Curated's Voice Mode lets you ask questions out loud and hear answers back. It's not just a novelty — there are real workflows where typing is the wrong input method.
When to use voice mode
Voice mode is the right tool when:
- You're in the field — site visits, customer walkthroughs, real estate scouting. You're looking at the place you're asking about; you don't want to look at a keyboard.
- You're driving or commuting (passenger seat — please don't drive while prompting). Hands-free Q&A on the way to a meeting.
- You're demoing the product — voice is the cleanest way to show someone what Curated does. The customer hears the question and the answer simultaneously instead of watching someone type.
- You're brainstorming — speaking out loud lowers the friction of asking a half-formed question. Some of the best prompts start as voice and get refined by text follow-ups later.
Voice mode is NOT great for:
- Long, structured prompts with specific filter values, dates, or jargon — typing is faster and more accurate.
- Reading the answer in detail — voice playback is for the gist; for a 5-paragraph PlaceStory, look at the screen.
- Noisy environments — speech-to-text degrades quickly when there's background chatter.
How to enable it
In any Space:
- Click the microphone icon in the prompt bar (top of the Map)
- The first time: your browser will ask for microphone permission. Allow it.
- The mic icon turns red — Curated is listening.
- Speak your prompt naturally. Pause when you're done; Curated detects end-of-speech automatically.
- The answer plays back through your speakers. The text version also appears in the conversation panel.
To turn voice mode off, click the microphone icon again. You can switch between voice and text mode mid-conversation — context carries over.
Prompt patterns that work in voice
Voice mode rewards conversational prompts more than text mode does. The same patterns from text follow-ups still apply, but with two adjustments:
Speak like you'd brief a colleague
❌ Bad voice prompt:
"Run point-in-polygon analysis on Census Block Groups in Travis County for median income above eighty thousand dollars."
✅ Better voice prompt:
"Show me the parts of Travis County where median income is above eighty thousand."
Curated will use the right layers and the right analysis. You don't need to specify the geometry type.
Use short follow-ups
Voice follow-ups should be ≤ one sentence. The longer you talk, the more chance speech-to-text mis-transcribes a number or a place name.
"Now narrow to ZIP codes above ninety thousand."
"Compare to last year."
"Read me the top three."
Ask Curated to summarize before reading aloud
When the answer is long, ask Curated to summarize before it reads back:
"Give me the two-sentence summary."
Voice mode is much better for the two-sentence summary than for paragraph 7 of a long brief.
Voice and accuracy
A few tips for getting the right transcription:
- Spell out unfamiliar place names if Curated keeps mis-hearing them. "That's M-A-N-O-R, Texas, not Manure."* (Yes, this happens.)
- Speak numbers naturally. "Ten thousand" works as well as "10,000." But for a complicated number like a ZIP code, "seven-eight-seven-zero-one" sometimes parses better than "seventy-eight-thousand-seven-hundred-one."
- Switch to text if a single prompt keeps failing. Mode-switching mid-conversation is fine.
Privacy
Voice mode sends audio to Curated's servers for transcription and processing. Audio is not stored beyond the request lifetime; only the resulting text transcript is logged (same as a typed prompt). See the Privacy Policy for the full data handling details.
Related
- Sign In and Ask Your First Question
- Refining Results with Follow-up Prompts
- Privacy Policy