Common questions, answered plainly.

Yes. Respondents can answer through a voice conversation, watch the corresponding fields fill in, review the captured information, and submit from the same page. Manual controls remain available for fields that need them.

Yes, in both directions. The assistant asks related questions together in one turn rather than one at a time, and anything a respondent volunteers beyond what was asked is matched to whatever fields it covers, in that same turn.

Yes. A respondent can say a correction naturally, such as "actually, make that three", and AmphiVox updates the current field value. The collected fields stay visible for review before submission.

No. Say someone gets halfway through, gives their name and email, then closes the tab. When they come back, those answers are still sitting on the form — so the assistant picks up from the next unanswered question instead of walking them through it again.

The same thing happens when your own site already knows a few details and passes them in: the respondent only gets asked about what is missing. Fields marked PII, and manual-only file, URL, and legal fields, are never filled this way.

Yes. Addresses are checked in their country context, including postal information and incompatible state/country combinations; a recognised postal code can also fill city and state. Phone numbers are validated against the spoken country or the field's configured default region and normalized into E.164 international format.

45+ languages and counting, including 10+ Indian languages such as Hindi, Tamil, Telugu, Kannada, Malayalam, Marathi, Bengali, Gujarati, Punjabi, and Odia. Across the Indian languages people mix with English as a matter of course — Hinglish, Tanglish and the like — and the assistant follows the mix rather than forcing one language. On the worldwide stack, answering in your own language works, and falling back to English for a word or a phrase is fine too. Answers are stored as structured English values.

Yes. Hosted forms and the embedded voice experience are responsive and work in modern mobile browsers. No app download is required, but the respondent must grant microphone permission to use voice.

Start from scratch in the builder, or import screenshots or a PDF. Imported fields remain editable, and AmphiVox shows notes for source content that needs review before publishing.

You can preview the form itself — the fields, their order, and what a respondent sees on the page — before anything goes live.

The voice conversation is not part of that preview. The assistant runs against a published form, so the way to hear it is to publish and open your own form link. If that run was only a test, you can close the form to responses or delete it afterwards.

Common text, contact, numeric, choice, date, address, URL, legal, and file-upload workflows are supported. Available validation and configuration depend on the field type.

Yes. Creators can mark a field as PII/manual-only. The agent does not ask the respondent to say that value, voice tools reject attempts to populate it, and the respondent enters it directly on the visible form.

This is a controlled handoff for configured fields. It should not be described as removing arbitrary sensitive information that a respondent voluntarily speaks into an active microphone session.

Yes. Forms can include file-upload fields, and uploaded files go to the creator's connected Google Drive folder. A form with file fields cannot accept responses until that Drive destination is connected.

Forms can be public or private. Private forms require an authenticated user with access, and creators can maintain an allowed-email list. When sign-in is required, creators can restrict each authenticated respondent to one response.

No. Recordings of a respondent's voice are not saved. Audio is streamed for speech recognition, converted to text, and discarded as the session runs.

The active session state is held in memory while the conversation runs. Diagnostic logs can include conversation text and collected answers and are retained for seven days. Submitted responses follow the retention rules of the selected destination.

You choose: AmphiVox storage, a connected Google Sheet, or a Microsoft Excel workbook. Files collected through file-upload fields are sent to your connected Google Drive folder.

Yes. AmphiVox uses an AI voice assistant to guide the conversation and interpret answers into the form's field structure. Respondent-facing experiences should disclose that AI interaction before the voice session begins.

Explain why you are collecting the information, provide any notices or consent your workflow requires, and link to your own privacy terms where appropriate. The AmphiVox Privacy Policy explains how the platform processes voice, text, submissions, and connected-account data.

This is product guidance, not a substitute for legal advice.

Yes. The JavaScript widget fills mapped inputs on an existing page, and the React package offers component and hook integrations for applications that need more direct state control.

Widget keys carry an approved-domain list. The AmphiVox iframe checks the embedding origin before creating a voice session, which prevents the same key from being used from an unapproved domain. Referrer-policy troubleshooting is covered in the developer docs.

Yes. Both widget integrations support prefill values for the next voice session. Prefill is a complete current snapshot rather than an append-only history, and it must arrive before the session starts.

Yes. Widget callbacks report field updates, session state, speaking state, and transcript events. Exact callback names and value shapes differ by integration and are documented in the relevant widget guide.

Responses can remain in AmphiVox storage or be written to Google Sheets or Microsoft Excel. File uploads use a separately connected Google Drive destination.

You may need to reconnect the provider or correct the destination configuration. Delivery credentials and troubleshooting are managed from authenticated form settings, not the public respondent experience.