
callback=sd_callback,
):
async with websockets.connect(uri, max_size=999999999) as ws: # type:ignore
idx: int = 0
while True:
async def g():
return await q.get()
2022-09-24 12:50:04 +00:00
logger.debug(f"Loop #: {idx}")
segment = None
2022-09-24 11:46:10 +00:00
try:
2022-09-24 12:50:04 +00:00
segment = await asyncio.wait_for(g(), timeout=0.5)
2022-09-24 11:46:10 +00:00
except asyncio.TimeoutError:
2022-09-24 12:50:04 +00:00
pass
if segment is not None:
logger.debug(f"Segment size: {len(segment)}")
await ws.send(segment)
logger.debug("Sent")
2022-09-24 11:46:10 +00:00
2022-09-24 12:50:04 +00:00
async def recv():
return await ws.recv()
2022-09-24 11:46:10 +00:00
2022-09-24 12:39:22 +00:00
while True:
2022-09-24 12:50:04 +00:00
try:
c = await asyncio.wait_for(recv(), timeout=0.5)
chunk = ParsedChunk.parse_raw(c)
print(f"{chunk.start:.2f}->{chunk.end:.2f}\t{chunk.text}")
except asyncio.TimeoutError:
2022-09-24 12:39:22 +00:00
break
2022-09-24 12:50:04 +00:00
2022-09-24 11:46:10 +00:00
idx += 1
def get_opts() -> argparse.Namespace:
parser = argparse.ArgumentParser()
parser.add_argument(
"--host",
required=True,
help="host of websocker server",
)
parser.add_argument(
"--port",
type=int,
required=True,
help="Port number of websocker server",
)
parser.add_argument(
"--mic",
)
parser.add_argument(
"--num_block",
"-n",
type=int,
default=160,
help="Number of operation unit",
)
parser.add_argument(
"--debug",
action="store_true",
)
return parser.parse_args()
async def main() -> None:
opts = get_opts()
basicConfig(
level=DEBUG if opts.debug else INFO,
format="[%(asctime)s] %(module)s.%(funcName)s:%(lineno)d %(levelname)s -> %(message)s",
)
global q
global loop
loop = asyncio.get_running_loop()
q = asyncio.Queue()
await transcribe_from_mic_and_send(
sd_device=opts.mic,
num_block=opts.num_block,
host=opts.host,
port=opts.port,
)
if __name__ == "__main__":
asyncio.run(main())