[Bug] Silence timeout doesn't close interface and start transcription; causes silent loss of dictated speech.

オープン
  • 何か問題がありますか？
#765 コメント 0 件 リアクション 0 件 担当者 1 名 GitHub で見る

メンテナーはふだん 1 日以内に返信

@LeanBitLab がすでに取り組んでいます。

2026年11月3日 から。

評価

この issue はまだ評価されていません。

説明

bug

This issue involves the LeanType Voice Plugin, but I'm reporting it here because the silence-timeout setting and recording UI are managed through the main keyboard application.

Describe the bug

  1. When the configured silence timeout is reached, audio capture stops (unverified whether it actually does stop, however the result is the same regardless) without providing any clear indication to the user.
  2. The visual waveform does drop to a fraction of its previous amplitude, even when speaking resumes at normal volume. However, this is not a sufficiently noticeable change to meaningfully indicate that recording has stopped, unless the user is already aware of the issue and specifically watching for it. Even at its reduced amplitude, the waveform continues to animate, giving the impression that recording is still in progress.
  3. The recording interface / waveform remains active, does not close, and transcription does not begin until the user selects the "check" icon to close the recording interface and begin transcription, or the "X" icon to cancel recording and transcription. In fact, the recording interface appears to remain active indefinitely until the user selects one of these 2 options / buttons. Only when the user manually stops the recording and confirms it does transcription begin (as far as I'm aware, this is the only way to initiate transcription). The resulting transcription includes only the speech captured before the silence timeout was reached.

The silence timeout therefore fails to:

  1. Provide any indication that it has been reached.
  2. Close the recording interface and stop the waveform animation.
  3. Automatically initiate transcription of the captured audio, as is typical of voice input in other keyboards.

This creates a misleading recording state that can result in substantial, unnoticed loss of dictated speech.

To Reproduce

  1. Set the Voice Plugin's silence timeout to 5 seconds (or any length of time).
  2. Start dictation and speak normally.
  3. Pause for approximately 5 seconds (or previously selected length), then resume speaking.
  4. Observe that recording interface does not close, nor is there any clear indication of the timeout being reached, as the waveform remains active (albeit at a reduced amplitude).
  5. Continue dictating, then manually stop and confirm the recording and transcription.
  6. Observe that the resulting transcription contains only the speech preceding the silence timeout, omitting everything dictated afterward.

Expected behavior

  1. When the silence timeout is reached, recording and the recording interface including the waveform animation should stop as well, clearly indicating that audio is no longer being captured. The interface should not continue displaying an active recording interface and waveform while subsequent speech is discarded.

  2. When the silence timeout is reached and recording stops, transcription should start automatically without the requirement for further user prompting. This is the default / typical behaviour of voice input features in other commonly used keyboards.

Screenshots
None attached. The issue is most apparent during an active recording.

App version & Flavor

  • LeanType: v4.3.1, Offline
  • Voice Plugin: 1.0.1
  • Whisper Base (~57 MB)

Device Info

  • Model: Samsung Galaxy S22 Ultra
  • OS version: Android 16
  • OneUI VersionL 8.0
主要言語
Kotlin
スター
1k
フォーク
59
平均マージ
14時間 9分
マージ済み PR（30日）
56

環境構築

はじめの一歩

  1. issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
  2. 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
  3. リポジトリをフォークし、ブランチを切って変更します。
  4. issue 番号を参照したプルリクエストを送ります。

LeanBitLab/LeanType のほかの issue

LeanBitLab/LeanType の issue をすべて見る

似ている issue

Kotlin の issue をもっと見る

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。