Product

Everything SpeakingForm does, in one place.

From 'describe your intake out loud' to 'a signed webhook lands in your CRM' — here's the whole pipeline, honestly labeled by what ships today.

How it works

Four steps from spoken words to signed data.

1

Describe

Talk (or type) the interview you want.

2

Publish

SpeakingForm compiles it into one configuration + a runtime.

3

Render

Share a link, embed a widget, or turn on voice.

4

Receive

Clean, redacted, structured responses land in your systems.

The same compiled configuration drives every channel — the manifest is the deliverable; the UI serves it. Read the architecture

Voice intake

Let them talk. Get a tidy row.

Customers record an answer in their own words; SpeakingForm transcribes, redacts, extracts, and (optionally) replies with synthesized voice — all from the same form. In-browser recording ships today; WhatsApp voice notes are on the roadmap.

Try a voice form

Every answer type, one renderer.

The web and embed renderer supports every field type an interview needs — including a native audio answer.

What the respondent does	Renderer widget
Types a short answer	`text`
Writes a longer answer	`textarea`
Records a spoken answer	`audio_answer`
Picks one option	`single_choice`
Picks several options	`multi_choice`
Answers yes / no	`boolean`
Enters a number	`number`
Picks a date	`date`
Gives a star rating	`rating`
Slides along a scale	`scale`
Uploads a file	`file_upload`

Talk to build

You talk. It builds the form.

Open Talk to build, describe your intake out loud, and watch the form assemble in a live panel beside you — with recommendations you can accept or dismiss, and click-to-edit on any field. About three minutes from 'hello' to a published, shareable form.

Talk to build

Answer-back

The form answers questions — from your documents.

Upload your policies, menus, or specs. When a respondent asks something, SpeakingForm retrieves the answer from your material, cites where it came from, and refuses plainly when it doesn't know. No invented promises.

See it answer

Do you do vegan buttercream?

Yes — we make a coconut-oil vegan buttercream on request, with no dairy or egg.

from: Allergens & substitutions, p.2

Privacy vault

Redaction happens before the model reads a word.

Names, phone numbers, addresses, and card-shaped strings are pulled into a vault and replaced with tokens before any language model reads the text — then rehydrated only into the record you own. Pin the vault to an EU jurisdiction when residency is a requirement.

Read the security posture

Live agent

A live voice agent that listens like a person. Landing

Full-duplex conversation — barge-in so customers can interrupt, and backchannel acknowledgements ('mm-hm', 'got it') so it feels alive. Rolling out on the Business tier.

Book a demo

Integrations

Your data, in your systems — signed.

Responses leave SpeakingForm the way your stack expects them: signed webhooks, plain exports, an embeddable widget, and versioned forms you can pin.

Signed webhooks

Every delivery is signed with HMAC-SHA256, so your endpoint can verify it came from us — not a spoofed payload.

Exports

Pull responses as CSV or NDJSON whenever you need them, from the dashboard or the API.

Embed

One script tag drops an isolated, styled widget into any page — no iframe wrangling.

Versioning

Every form keeps its version history, with an active-version pin so live responses never shift under you.

Read the docs