Push-to-Talk Voice on iOS

Typing on a phone is fine for short replies and terrible for thinking out loud. The Flowly iOS app's push-to-talk voice mode makes the agent feel like a friend on speakerphone — hold the button, talk, listen back, hang up.

What it does

Hold-to-record interaction model (no wake word, no always-listening)
Whisper-based STT runs in-app for speed
Replies arrive as voice (TTS) plus visible transcript on the lock screen
Conversation continues in your shared memory — switch to Desktop later and the context is there
Special voice-mode behaviour: replies are sanitised for speech-readability (no markdown, no code blocks)

What you'll need

Flowly iOS app
Signed in to your Flowly account
A few minutes to calibrate

Setup

Download Flowly from the App Store. Sign in. The app connects to your deployed gateway over the relay — no localhost setup needed.

2. Enable voice mode

In settings, toggle Voice mode. Grant microphone permission.

3. Capture voice memos

Hold the mic button on the home screen. Speak. Release. The agent processes your audio, replies in voice + transcript.

Try:

"Note for tomorrow's standup: I'm blocked on the Stripe webhook issue from yesterday — Erkan was going to look at it but I haven't heard back."

The agent stores it in memory tagged appropriately. Tomorrow morning when you ask "what's on my plate today?", that note surfaces.

4. Use during walks/drives

Voice mode shines when keyboarding isn't an option. Some patterns that work:

Brain dump: "I've been thinking about whether to go SaaS or open-core for Flowly, here's my latest take..."
Spontaneous to-do: "Add to tomorrow's list: call the accountant about the VAT registration."
Quick fact lookup: "What was the agenda for last week's product meeting?"
Memory queries: "What did I save about AI agent benchmarks?"

The agent doesn't read code, links, or technical details aloud — it summarises and tells you to check Desktop for full output.

Tips

Voice mode replies are short. Long answers get truncated to ~30 seconds of TTS with a "see Desktop for the full answer." Use voice for quick exchanges, switch to Desktop for deep work.
Background noise matters. Whisper handles ambient noise well but struggles with crosstalk. In a coffee shop, you'll mishear occasionally; in a quiet car, near-perfect.
Hold to record, release to send. Slip-of-the-thumb is a cancellable gesture; release while sliding off the button.
Don't dictate sensitive content in public. TTS replies are audible to anyone nearby. The transcript on your lock screen is visible too. Use Desktop for confidential.
Bluetooth quality matters. Cheap earbuds clip your speech enough that STT degrades. Good earbuds are a 20% accuracy boost in noisy environments.