Voiceboard Support / 支持
Latest version / 最新版本: 1.0
English
If you need help with Voiceboard, here are the most common issues and how to resolve them.
Setup
- Install Voiceboard from the App Store.
- Open the Voiceboard app and complete the onboarding privacy disclosure.
- Configure your AI provider in Settings → Model. (See AI Provider Setup below.)
- Enable the keyboard: Settings → General → Keyboard → Keyboards → Add New Keyboard → Voiceboard.
- Toggle “Allow Full Access” ON for Voiceboard. (Required for the keyboard to talk to the main app.)
- In any text field (Notes, Mail, Messages…), tap and hold the globe key to switch to the Voiceboard keyboard.
AI Provider Setup (BYOK — Bring Your Own Key)
Voiceboard does not bundle any AI provider API key. You provide your own key for whichever provider you trust.
- OpenAI: Get a key at platform.openai.com → API keys. Paste into Settings → Account → OpenAI API Key.
- Mimo (小米 MiMo): Get a key at platform.xiaomimimo.com. Switch the active model to a Mimo preset in Settings → Model; the baseURL and endpoint are pre-filled for you. Paste your Mimo key.
- Other providers: edit baseURL and endpoint mode in Settings to match your chosen provider’s documentation.
In “Transcribe” mode, no API key is needed; speech-to-text runs entirely on-device.
Troubleshooting
Mic permission denied Settings → Privacy & Security → Microphone → Voiceboard → ON.
Speech recognition asset downloading takes a long time First-time setup downloads ~80 MB of language models. Wi-Fi is recommended. Models are cached for subsequent use.
Keyboard mic does nothing Verify “Allow Full Access” is ON: Settings → General → Keyboard → Keyboards → Voiceboard → Allow Full Access.
“Polish unavailable. Inserted raw transcript.” Your AI provider’s API call failed. Check your API key in Settings → Account, your network connection, and your provider’s status page.
“Parameter rejected: max_output_tokens. Check Models settings.” Your active model has a parameter the provider doesn’t accept. Edit the model in Settings → Model to remove the offending parameter.
Background mic always on That’s intentional — by default, Voiceboard keeps the mic warm for 5 minutes after each dictation so the next tap is instant. To stop immediately: Settings → Microphone → “Stop background mic now”. To change the timeout window or disable warming: Settings → Microphone.
Privacy
See our Privacy Policy.
Requirements
- iPhone 12 or newer (Neural Engine required for on-device speech)
- iOS 26.4 or later
- “Allow Full Access” enabled for the keyboard extension
- Internet connection only when using AI cleanup features
Contact
Email: alexcopd@gmail.com
For bug reports, please include:
- iPhone model + iOS version
- App version (in Settings)
- Steps to reproduce
- Expected vs actual behavior
We try to respond within 48 hours.
简体中文
如果你在使用 Voiceboard 时遇到问题,这里有最常见问题的解决方法。
设置
- 从 App Store 安装 Voiceboard。
- 打开 Voiceboard 应用,完成 onboarding 隐私说明的浏览。
- 在 Settings → Model 中配置你的 AI 服务商。(参见下方的 AI 服务商配置)
- 启用键盘:设置 → 通用 → 键盘 → 键盘 → 添加新键盘 → Voiceboard。
- 把 Voiceboard 的 “允许完全访问” 切到开。(键盘扩展需要它来与主应用通信)
- 在任何文字输入框(备忘录、邮件、信息…)中,长按地球键切换到 Voiceboard 键盘。
AI 服务商配置(BYOK — 自带 API Key)
Voiceboard 不内嵌任何 AI 服务商的 API Key。请使用你信任的服务商的自有 Key。
- OpenAI:在 platform.openai.com → API keys 处获取 key,粘贴到 Settings → Account → OpenAI API Key。
- Mimo(小米 MiMo):在 platform.xiaomimimo.com 获取 key,在 Settings → Model 中切换活跃模型为 Mimo 预设,baseURL 和 endpoint 会自动填写,然后粘贴你的 Mimo key。
- 其他服务商:在 Settings 中修改 baseURL 和 endpoint mode,以匹配你选定服务商的文档。
在 “Transcribe” 模式 下,无需 API Key;语音转文字完全在本机运行。
故障排除
麦克风权限被拒 设置 → 隐私与安全 → 麦克风 → Voiceboard → 开。
语音识别模型下载耗时长 首次设置会下载约 80 MB 的语言模型。建议使用 Wi-Fi。下载完成后会缓存,后续使用无需再下载。
键盘麦克风按了没反应 检查”允许完全访问”是否开启:设置 → 通用 → 键盘 → 键盘 → Voiceboard → 允许完全访问。
“润色不可用,已插入原始转写” 你 AI 服务商的 API 调用失败。请检查 Settings → Account 中的 API Key、网络连接,以及服务商的状态页面。
“参数被拒:max_output_tokens。请检查 Model 设置” 你的活跃模型有一个服务商不接受的参数。在 Settings → Model 中编辑模型,移除有问题的参数。
后台麦克风一直在线 这是设计如此 — 默认情况下 Voiceboard 在每次口述后保持麦克风预热 5 分钟,以便下次点击立即响应。立即停止:Settings → Microphone → “立即停止后台麦克风”。修改超时窗口或关闭预热:Settings → Microphone。
隐私
请查阅我们的隐私政策。
系统要求
- iPhone 12 或更新机型(端侧语音识别需要神经引擎)
- iOS 26.4 或更高版本
- 键盘扩展启用”允许完全访问”
- 仅在使用 AI 处理功能时需要联网
联系我们
发送 bug 报告时,请附上:
- iPhone 型号和 iOS 版本
- 应用版本(在 Settings 中查看)
- 复现步骤
- 预期行为 vs 实际行为
我们会在 48 小时内回复。