From 72ead71eebc4c6ab7d9507fbf58a05b48c26603c Mon Sep 17 00:00:00 2001 From: Chen Hua Date: Mon, 8 Jan 2024 15:33:55 +0800 Subject: [PATCH] feat: set initial_prompt and vad_parameters in the first message --- whisper_live/server.py | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) diff --git a/whisper_live/server.py b/whisper_live/server.py index 828362f..81ca9f3 100644 --- a/whisper_live/server.py +++ b/whisper_live/server.py @@ -102,7 +102,9 @@ class TranscriptionServer: language=options["language"], task=options["task"], client_uid=options["uid"], - model_size=options["model_size"] + model_size=options["model_size"], + initial_prompt=options["initial_prompt"], + vad_parameters=options["vad_parameters"] ) self.clients[websocket] = client @@ -190,7 +192,9 @@ class ServeClient: multilingual=False, language=None, client_uid=None, - model_size="small" + model_size="small", + initial_prompt=None, + vad_parameters=None ): """ Initialize a ServeClient instance. @@ -218,6 +222,8 @@ class ServeClient: self.language = language if self.multilingual else "en" self.task = task self.websocket = websocket + self.initial_prompt = initial_prompt + self.vad_parameters = vad_parameters or {"threshold": 0.5} device = "cuda" if torch.cuda.is_available() else "cpu" @@ -352,11 +358,11 @@ class ServeClient: # whisper transcribe with prompt result, info = self.transcriber.transcribe( input_sample, - initial_prompt=None, + initial_prompt=self.initial_prompt, language=self.language, task=self.task, vad_filter=True, - vad_parameters={"threshold": 0.5} + vad_parameters=self.vad_parameters ) if self.language is None: