Voice Control

Voice is WhaleLink's primary input. The core is one line: hold to talk, release to send.

1. Basics

• Hold the voice button at the bottom to start talking
• Release to send immediately
• To cancel this input, slide up and away before releasing

The transcription is shown as text first, so you can confirm before sending.

2. Voice input hints

The UI shows the current state (listening, recognizing, recognized). To turn these hints off or adjust them, go to Settings → Voice input hints.

3. What to say

The agent understands complete instructions best. A few practical tips:

• One thing at a time. "Check yesterday's logs" works better than "How is that thing going".
• Name the object and scope. Say which project, file, or machine.
• State the goal first for multi-step work. e.g. "Run the tests first, and commit if they pass" — let it break down the steps.
• Say proper nouns more slowly. Names, paths, and commands are recognized better at a slower pace.

4. When to type

Voice is great for quick commands; type directly when:

• You need to enter code, paths, or commands
• It is noisy, or you cannot speak
• You need to paste existing text

You can mix the two anytime — the input box and voice button share the same area.

5. Inaccurate recognition

• Say it again, clearly pronouncing the key words
• Switch to typing
• Check whether the system granted microphone permission
• Make sure the phone's network is fine (recognition needs connectivity)

If the system has no usable speech recognition, the UI will tell you. Related settings are under Settings → Read Aloud / Voice.

6. Next

• Have the agent read replies aloud → Read Aloud
• Learn about the entry points on the agent page → Agent Page