From bc6de735d85d49500d7ec1bddad9192401c40b0c Mon Sep 17 00:00:00 2001
From: dimitrievgs <dimitriev.gs@ya.ru>
Date: Wed, 29 Apr 2026 23:55:33 +0300
Subject: [PATCH] basic transcribe for vosk

---
 ...transcribe.py => run-transcribe-parsing.py |  0
 run-transcribe-vosk.py                        | 68 +++++++++++++++++++
 2 files changed, 68 insertions(+)
 rename run-transcribe.py => run-transcribe-parsing.py (100%)
 create mode 100644 run-transcribe-vosk.py

diff --git a/run-transcribe.py b/run-transcribe-parsing.py
similarity index 100%
rename from run-transcribe.py
rename to run-transcribe-parsing.py
diff --git a/run-transcribe-vosk.py b/run-transcribe-vosk.py
new file mode 100644
index 0000000..050deb8
--- /dev/null
+++ b/run-transcribe-vosk.py
@@ -0,0 +1,68 @@
+import sounddevice as sd
+from vosk import Model, KaldiRecognizer
+import json
+import sys
+import numpy as np
+
+# 1. Настройки
+MODEL_PATH = r"D:\Work\Software_Development\Projects\vosk\vosk-model-small-ru-0.22"
+# Попробуйте 48000, если снова будет ошибка -9997
+SAMPLERATE = 48000 
+
+model = Model(MODEL_PATH)
+
+# 2. Поиск ID (Potato использует имена B1 и B2)
+def find_devices():
+    id_gm = None  # VoiceMeeter B1
+    id_pcs = None # VoiceMeeter B2
+    for i, d in enumerate(sd.query_devices()):
+        if "Voicemeeter Out B1" in d['name']: id_gm = i
+        if "Voicemeeter Out B2" in d['name']: id_pcs = i
+    return id_gm, id_pcs
+
+id_gm, id_pcs = find_devices()
+
+if id_gm is None or id_pcs is None:
+    print("Ошибка: Устройства B1 или B2 не найдены!")
+    sys.exit()
+
+# 3. Распознаватели
+rec_gm = KaldiRecognizer(model, SAMPLERATE)
+rec_pcs = KaldiRecognizer(model, SAMPLERATE)
+
+def create_callback(rec, role):
+    def callback(indata, frames, time, status):
+        # Если звук тихий, не мучаем Vosk
+        if np.max(np.abs(indata)) < 0.001:
+            return
+
+        if rec.AcceptWaveform(bytes(indata)):
+            text = json.loads(rec.Result())['text']
+            if text:
+                print(f"[{role}]: {text}")
+        '''else:
+            # Частичный результат для динамики
+            partial = json.loads(rec.PartialResult())['partial']
+            if partial:
+                # Используем \b для удаления "слышит" при финальном результате
+                sys.stdout.write(f"\r[{role} говорит...]: {partial}".ljust(60))
+                sys.stdout.flush()'''
+    return callback
+
+# 4. Параллельный запуск
+print(f"--- ЗАПУСК ТРАНСКРИБАЦИИ TTRPG ---")
+print(f"GM (Микрофон B1) ID: {id_gm}")
+print(f"PCs (Система B2) ID: {id_pcs}")
+
+try:
+    with sd.RawInputStream(samplerate=SAMPLERATE, blocksize=8000, device=id_gm,
+                           dtype='int16', channels=1, callback=create_callback(rec_gm, "GM")), \
+         sd.RawInputStream(samplerate=SAMPLERATE, blocksize=8000, device=id_pcs,
+                           dtype='int16', channels=1, callback=create_callback(rec_pcs, "PCs")):
+        
+        print("\nСистема готова. GM может начинать сессию, PCs могут перебивать.")
+        while True:
+            sd.sleep(1000)
+
+except Exception as e:
+    print(f"\nОшибка: {e}")