can`t start server in windows

[systemoutprintlnhelloworld]

here is the log after this code
from whisper_live.server import TranscriptionServer
server = TranscriptionServer()
server.run("0.0.0.0", 9090)

Traceback (most recent call last):
File "C:\Program Files\WindowsApps\PythonSoftwareFoundation.Python.3.10_3.10.3056.0_x64__qbz5n2kfra8p0\lib\runpy.py", line 187, in _run_module_as_main
mod_name, mod_spec, code = _get_module_details(mod_name, _Error)
File "C:\Program Files\WindowsApps\PythonSoftwareFoundation.Python.3.10_3.10.3056.0_x64__qbz5n2kfra8p0\lib\runpy.py", line 110, in _get_module_details
import(pkg_name)
File "D:\Code\whisper\WhisperLive-0.0.7\main.py", line 2, in
server = TranscriptionServer()
File "D:\Code\whisper\WhisperLive-0.0.7\whisper_live\server.py", line 38, in init
self.vad_model = VoiceActivityDetection()
File "D:\Code\whisper\WhisperLive-0.0.7\whisper_live\vad.py", line 21, in init
self.session = onnxruntime.InferenceSession(path, providers=['CPUExecutionProvider'], sess_options=opts)
File "C:\Users\25813\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0\LocalCache\local-packages\Python310\site-packages\onnxruntime\capi\onnxruntime_inference_collection.py", line 419, in init
self._create_inference_session(providers, provider_options, disabled_optimizers)
File "C:\Users\25813\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0\LocalCache\local-packages\Python310\site-packages\onnxruntime\capi\onnxruntime_inference_collection.py", line 452, in _create_inference_session
sess = C.InferenceSession(session_options, self._model_path, True, self._read_config_from_model)
onnxruntime.capi.onnxruntime_pybind11_state.Fail: [ONNXRuntimeError] : 1 : FAIL : Load model from C:\Users\25813/.cache/whisper-live/silero_vad.onnx failed:C:\a_work\1\s\onnxruntime\core\graph\model.cc:134 onnxruntime::Model::Model ModelProto does not have a graph.

i have installed requirement from client.txt and server.txt,and i can not installed them from setup.sh,cause i am running server and client both in the same windwos system,as well as setup.py ,it error Perhaps my account does not have write access to this directory

and i don`t installed cuda,is this error linked with this or something else?
Lokk forward for your early reply

[makaveli10]

@systemoutprintlnhelloworld can you check if the onnx model exists
C:\Users\25813/.cache/whisper-live/silero_vad.onnx

[systemoutprintlnhelloworld]

it`s an empty file

[makaveli10]

can you try changing the download path here

WhisperLive/whisper_live/vad.py

Line 98 in 410b91d

target_dir = os.path.expanduser("~/.cache/whisper-live/")

and see if that works

[systemoutprintlnhelloworld]

thanks for your reply!
it`s a network problem
i just download it manually and put it in the right place

[systemoutprintlnhelloworld]

here is another question:
why i open a video in broswer and click this extension to record after the server is runing in locahost:9090,the extension's console runs error like
The message port closed before a response was received
or
Cannot access contents of the page. Extension manifest must request permission to access the respective host.
like this picture

i try to run the extension in chrome either
but error as same as edge
by the way,the network column is empty so their is no request between B and S

could u please tell me how to fix it,or is their anything i ignore to present?

[makaveli10]

Just curious does these error cause issues in transcriptions?
Anyway, we fixed it in the latest PR #58 , just do a git pull and that should make these issues go away.

[rpurinton]

FYI the reason this doesn't work on windows is subprocess.run(["wget", "-O"

workaround is download this: https://github.com/snakers4/silero-vad/raw/master/files/silero_vad.onnx

and place it manually in your home directory ~.cache\whisper-live
for example for me the folder is C:\Users\mir4.cache\whisper-live

this might not be the only thing necessary for windows setup though, it's just the first problem i ran into running run_server.py.

notes
i also had problems with the requirements server.txt complaining torch 1 is not available. i'm going to try with torch 2 which i already had installed.

be sure to allow python in windows firewall.

i was able to get the server running and confirmed 9090.

i created a run_client.py

from whisper_live.client import TranscriptionClient

client = TranscriptionClient(
    "127.0.0.1", "9090", is_multilingual=False, lang="en", translate=False
)
client()

i tried speaking and it did display my words, though it was slow. since i'm running on a VM and even though GPU/CUDA passthrough is enabled, i know whisper-faster runs faster in CPU mode on my machine with 4 threads on the tiny.en model.

i am going to dig into see what i need to do to update this stuff on the server. thanks for making this project its exactly what i needed for mine