gitea/conjurer
1
0
Fork 0
mirror of https://github.com/migatu/conjurer.git synced 2026-07-19 00:02:11 +00:00
Code Issues Packages Projects Releases Wiki Activity

asynv

Browse Source
This commit is contained in:
mtuszowski
2024-09-20 16:26:01 +02:00
parent f1d73fc0a1
commit 26c34c6fa9
1 changed files with 22 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files
voice_recognition.py
+22 -31
View File
@@ -5,7 +5,6 @@ import assemblyai as aai
import discord import discord
from discord.ext import commands, voice_recv, tasks from discord.ext import commands, voice_recv, tasks
from discord.opus import Decoder as OpusDecoder from discord.opus import Decoder as OpusDecoder
from queue import Queue, Empty
import time import time
import threading import threading
# Replace with your API key # Replace with your API key
@@ -127,11 +126,10 @@ class SRBuffer(voice_recv.AudioSink):
class Transcriber(commands.Cog): class Transcriber(commands.Cog):
def __init__(self, bot, worker, queue): def __init__(self, bot):
self.bot = bot self.bot = bot
self.threads = [] self.threads = []
self.comm_queue = queue self.comm_queue = asyncio.Queue()
self.worker = worker
self.wsink = SRBuffer(self.comm_queue) self.wsink = SRBuffer(self.comm_queue)
@commands.hybrid_command(name="transcribe") @commands.hybrid_command(name="transcribe")
@@ -151,7 +149,7 @@ class Transcriber(commands.Cog):
logger.info(self.bot.voice_clients) logger.info(self.bot.voice_clients)
self.check_data.start() self.check_data.start()
logger.info("Start worker!~!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!") logger.info("Start worker!~!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!")
self.worker.start() asyncio.run(self.transcribe_output_queue())
vc.listen(self.wsink) vc.listen(self.wsink)
@@ -196,35 +194,28 @@ class Transcriber(commands.Cog):
self.comm_queue.put(stop_token) self.comm_queue.put(stop_token)
await ctx.voice_client.disconnect() await ctx.voice_client.disconnect()
async def transcribe_output_queue(self):
#this needs to be async and transcript fun awaited logger.info("Transcript worker start")
def transcribe_output_queue(queue): config = aai.TranscriptionConfig(speaker_labels=True, language_code="pl")
logger.info("Transcript worker start") transcriber = aai.Transcriber()
config = aai.TranscriptionConfig(speaker_labels=True, language_code="pl") while True:
transcriber = aai.Transcriber() item = self.queue.get()
while True: if item.type == "STOP":
item = queue.get() logger.info("Queue ended")
if item.type == "STOP": break
logger.info("Queue ended") elif item.type == "send_file":
break transcript = ""
elif item.type == "send_file": for iter in item.data:
transcript = "" transcript = transcriber.transcribe(iter, config=config)
for iter in item.data: for utterance in transcript.utterances:
transcript = transcriber.transcribe(iter, config=config) logger.info("%s : %s", item.user, utterance.text)
for utterance in transcript.utterances: elif item.type == "user_cleanup":
logger.info("%s : %s", item.user, utterance.text) logger.info("User %s disconnected - cleanup action")
elif item.type == "user_cleanup": logger.info("End transcript worker")
logger.info("User %s disconnected - cleanup action")
logger.info("End transcript worker")
async def setup(bot): async def setup(bot):
logger = logging.getLogger("discord") logger = logging.getLogger("discord")
logger.info("Loading voice transcribed module phase one") await bot.add_cog(Transcriber(bot))
worker = threading.Thread(target=transcribe_output_queue)
logger.info("Loading voice transcribed phasetwo")
queue = Queue()
logger.info("Loading voice transcribed module phase three")
await bot.add_cog(Transcriber(bot, worker, queue))
logger.info("Loading voice transcribed module done") logger.info("Loading voice transcribed module done")
# 1. zapisuj kwestie człowieka do pliku w którym będzie można stwierdzić kto co powiedział (callback z basicaudio + write z wavesinka). Zamykaj plik i wysyłaj do transkrypcji w momencie ciszy dłuższej niż 0.5s. Jeśli człowiek nadaje cały czas dawaj sygnał że nie można go transkrybować - i wyłaczaj zapis. # 1. zapisuj kwestie człowieka do pliku w którym będzie można stwierdzić kto co powiedział (callback z basicaudio + write z wavesinka). Zamykaj plik i wysyłaj do transkrypcji w momencie ciszy dłuższej niż 0.5s. Jeśli człowiek nadaje cały czas dawaj sygnał że nie można go transkrybować - i wyłaczaj zapis.