Jialin27 downloadsTranscribe audio recordings and organize them into summaries, action items, and Markdown notes with AI.
Turn voice recordings into searchable transcripts, summaries, action items, and Markdown notes with configurable AI services. RecordNote works with recordings captured by Obsidian Quick Capture on iPhone and audio files available in your vault.
项目地址 / Repository: github.com/0xcjl/RecordNote
Author: Jialin · Email: [email protected] · X: @0xjialin
RecordNote 是一个 Obsidian 第三方插件,用于将录音转写为 Markdown,并生成摘要、待办和建议。项目处于早期测试阶段。
Capture a recording → open its note → transcribe and organize → review or share Markdown.
RecordNote turns audio into timestamped transcripts and separate AI summaries with key points, decisions, action items, and suggestions. It adds links to results in the source note, keeps original audio, processes audio in segments, and saves progress to support retries. You can reorganize an existing transcript while keeping earlier summaries. Markdown export offers the transcript, summary, or a combined file; available sharing and save options depend on the device.
Obsidian Quick Capture handles recording on iPhone. RecordNote processes the saved audio afterward. You can also choose an audio file manually, including an exported voice memo. On Mac, use the same note processing workflow for audio available on that device.
The community directory submission for the recordnote ID is subject to review. For manual installation, download main.js, manifest.json, and styles.css from the GitHub release, place them in .obsidian/plugins/recordnote/ within your vault, restart Obsidian, and enable RecordNote under Community plugins. If upgrading from voice-organizer-local, keep the old plugin files until you have verified that settings and recordings are available in RecordNote; disable the old plugin before enabling the new one. The minimum Obsidian version is specified in manifest.json. Mobile installation may also use a supported plugin synchronization method.
This workflow builds on A-Lin's Quick Capture post and the official iOS guide. iPhone voice capture requires Obsidian 1.15 or later and iOS 26 or later.
Siri and Spotlight can also open Quick Capture. Automatic note titles are available in its capture settings. Labels vary with language and version; widgets are unavailable when Obsidian's Require Face ID option is enabled.
Transcription and text organization can use separate providers and API keys. Transcription supports MiniMax, OpenAI, or a custom OpenAI-compatible audio transcription endpoint. Text organization supports MiniMax, OpenAI, Gemini, Grok, DeepSeek, Claude, or a custom OpenAI-compatible Chat Completions endpoint. Enter the appropriate model ID, choose the correct MiniMax region if applicable, and select keys through Obsidian's secret storage. Custom endpoints require HTTPS. Service capabilities, account access, and pricing vary.
The settings connection test sends fixed sample text to the text provider and may incur a charge; it does not test transcription. Consumer subscription login and local Agent integration are not implemented.
Recordings remain at their original locations. Results and progress use the vault's RecordNote folder for new installations. Existing installations continue using 语音整理 until migrated; existing data remains accessible. The file picker can read an audio file that you select outside the vault; on supported Mac environments, the export dialog can save Markdown outside the vault. Audio is sent to the selected transcription provider and transcript text to the selected text provider. API access may incur charges.
To keep audio out of Obsidian Sync, disable audio synchronization on each device. The plugin does not change synchronization settings or delete previously uploaded recordings. Check other sync services separately. Audio is decoded in device memory; files larger than 100 MB are rejected. Long recordings, background execution, and recovery are subject to device limits. Review output for transcription and AI errors. The latest open-audio-file fix and real long recordings still need iPhone acceptance testing.
If an older version created a 语音整理 folder, back up the entire vault, including device-only audio and progress files, before running 迁移语音整理文件夹到 RecordNote on the mobile device that holds the original recordings. Enable Obsidian’s Automatically update internal links first. The command refuses an existing RecordNote destination, checks file hashes and indexed links, and leaves a migration audit containing file paths and hashes in the plugin folder. If verification reports an error after the folder has moved, keep the files and backup and inspect the audit before trying recovery. Do not run competing sync or editing during migration. Excluded audio and other files remain device-local.
Author: Jialin · Email: [email protected] · X: @0xjialin · GitHub Issues. Licensed under the MIT License.
快速录音 → 打开录音笔记 → 一键转写和整理 → 查看、分享 Markdown。
Obsidian 的 Quick Capture 负责录音;RecordNote 负责后续处理。在 iPhone 上可以从小组件启动录音,也可以在 Mac 上处理仓库中的音频或手动选择录音。生成的结果保留在自己的 Obsidian 仓库中,原录音不因处理而删除。转写和总结分开保存,方便分别查看或导出。
在 Obsidian 命令面板搜索以下命令,或将常用命令加入移动端菜单:
RecordNote 的 recordnote ID 在社区目录的提交需要审核。也可手动安装:下载 main.js、manifest.json 和 styles.css,放入仓库的 .obsidian/plugins/recordnote/ 文件夹,重启 Obsidian 后在第三方插件设置中启用 RecordNote。从 voice-organizer-local 升级时,先停用旧插件,并保留旧插件文件,直到确认新插件的设置和录音均可正常使用。最低 Obsidian 版本见 manifest.json。在移动设备上可通过支持插件同步的方式传递这些插件文件,再按提示重启。
这个用法参考了 阿蔺 A-Lin 分享的 Obsidian 语音速记帖子,并结合 Obsidian 官方 iOS 使用说明。录音和锁屏入口由 Obsidian 提供,RecordNote 负责录音结束后的转写与整理。
Siri 和 Spotlight 也可用于打开 Quick Capture;可在 Quick Capture 的捕获位置设置中把笔记标题设为自动。具体入口名称可能随系统语言和 Obsidian 版本变化。小组件在 Obsidian 开启“需要 Face ID”时不可用,详见官方说明。
转写和文字整理可以使用不同的服务和密钥。选择 MiniMax 时,先选对应区域,在密钥库选择密钥,并填写文字模型 ID;其他服务分别填写对应的密钥与模型。自定义转写接口需兼容 OpenAI Audio Transcriptions,自定义文字接口需兼容 OpenAI Chat Completions,并使用 HTTPS。设置中的“测试文字服务”只测试文字接口,会发送固定测试文字,可能产生费用;它不代表转写接口也已验证。
插件使用 API 密钥;目前没有实现服务商消费级订阅登录或本机 Agent 接入。API 可能收费,套餐是否包含对应接口由服务商决定。默认 MiniMax 文字模型可在设置中更改。
原始录音保留在原位置;新安装时,处理结果和进度保存在仓库的 RecordNote 文件夹。已有旧版文件夹时,在迁移前会继续使用 语音整理,已有数据仍可访问。手动选择录音时可读取用户在文件选择器中指定的仓库外音频;在支持的 Mac 环境中,导出功能可将用户选定的 Markdown 保存到仓库外位置。处理时录音会发送给所选转写服务,转写文字会发送给所选文字整理服务。关闭 Obsidian Sync 的音频同步需要在每台设备的同步设置中配置;插件本身不会关闭同步,其他同步服务也需要单独检查。同步设置变更不会自动删除此前已上传的音频。
长音频会先在设备内解码,文件超过 100 MB 会被拒绝;长录音的内存消耗、后台执行和恢复能力受设备及系统限制。输出可能包含识别或总结错误,请核对重要信息。最新版本对直接打开音频文件的处理入口有本地自动化测试,但尚未完成该修复的 iPhone 真机验收。
旧版如果已有 语音整理 文件夹,请先备份整个仓库,包括只存在于手机上的录音和进度文件,并开启 Obsidian 的“始终更新内部链接”。在保存原录音的移动设备上运行“迁移语音整理文件夹到 RecordNote”。插件会在目标文件夹已存在时停止,并核对文件哈希和已索引链接;插件目录中会保留包含路径和哈希的迁移核对记录。如果文件已移动但核对报错,请保留现有文件及备份,检查记录后再处理。迁移期间避免其他设备同时编辑或同步;同步排除的音频等文件仍仅保存在本机。
运行 node --test audio-chunks.test.mjs missing-audio.test.cjs plugin.test.cjs settings.test.cjs migration.test.cjs 执行本地测试。
作者:Jialin。问题反馈可通过 GitHub Issues 或 邮件 联系;也可访问 X @0xjialin。
本项目采用 MIT License。