VoxDrop 支持

最后更新：2026-09-03

VoxDrop 是一款极简的语音笔记工具：录音 → 通过你配置的第三方语音转写服务转成文字 → 经 HTTPS Webhook 投递到你指定的地址。本页介绍如何使用与常见问题。

快速上手

  1. 打开右上角 设置，在“转写服务”里选择服务商并填写凭据（服务地址、API Key、模型等）。
  2. 可选：在“Webhook 投递”里填一个 HTTPS 地址（以及可选的 Bearer 密钥）。
  3. 打开“数据共享与隐私”，确认将发送的数据和实际目标域名，然后点按授权并启用。更换服务商、凭据或地址后需要重新授权。
  4. 回到主界面，点击底部按钮开始录音；再次点击结束。
  5. 录音结束后会在后台自动转写并投递，你可以立即录下一段。每条记录会显示"已转写 / 已投递"状态。

各服务商配置详解

下面按服务商说明“设置里各字段填什么”，以及配置时常见的疑问。所有地址必须使用 HTTPS。

自定义 STT API

这是一个中立的用户自定义 HTTP 接口，不绑定或代售任何服务。服务可用性会遵守当前 App Store 地区的适用要求。

阿里云 Qwen3-ASR（百炼）

建议填法：

常见疑问：

火山引擎豆包（录音文件识别 · 极速版）

建议填法（新版「语音技术」控制台）：

常见疑问：

常见问题

支持哪些语音转写服务？

目前支持三类：用户填写的自定义 STT API（/audio/transcriptions）、阿里云百炼 Qwen3-ASR、火山引擎豆包（Seed-ASR / 录音文件识别）。凭据由对应服务签发；服务可用性可能因 App Store 地区而异。

我的 API Key 安全吗？

你填写的 API Key、Access Token、Webhook 密钥等只保存在本机 iOS 钥匙串（Keychain），不会上传到任何服务器，开发者也无法读取。

为什么需要“数据共享与隐私”授权？

VoxDrop 不会把“填写配置”默认为数据共享许可。首次启用时会列出音频、热词、转写文本等实际发送内容，以及转写服务和 Webhook 的目标域名。只有你主动点按授权并启用后才会发送；拒绝不会产生对应网络请求。更换服务商、凭据或地址会使旧授权自动失效，你也可以随时在设置中撤销。

Webhook 收到的是什么？

每条录音转写成功后，会把结果作为一段 JSON 通过 HTTPS POST 到你填写的地址。

请求头：

请求体字段：

示例（字段按字母排序）：

{
  "created_at" : "2026-08-29T14:03:21.472Z",
  "duration_seconds" : 12.4,
  "id" : "9F3B2C1A-4D5E-6789-ABCD-EF0123456789",
  "model" : "qwen3-asr-flash",
  "source" : "VoxDrop",
  "text" : "周一上午十点的项目评审会，记得提前把演示环境准备好。"
}

说明：地址必须是 HTTPS；只有转写成功且结果非空才会投递；服务端返回 2xx 视为成功，其它状态码或超时（60 秒）会标记为投递失败，可在记录详情里手动重试。未填 Webhook 地址则跳过投递。

怎么使用热词 / 专有名词？

在"设置 → 热词 / 术语"里选择一个 JSON 文档（可放在 iCloud 或其他 App）。每次启动会自动重新加载最新版本，并按所选服务商自动转换为对应接口格式。示例格式：

{
  "context": "可选的自由背景文本",
  "hotwords": [
    { "term": "VoxDrop", "weight": 10, "lang": "en" },
    { "term": "语音识别", "weight": 8 },
    "赛德克巴莱"
  ]
}

转写或投递失败了怎么办？

失败的记录会显示错误状态，进入该记录详情后可手动重试失败的步骤。常见原因：凭据填写有误、模型名不对、Webhook 地址不可达或非 HTTPS。火山引擎的"权限 / 资源未开通"多为 Resource ID 填错，详见上文「各服务商配置详解」。

录音时屏幕为什么保持常亮？

为避免录音中途锁屏中断，录音期间会临时保持屏幕常亮，结束后自动恢复。

隐私

本应用无开发者运营的服务器。经你明确授权后，音频与文字仅发往授权页面列出的转写服务和 Webhook；开发者不会收到这些数据。详见 隐私政策。

联系我们

问题、反馈或建议，请邮件联系：voxdrop@yuruiz.com。我们会尽快回复。

VoxDrop Support

Last updated: September 3, 2026

VoxDrop is a minimalist voice-notes tool: record → transcribe via a third-party speech-to-text service you configure → deliver the text to an endpoint you choose over an HTTPS webhook. This page covers how to use it and common questions.

Getting started

  1. Open Settings (top-right), pick a provider under “Transcription Service,” and enter its base URL, API key, model, and any other required fields.
  2. Optional: under “Webhook Delivery,” enter an HTTPS URL and optional bearer secret.
  3. Open “Data Sharing & Privacy,” review the exact data and destination domains, then tap Allow & Enable. Changing a provider, credential, or destination requires permission again.
  4. On the main screen, tap the button to start recording; tap again to stop.
  5. After recording, transcription and delivery run automatically in the background — you can immediately record the next clip. Each item shows its "Transcribed / Delivered" status.

Provider setup in detail

What to put in each field per provider, plus the questions people most often hit. Every destination must use HTTPS.

Custom STT API

This is a provider-neutral, user-configured HTTP interface. Availability follows requirements applicable to the current App Store storefront.

Alibaba Cloud Qwen3-ASR (Bailian)

Recommended fields:

Common questions:

Volcengine Doubao (audio file recognition · flash)

Recommended fields (new "Speech" console):

Common questions:

FAQ

Which transcription services are supported?

Three kinds today: a user-entered Custom STT API (/audio/transcriptions), Alibaba Cloud Qwen3-ASR, and Volcengine Doubao (Seed-ASR). Credentials are issued by the corresponding service; availability may vary by App Store storefront.

Are my API keys safe?

Credentials you enter (API keys, access tokens, webhook secrets) are stored only in the on-device iOS Keychain. They are never uploaded to any server and are not accessible to the developer.

Why is Data Sharing & Privacy permission required?

VoxDrop does not treat entering configuration as permission to share data. Before first use, it lists the audio, hotwords, transcript, and metadata that may be sent and displays the actual transcription and webhook destination domains. Nothing is sent until you affirmatively tap Allow & Enable. Changing a provider, credential, or destination invalidates the old permission, and permission can be revoked in Settings at any time.

What does my webhook receive?

After a recording is successfully transcribed, the result is POSTed as a single JSON body over HTTPS to your URL.

Request headers:

Body fields:

Example (keys sorted alphabetically):

{
  "created_at" : "2026-08-29T14:03:21.472Z",
  "duration_seconds" : 12.4,
  "id" : "9F3B2C1A-4D5E-6789-ABCD-EF0123456789",
  "model" : "qwen3-asr-flash",
  "source" : "VoxDrop",
  "text" : "Prep the demo environment before Monday's 10am project review."
}

Notes: the URL must be HTTPS; delivery happens only when transcription succeeds with non-empty text; a 2xx response counts as success, while any other status or a timeout (60s) is marked as a failed delivery you can retry from the item's detail. With no webhook URL set, delivery is skipped.

How do I use hotwords / custom terms?

In "Settings → Hotwords", pick a JSON document (it may live in iCloud or another app). It is re-read on each launch and converted to each provider's format automatically. Example:

{
  "context": "optional free-form background text",
  "hotwords": [
    { "term": "VoxDrop", "weight": 10, "lang": "en" },
    { "term": "语音识别", "weight": 8 },
    "Seediq Bale"
  ]
}

Transcription or delivery failed — what now?

Failed items show an error state; open the item's detail to retry the failed step manually. Common causes: wrong credentials, incorrect model name, or a webhook URL that isn't reachable or isn't HTTPS. Volcengine "resource not granted" errors are usually a wrong Resource ID — see "Provider setup in detail" above.

Privacy

The App has no developer-operated backend. After your explicit permission, audio and text are sent only to the transcription service and webhook shown on the permission screen; the developer does not receive them. See the Privacy Policy.

Contact

Questions, feedback, or suggestions: email voxdrop@yuruiz.com. We'll get back to you as soon as we can.