Local Audio Plus

Obsidian desktop plugin for transcribing local audio links in notes and inserting transcripts.

Automatic scanning is off by default. Opening, saving, or modifying notes will not start transcription unless the user enables Automatic processing in the plugin settings. Users can always start transcription manually from the audio link button, the audio link context menu, the ribbon button, or the command palette.

Supported providers:

• OpenAI Audio Transcriptions
• Tencent Cloud recording file recognition (CreateRecTask with local base64 audio upload)
• iFlytek recording file transcription (/v2/upload + /v2/getResult)

OpenAI normal transcription can use gpt-4o-mini-transcribe for lower cost or gpt-4o-transcribe for higher accuracy. Speaker diarization uses gpt-4o-transcribe-diarize.

Tencent Cloud local upload uses SourceType=1, so the original audio file must be no larger than 5 MB.

iFlytek uses APPID, APIKey, and APISecret from the "录音文件转写大模型" console. In the API documentation, APIKey is also called accessKeyId, and APISecret is also called accessKeySecret. APISecret is signed as the original UTF-8 string; do not Base64 decode it. The default language mode is autodialect; autominor requires manual enablement from iFlytek.

The shared speaker diarization setting is used by supported providers. For iFlytek, enabling it sends roleType=1 for general role separation, and the optional speaker count is sent as roleNum when it is between 1 and 10.

Privacy and network disclosure

Local Audio Plus reads local audio files referenced by notes in your vault and sends the selected audio content to the transcription provider you configure.

Network requests may be sent to:

• OpenAI API, using the configured OpenAI base URL.
• Tencent Cloud ASR.
• iFlytek recording file transcription.
• A user-configured HTTP, HTTPS, SOCKS5, or SOCKS5H proxy, if enabled.

API credentials are stored in this plugin's Obsidian plugin data inside your vault. This plugin does not include telemetry, ads, or analytics.

License

Apache License 2.0. See LICENSE.

发布到 Obsidian 插件市场

发布前确认：

• manifest.json、package.json 和 package-lock.json 里的版本一致。
• GitHub release 的 tag 和 manifest.json 里的 version 完全一致，例如 1.0.4。
• 仓库根目录包含 README.md、LICENSE、manifest.json 和源码。

生成发布文件：

npm run release

这个命令会先执行生产构建，然后生成 Obsidian 插件市场需要上传到 GitHub Release 的独立文件：

dist/main.js
dist/manifest.json
dist/styles.css

同时也会生成离线安装 zip：

dist/local-audio-plus-<version>.zip

在 GitHub 创建 release 时，上传 dist/main.js、dist/manifest.json 和 dist/styles.css。然后到 https://community.obsidian.md 登录 Obsidian 账号，连接 GitHub 账号，在 Plugins 里提交这个仓库。

发布离线安装包

生成离线发布 zip：

npm run release

这个命令会先执行生产构建，然后生成：

dist/local-audio-plus-<version>.zip

<version> 来自 manifest.json 里的版本号，例如 1.0.4 会生成：

dist/local-audio-plus-1.0.4.zip

发布 GitHub release 时请创建与 manifest.json 版本完全一致的 tag：

git tag 1.0.4

zip 内结构如下：

local-audio-plus/
  main.js
  manifest.json
  styles.css
  README.md
  install.ps1
  install.sh

其中：

• install.ps1：Windows PowerShell 自动安装脚本
• install.sh：Linux 自动安装脚本
• README.md：给离线安装用户看的说明

安装脚本会读取 Obsidian 本地配置，优先选择当前打开的 vault；如果没有当前打开的 vault，会选择最近打开的 vault。脚本会复制插件文件到：

<vault>/.obsidian/plugins/local-audio-plus/

并把插件 ID 写入：

<vault>/.obsidian/community-plugins.json

Windows 用户解压 zip 后，在 local-audio-plus 文件夹里运行：

Set-ExecutionPolicy -Scope Process Bypass
.\install.ps1

Linux 用户解压 zip 后，在 local-audio-plus 文件夹里运行：

chmod +x ./install.sh
./install.sh