Documentation

Talk to Action Documentation

Last updated: July 21, 2026

1. What is Talk to Action?

Talk to Action is a voice-first capture app for turning spoken thoughts into usable notes, tasks, shopping items, calendar drafts, AI prompts, and hand-offs to other apps.

The core idea is simple:

Speak it. AI organizes. You get things done.

Talk to Action is not meant to replace your note app, task manager, calendar, or AI assistant. It is the fast entry point before those tools. You open the app, speak naturally, review the text, choose or confirm the capture type, and send it to the right destination.

Talk to Action is a voice-first capture and routing app. Instead of deciding where a thought belongs before you record it, you capture first and route it afterward.

2. Core principles

Talk to Action is designed around five principles.

1. Fast capture first
   The app should feel light. Opening the app, recording, editing, and sending should take as few steps as possible.
2. Capture should not wait for classification
   Speech is converted to text and preserved as a draft before any routing decision becomes final. Classification is treated as a helpful suggestion, not a point of failure.
3. Local-first and offline-friendly
   Speech recognition uses the operating system speech input where possible, and capture/routing logic is designed to work without a server-side AI pipeline. Some destination apps may use the network after the text is handed off to them.
4. Multiple destinations, one capture
   A single spoken input can be sent to one or more destinations. This is the main difference from a simple voice memo app.
5. Global and multilingual by design
   The app is not only for Japanese users. English and Japanese are the first target languages, with future expansion to additional languages.

3. Current product scope

The current Android beta focuses on the following scope:

• Voice-to-text capture using the Android system speech recognizer.
• Editable draft text after recognition.
• Capture type selection: Note, Task, and Shop.
• Destination actions from an Actions screen.
• Obsidian hand-off with category-specific routing options.
• Google Keep / Android share hand-off.
• Clipboard copy.
• Recent capture history.
• Settings persistence across app restarts.
• English and Japanese usage as the initial practical language set.

Planned and experimental destinations include AI apps, task managers, calendar apps, team chat, and webhook automation. Availability depends on the installed apps on the device and the connection method supported by each service.

4. Main workflow

The basic workflow is:

1. Capture — tap the microphone and speak.
2. Transcribe — Android converts your speech to editable text.
3. Review — confirm or fix the text and capture type.
4. Choose an action — use the default send action or open the destination list.
5. Send — hand the text to Obsidian, Keep, Clipboard, an AI app, a task app, calendar, or another destination.
6. Reuse history — refer to recent captures when needed.

In practice:

Open Talk to Action
→ Tap and speak
→ Review the draft
→ Choose Note / Task / Shop
→ Send to the selected destination

5. Capture screen

The Capture screen is the main screen. It is intentionally simple because the primary job is to capture a thought before it disappears.

Typical elements:

• Large microphone / primary action button.
• Current draft text.
• Capture type buttons: Note, Task, Shop.
• Send action button.
• Shortcut buttons for common text insertion.
• Access to recent history.
• Swipe access to the destination Actions screen.

Recording behavior

1. Tap the microphone button.
2. Speak naturally.
3. The system speech recognizer listens and returns text.
4. The text appears in the draft field.
5. You can edit the text before sending.

If speech recognition fails or produces imperfect text, the capture is still editable. Talk to Action treats the recognized text as a draft, not as a final result.

6. Draft editing

After transcription, you can edit the draft directly. Use this to:

• Correct recognition mistakes.
• Add missing context.
• Remove filler words.
• Change Japanese/English wording.
• Turn a raw thought into a clearer task or note.

The app also includes quick input helpers, such as date, time, symbols, and clear actions, so small corrections can be made quickly.

7. Capture types

Talk to Action uses three simple capture types in the initial product.

The type is not meant to be perfect. It is a routing hint. You can change it manually before sending.

Note

Use Note for anything you want to remember or develop later:

Capture the idea before it disappears.

Common use cases:

• Project ideas.
• Reading notes.
• Quick meeting notes.
• Personal observations.
• Ideas to expand later in Obsidian or an AI app.

Task

Use Task for something you need to do:

Check the Google Play registration steps tomorrow morning.

Common use cases:

• Follow-up items.
• Personal tasks.
• Project todos.
• Work items to send to a task manager.

Talk to Action does not need to fully understand dates in the first release. If you say “tomorrow,” that word remains in the text so your destination app or later review process can handle it.

Shop

Use Shop for buying items:

Buy milk, eggs, and bread.

Common use cases:

• Grocery items.
• Household supplies.
• Things to remember while shopping.

8. Destination Actions screen

Talk to Action separates frequent sending from detailed configuration.

• The main screen is for capturing and editing.
• The Actions screen is for running destination actions.
• Destination-specific settings are accessed from each destination’s edit area.

The Actions screen keeps sending and configuration separate:

Main screen
  → swipe left
Actions screen
  → tap a destination to send
  → edit a destination when configuration is needed

Only enabled destinations should be shown in the Actions screen. This keeps the main capture experience clean even as more integrations are added.

9. Default send behavior

The default mapping is:

You can override the destination for a specific draft. Classification is only the initial suggestion.

10. Supported and planned destinations

The Android beta documentation currently covers the following destination families.

Some services are intentionally not prioritized for direct API integration in the initial product because they require OAuth, server-side secrets, workspace configuration, or a high-friction setup. Talk to Action prefers simple share-based hand-off first, then deeper integrations only where they can be implemented safely.

11. Obsidian setup

Obsidian is the primary knowledge destination.

Basic idea

Talk to Action sends text to Obsidian. Obsidian remains the place where you organize, link, review, and develop your knowledge.

A typical Obsidian setup uses different routing rules per capture type:

Obsidian routing options

The app design supports category-specific Obsidian routes:

• Route per type: Note / Task / Shopping.
• Target kind: Daily note, existing file, or new file.
• Append or prepend behavior.
• Optional heading targeting.
• Insert after a heading or at the end of a heading section.
• Line prefix such as no prefix, - , or - [ ] .
• Optional current time prefix for Note captures.

Standard URI vs Advanced URI

Basic file append can be handled by Obsidian URI. More precise heading insertion may require the Obsidian Advanced URI plugin on the Android device.

Use the simpler route first. Enable heading-specific behavior only if your Obsidian mobile setup supports it.

Example Obsidian configurations

Daily notes for ideas

Type: Note
Target: Daily note
Position: Append
Prefix: -
Time prefix: On

Result idea:

- 09:15 Capture the idea before it disappears.

Task inbox

Type: Task
Target: Task.md
Heading: Inbox
Position: Section end
Prefix: - [ ]

Result idea:

- [ ] Check the Google Play registration steps tomorrow morning.

Shopping list

Type: Shop
Target: Shopping.md
Heading: Buy
Position: Section end
Prefix: -

Result idea:

- Milk, eggs, and bread

12. Google Keep setup

Google Keep is useful for quick mobile lists, especially shopping.

Recommended use:

• Enable Google Keep as a destination.
• Use it as the default or common destination for Shop captures.
• If direct Keep hand-off is unavailable, use the Android share sheet fallback.

Because Keep is controlled by Google’s app behavior, Talk to Action may open a share screen instead of silently creating a note. Confirm and save inside Keep when required.

13. Clipboard destination

Clipboard is the safest universal fallback.

Use Copy when:

• The desired destination is not supported yet.
• A share action fails.
• You want to paste the text manually.
• You are working offline and want to avoid external app behavior.

The clipboard destination does not require login, network access, or another app.

14. AI app destinations

Talk to Action can be used as an entry point into AI apps.

Instead of opening ChatGPT, Claude, or Gemini and typing a prompt, you can speak the idea first, review it, then send it to the AI app.

Good uses:

• “Turn this rough thought into a clear message.”
• “Help me break this into next actions.”
• “Summarize this idea.”
• “Translate this into English.”
• “Challenge this plan.”

Current expected behavior is hand-off, not automatic response retrieval. Talk to Action sends the draft text to the AI app where supported; the AI conversation happens in that app.

15. Calendar and task destinations

Calendar and task destinations are best for spoken commitments:

Schedule a meeting with Tanaka next Tuesday at 10.

Follow up with the designer about the landing page.

In the initial design, calendar creation should open the system calendar event creation screen. This avoids silently creating incorrect events and keeps the user in control.

Task manager integrations may start with Android share actions and later move to official APIs where authentication can be handled safely.

16. Webhook destination

Webhook support is for advanced users and automation workflows.

A webhook can send the draft as an HTTPS POST to:

• Zapier.
• Make.
• IFTTT.
• A custom API.
• An internal workflow endpoint.

Typical payload fields may include:

{
  "text": "Draft text",
  "type": "note",
  "createdAt": "2026-07-21T15:41:00+09:00",
  "source": "Talk to Action"
}

Use webhook only with endpoints you trust. If the endpoint is external, the capture text leaves your device.

17. Settings

Settings are for app-level configuration and destination enablement.

Common settings include:

• Theme.
• Enabled destinations.
• Default destination / default send action.
• Destination-specific edit screens.
• Obsidian route settings.
• Google Keep behavior.
• Clipboard format.
• Optional experimental features.

The guiding rule is: Settings configure; Actions execute.

18. History

Talk to Action keeps recent captures so you can recover from accidental navigation, failed hand-off, or later review.

History is useful when:

• You sent to the wrong destination.
• You want to resend a draft.
• You need to copy a previous capture.
• You want to verify what was captured.

Future paid/pro features may include deeper history search, filters, resend controls, daily review, and longer retention.

19. Privacy and offline behavior

Talk to Action is designed to minimize server dependency.

Important points:

• The app’s own capture and routing logic is local-first.
• System speech recognition is provided by the operating system; exact offline availability depends on device, language, and OS settings.
• Obsidian and Clipboard can work without Talk to Action sending text to a custom server.
• Cloud destinations such as Google Keep, AI apps, Slack, or webhooks may transmit text according to those apps or services.
• Webhook destinations explicitly send text to the configured URL.

If privacy is critical, use Obsidian or Clipboard routes and verify your device’s speech recognition settings.

20. Troubleshooting

The microphone does not start

• Check Android microphone permission.
• Confirm that speech recognition is available on the device.
• Restart the app and try again.
• Check whether another app is using the microphone.

Transcription is wrong

• Edit the draft before sending.
• Try shorter phrases.
• Speak closer to the microphone.
• Confirm the device language and keyboard/speech settings.

Obsidian does not receive the text

• Confirm Obsidian is installed on the Android device.
• Check the vault name and target file settings.
• Start with a simple append-to-file route.
• If using heading-specific insertion, confirm Advanced URI support is installed and configured.

Google Keep does not save automatically

• This can be normal. Keep may open a share or confirmation screen.
• Confirm the note inside Keep.
• Use Clipboard if Keep’s share behavior is unreliable.

A destination is missing from Actions

• Open Settings and enable that destination.
• Confirm the destination app is installed.
• Some destinations are experimental or future integrations.

Sending fails

• The draft should remain available.
• Copy it to Clipboard as a fallback.
• Try a simpler destination such as Obsidian append or Clipboard.

21. Beta feedback

When reporting feedback, include:

• Device name and Android version.
• Spoken language.
• What you said, if you can share it.
• The capture type selected.
• Destination used.
• Expected result.
• Actual result.
• Whether the draft remained in history.

Short reports are welcome. The most useful feedback is about moments where capture felt slow, confusing, or unsafe.

22. Roadmap overview

The planned direction is:

1. Stabilize Android MVP.
2. Improve capture, edit, routing, and destination settings.
3. Expand destinations carefully.
4. Add freemium limits and Pro features.
5. Explore BYOK AI features for users who want cloud AI enhancements.
6. Build iOS version with feature parity where possible.
7. Explore watch-based capture for Android Watch and Apple Watch.

Future Pro candidates include:

• Multiple simultaneous destinations.
• Longer history.
• Resend from history.
• Category-specific routing rules.
• Search and filters.
• Daily review.
• Custom categories.
• Profiles for work/personal contexts.
• Widgets and shortcuts.
• Send queue and retry.
• Watch capture.