Mercurial

annotate dictation/audio.py @ 281:c57149ad216e default tip

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Copilot-Session: f68442b1-fa8f-46a0-9689-81710613bbd4
author MrJuneJune <me@mrjunejune.com>
date Tue, 18 Aug 2026 22:18:15 -0700
parents 49e9e591c9bb
children
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
275
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
1 from __future__ import annotations
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
2
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
3 from dataclasses import dataclass
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
4 from enum import Enum
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
5
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
6 import numpy as np
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
7
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
8
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
9 class AudioEventKind(str, Enum):
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
10 SPEECH_STARTED = "speech_started"
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
11 PARTIAL_READY = "partial_ready"
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
12 FINAL_READY = "final_ready"
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
13
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
14
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
15 @dataclass(frozen=True)
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
16 class AudioEvent:
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
17 kind: AudioEventKind
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
18 samples: np.ndarray | None = None
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
19
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
20
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
21 class AudioSegmenter:
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
22 def __init__(
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
23 self,
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
24 *,
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
25 sample_rate: int = 16000,
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
26 speech_threshold: float = 0.012,
280
49e9e591c9bb Add persistent dictation, prewarmed WebRTC speech input, Copilot SDK routing, animated conversation lifecycle controls, parking, and architecture coverage.
MrJuneJune <me@mrjunejune.com>
parents: 275
diff changeset
27 silence_ms: int = 400,
49e9e591c9bb Add persistent dictation, prewarmed WebRTC speech input, Copilot SDK routing, animated conversation lifecycle controls, parking, and architecture coverage.
MrJuneJune <me@mrjunejune.com>
parents: 275
diff changeset
28 partial_interval_ms: int = 500,
275
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
29 max_utterance_seconds: int = 30,
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
30 pre_roll_ms: int = 200,
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
31 ) -> None:
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
32 self.sample_rate = sample_rate
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
33 self.speech_threshold = speech_threshold
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
34 self.silence_samples = sample_rate * silence_ms // 1000
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
35 self.partial_samples = sample_rate * partial_interval_ms // 1000
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
36 self.max_samples = sample_rate * max_utterance_seconds
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
37 self.pre_roll_samples = sample_rate * pre_roll_ms // 1000
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
38 self._pre_roll = np.empty(0, dtype=np.float32)
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
39 self._utterance = np.empty(0, dtype=np.float32)
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
40 self._silence_count = 0
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
41 self._last_partial_size = 0
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
42 self._speaking = False
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
43
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
44 @property
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
45 def speaking(self) -> bool:
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
46 return self._speaking
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
47
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
48 def feed(self, samples: np.ndarray) -> list[AudioEvent]:
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
49 chunk = np.asarray(samples, dtype=np.float32).reshape(-1)
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
50 if chunk.size == 0:
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
51 return []
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
52 rms = float(np.sqrt(np.mean(np.square(chunk), dtype=np.float64)))
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
53 is_speech = rms >= self.speech_threshold
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
54 events: list[AudioEvent] = []
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
55
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
56 if not self._speaking:
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
57 if not is_speech:
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
58 self._append_pre_roll(chunk)
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
59 return events
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
60 self._speaking = True
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
61 self._utterance = np.concatenate((self._pre_roll, chunk))
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
62 self._pre_roll = np.empty(0, dtype=np.float32)
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
63 self._silence_count = 0
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
64 self._last_partial_size = 0
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
65 events.append(AudioEvent(AudioEventKind.SPEECH_STARTED))
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
66 else:
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
67 self._utterance = np.concatenate((self._utterance, chunk))
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
68
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
69 self._silence_count = 0 if is_speech else self._silence_count + chunk.size
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
70 utterance_size = self._utterance.size
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
71
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
72 if (
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
73 utterance_size - self._last_partial_size >= self.partial_samples
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
74 and self._silence_count < self.silence_samples
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
75 ):
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
76 self._last_partial_size = utterance_size
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
77 events.append(
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
78 AudioEvent(
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
79 AudioEventKind.PARTIAL_READY,
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
80 self._utterance.copy(),
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
81 )
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
82 )
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
83
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
84 if (
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
85 self._silence_count >= self.silence_samples
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
86 or utterance_size >= self.max_samples
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
87 ):
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
88 final = self._utterance.copy()
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
89 self._reset()
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
90 events.append(AudioEvent(AudioEventKind.FINAL_READY, final))
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
91
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
92 return events
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
93
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
94 def flush(self) -> AudioEvent | None:
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
95 if not self._speaking or self._utterance.size == 0:
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
96 self._reset()
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
97 return None
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
98 final = self._utterance.copy()
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
99 self._reset()
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
100 return AudioEvent(AudioEventKind.FINAL_READY, final)
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
101
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
102 def _append_pre_roll(self, chunk: np.ndarray) -> None:
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
103 self._pre_roll = np.concatenate((self._pre_roll, chunk))
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
104 if self._pre_roll.size > self.pre_roll_samples:
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
105 self._pre_roll = self._pre_roll[-self.pre_roll_samples :]
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
106
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
107 def _reset(self) -> None:
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
108 self._pre_roll = np.empty(0, dtype=np.float32)
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
109 self._utterance = np.empty(0, dtype=np.float32)
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
110 self._silence_count = 0
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
111 self._last_partial_size = 0
78699f810817 Add Qwen3-VL and WebRTC dictation services
MrJuneJune <me@mrjunejune.com>
parents:
diff changeset
112 self._speaking = False