TypelessForm is a drop-in voice widget that site owners add to HTML forms — their visitors speak one sentence and AI fills every field at once. One script tag. Works with React, Vue, Angular, WordPress.
No credit card · React, Vue, Angular, WordPress, plain HTML · 25 interface languages
98.5%: 195 field observations, 27 utterances, 3 forms, en/ru/de, measured 2026-09-21. Speech recognition is not in that number. The method
“Hi, it’s Anna…”
Hi, it’s Anna. Manicure on October 9 at 3 pm. Phone 555‑0147.
Your client checks it and taps Send.
Opens a demo booking form · no sign-up
The four pictures above are a simplified story. Inside, TypelessForm works in three steps: it scans the form, turns the speech into text and puts each value into its field.
Setup in three steps · Is user data secure? · Pricing
TypelessForm fills a form in three automated steps: it scans the page and detects every input field, transcribes the visitor's sentence with OpenAI's speech model, not limited to our 25 interface languages, then maps each value to its matching field with an OpenAI language model — in a single request once the visitor stops speaking, with no backend changes.
TypelessForm processes voice input through a three-stage AI pipeline. First, the widget scans the page and detects all form fields — text inputs, email fields, date pickers, phone fields, dropdowns, and textareas — using label text, name attributes, and placeholder values. Second, when the user clicks the Fill by Voice button and speaks, the audio is transcribed by OpenAI's speech model, which is not limited to those 25 interface languages. Third, the transcription is passed to an OpenAI language model, which extracts named entities and maps each value to the correct field: "My name is Sarah, email sarah@example.com, arriving March 15" maps to first-name, email, and date fields simultaneously. We publish no latency figure: we have not measured it on real devices and networks. No backend changes are required on the embedding site — the widget communicates directly with the TypelessForm API using the site owner's API key.
AI automatically detects all form fields on the page — names, emails, addresses, dates, textareas
Users say their info in any language, any order. "My name is John, email john@example.com, born March 5th"
AI fills all detected fields. User reviews, edits if needed, and submits. You keep full control.
Click the Fill by Voice button, speak naturally, and watch the form fill itself. The demos below show four real-world forms — hotel booking, contact, dental appointment, and a beauty-salon visit booking — each filled from a single spoken sentence. In the hotel booking demo, a visitor says "John Smith, john@example.com, checking in March 20th, checking out March 25th, 2 adults, double room" and TypelessForm fills all 6 fields simultaneously in a single AI pass. The widget works by first transcribing speech via OpenAI's speech-to-text model (speech is not limited to those 25 interface languages), then using an OpenAI language model to extract named entities and map each value to the correct form field — first name, last name, email, dates, guest count, and room type all in a single AI pass. No field-by-field dictation. No clicking between inputs. The same one-shot approach works for your forms: text, email, tel, date, select, and textarea fields are all supported, while passwords, file uploads, CAPTCHA and hidden fields are deliberately left alone.
TypelessForm turns a single spoken sentence into structured JSON and fills every form field at once. The model is speak once, all fields fill at once — no field-by-field dictation. An OpenAI language model parses the utterance into named entities and maps each to its matching input, producing a fully populated, reviewable form before submission.
Users speak naturally. AI extracts structured JSON and fills every field.
"My name is Sarah Chen, email sarah@chen.com, checking in March 15th for 3 nights, double room, non-smoking"
{
"name": "Sarah Chen",
"email": "sarah@chen.com",
"check_in": "2026-03-15",
"nights": 3,
"room_type": "Double",
"smoking": false
}→ 6 form fields fill at once. Zero typing.
"Car accident March 1st, Toyota Camry 2022, front bumper damage"
→ Date, Vehicle, Year, Damage filled"Hi, I'm Maria from Berlin, maria@startup.de, need help with React forms"
→ Name, City, Email, Message filled"Anna Mueller, born Jan 12 1985, allergic to penicillin, taking metformin"
→ Name, DOB, Allergies, Medications filledThe demos below fill four real forms — hotel booking, contact, dental appointment, and a beauty-salon visit booking — from a single spoken sentence each. OpenAI’s speech-to-text model transcribes the speech, is not limited to those 25 interface languages, and an OpenAI language model maps the transcript to every visible field at once, so nothing needs typing or field-by-field clicking.
This is a live demo — please use example data, not real personal or medical details. Anything you speak is sent to our speech-to-text provider for processing.
We keep no voice recordings on our own infrastructure — audio is discarded as soon as the response returns. OpenAI, our speech provider, does not keep the audio. For its own abuse monitoring it keeps the requests and responses of its language-model calls for up to 30 days: the transcript text, the descriptions of the form fields and the filled values. We cannot have it deleted sooner. It does not use this data to train its models. Is user data secure? · Privacy Policy
Liked it? Start freeAdd it to your site →Pricing →
TypelessForm combines six capabilities: voice recognition beyond its 25 interface languages, standard HTML form compatibility, one-shot field mapping, audio handling that keeps no recordings on our servers (OpenAI, our speech provider, does not keep the audio but keeps the text of requests and responses for up to 30 days for abuse monitoring), review-before-submit control, and GDPR-compliant per-key data isolation.
TypelessForm provides six core capabilities for HTML form voice input. Voice recognition is not limited to the 25 interface languages: a visitor can speak another language — English, Spanish, German, French, Arabic, Japanese, and more — while the form fills in the language it was built in. Universal form compatibility means it works with standard inputs, Angular reactive forms, Vue components, React-controlled fields, and WordPress plugins — no refactoring needed. One-shot field mapping fills every detected field simultaneously from a single spoken sentence, rather than field-by-field dictation. We keep no voice recordings on our own infrastructure: audio is discarded as soon as the response returns. OpenAI, our speech provider, does not keep the audio. For its own abuse monitoring it keeps the requests and responses of its language-model calls for up to 30 days: the transcript text, the descriptions of the form fields and the filled values. We cannot have it deleted sooner. It does not use this data to train its models. Audio is sent to OpenAI's speech-to-text model and the resulting text to an OpenAI language model for field mapping. The API runs in the EU (Google Cloud europe-central2), while OpenAI (the speech-to-text model and GPT) processes audio in the United States under Standard Contractual Clauses. Site owners assess this flow against their regulatory scope before enabling on a given form. Review-before-submit gives users full control — all filled fields are editable before submission. GDPR-compliant API key architecture ensures each site owner's data is isolated with per-key rate limits and usage analytics.
Understands natural speech. Users speak in their language, forms get filled in the form's language.
Drop-in script. Auto-detects fields. React, Vue, Angular, WordPress, plain HTML — no changes needed.
Voice is faster than typing on small keyboards. Optimized for touch screens and on-the-go usage.
Passwords, card numbers, SSN, OTP codes — detected and not filled by voice; digits said aloud stay in the audio, so type them. A value already in a field not marked data-ai-private goes to our API with the request that fills the form; we do not store it and do not pass it to OpenAI. GDPR compliant. No voice recordings on our servers; OpenAI does not keep the audio, but for abuse monitoring it keeps the requests and responses of its language-model calls for up to 30 days.
Hints appear next to fields in the user's language. "Say your name", "Tell us your email" — contextual help.
Field detection on the page and one-shot auto-fill from a single voice utterance. No page reload needed.
Adding voice to your forms takes three steps. Get a free API key, drop one script tag onto the page holding your existing form, and the widget auto-detects every field. No backend changes and no form rebuild: it works on React, Vue, Angular, WordPress, and plain HTML.
Simple integration process — no technical expertise required
Sign up via Google in our developer portal; the free API key comes with the account — no credit card, no approval wait.
Paste a single line of HTML before </body>. Works with sites on React, Vue, Angular, WordPress, plain HTML. No build step needed. If your site uses a Content-Security-Policy, just add our domains to it (see FAQ).
A Fill by Voice button appears on your forms. Visitors click, speak naturally, including languages beyond the 25 the interface ships in, and AI fills every detected field automatically.
TypelessForm fits contact, signup, checkout, booking, and multi-language forms. The API runs in the EU (Google Cloud europe-central2); OpenAI's LLM processes audio in the US under Standard Contractual Clauses. It is not built for protected health information, payment-card data, or government ID numbers, since audio leaves the EU for a US-based LLM provider. Passwords and card numbers are not voice-filled.
What TypelessForm fits, and what it does not. You decide if this fits your form.
Contact, signup, checkout, booking, insurance, event registration, accessibility (WCAG 2.2 / EAA / ADA), B2B lead-gen, international multi-language sites.
Forms collecting protected health information (HIPAA / PHI), payment card data (PCI-DSS — use your processor's hosted fields), government-issued ID numbers (passport / national ID / SSN as primary capture), or any regulated field where audio cannot leave the originating jurisdiction. The widget transcribes audio via OpenAI's speech-to-text model and maps fields via GPT — if your data class cannot be processed by a US-based LLM provider, do not use TypelessForm for that form. You decide if this fits your form. PII-safe defaults (passwords, card numbers, SSN, OTP) are not filled by voice.
TypelessForm pricing starts with a free pilot of 200 form fills, then three paid monthly plans: Starter at €29, Professional at €99, and Enterprise at €199, scaling by submission volume and domains. Paid tiers are available today and activated through Purchase via Invoice. No setup fees, cancel anytime, and voice input in 25 interface languages is included.
No setup fees. Cancel anytime. Upgrade when you grow.
Try TypelessForm at no cost
Perfect for small websites
For growing businesses
For large organizations
Before you integrate: TypelessForm adds voice to your forms in three steps with one script tag, recognizes 25 interface languages and more, and does not fill sensitive fields like passwords, card numbers, and SSNs by voice. The questions below cover CSP setup, mobile support, and pricing in detail.
Yes. TypelessForm works with standard HTML forms on any website. Your users speak naturally, and the system fills the corresponding input fields — without requiring changes to your backend or form structure.
Three steps: get an API key, add our script tag to your page, and set the key. Works with npm install and one line of code. No complex setup required.
CSP note: Most sites don't need this. But if your site uses a Content-Security-Policy — don't replace it, just add our domains where your CSP is configured (<meta> tag, firebase.json, nginx.conf, WordPress functions.php, or CDN dashboard). The ... below means your existing domains — keep them!
script-src ... https://ai-form-copilot-eu.web.app; connect-src ... https://europe-central2-ai-form-copilot-eu.cloudfunctions.net; media-src ... blob:;
script-src is needed only for the script-tag install; with npm install the widget is part of your build.
25 interface languages including English, Spanish, French, German, Italian, Portuguese, Russian, Chinese, Japanese, Arabic, Polish, and many more. Voice recognition works seamlessly across all.
Yes. We use encryption, comply with GDPR, and keep no voice recordings on our own infrastructure — audio is discarded as soon as the response returns. OpenAI, our speech provider, does not keep the audio. For its own abuse monitoring it keeps the requests and responses of its language-model calls for up to 30 days: the transcript text, the descriptions of the form fields and the filled values. We cannot have it deleted sooner. It does not use this data to train its models. Passwords, card numbers, SSN and other sensitive fields are not filled by voice; digits a person says aloud stay in the audio, and the descriptions of the fields go to OpenAI for classification. A value already in a field not marked data-ai-private goes to our API with the request that fills the form; we do not store it and do not pass it to OpenAI.
Absolutely. TypelessForm is optimized for mobile devices and tablets. Voice input is especially effective on mobile — speaking is faster than typing on small keyboards.
Contact forms, registration forms, checkout forms, surveys, booking forms, and any custom form. It auto-detects text inputs, emails, phone fields, dates, textareas, and more.
A voice form filler is two systems, and we publish a figure for the one we control. Given a correct transcript, 98.5% of the fields a speaker named are filled correctly — 195 field observations across 27 utterances and 3 forms in English, Russian and German, measured 2026-09-21 on a frozen corpus we publish. Speech recognition is not inside that number: we have not measured it, so we quote no figure for it, and recognition quality depends on the microphone and the noise around it. You always review the filled values before submitting. The method, the corpus and the raw runs.
Free Pilot with 200 fills is available now. Paid plans (checked 3 October 2026) — Starter €29/mo (800 fills), Professional €99/mo (3,000 fills), Enterprise €199/mo with dedicated support — can be purchased via invoice. No setup fees. Purchase via Invoice to get started.
TypelessForm lets a visitor speak in their own language and fill a form written in another. Language is detected automatically, spoken answers map onto your existing fields, and dates, phone numbers and addresses are converted into the format each field expects. It ships 25 interface languages and accepts spoken input beyond them.
Every supported language, the detection-to-normalization pipeline, and how this category of tool compares: Multilingual Voice Form Filling for International Websites (2026).
TypelessForm is the one-shot voice form filling widget for HTML forms. It enables users to fill multiple form fields simultaneously by speaking a single sentence. The widget uses OpenAI's speech-to-text model and GPT for intelligent field mapping, automatically detecting form fields on any web page via DOM analysis.
How visible is TypelessForm inside AI answers? On 13 August 2026 it was present in 11 of 12 checks — three category questions put to ChatGPT, Perplexity, Gemini and Claude — with 11 answers naming it in the text. The one miss was Gemini on the multilingual question: same model, same wording as a July run that did name us, and this time it didn't. That is our own measurement, not a market ranking or a third-party award: Webappski builds both TypelessForm and the open-source tracker that ran the check. The same run scores a Unified Visibility Index of 87 out of 100 — again our own index, from our own tool, down from 92 in July. Six of the twelve answers came back through the engines' own APIs; the three Perplexity answers were read from its app and entered by hand; the three Claude answers came from a fresh, context-free AI session we ran ourselves — not the founder's own claude.ai account, whose saved chat history could otherwise nudge a personalized answer toward brands it already recognizes. Either way, what the tracker records is not the same as what any given person is shown, and it does not cover Google AI Overviews or Copilot. We publish the run rather than the headline: read how the number was measured, or open the raw report.
TypelessForm publishes how the widget works, how it compares to other voice-to-form tools, and its own measured AI-visibility runs as they happen — not marketing copy, but the same data behind the numbers on this page. New posts appear here as soon as they're live.
An accuracy figure means nothing without the sample, the date and the method behind it. Ours used to be a speech vendor’s benchmark quoted as our own, so we removed it and measured instead: a frozen corpus of 27 spoken utterances across three forms and three languages, 195 field observations, three passes. Given a correct transcript, 98.5% of named fields are filled correctly and nothing is invented for fields nobody mentioned. Here is the whole method, including the three defects the first run found and the improvement the measurement threw out.
A visibility score is not a fact until you know which instrument produced it. On 1 September 2026 the tracker Webappski built reported TypelessForm down from 92% to 50% and listed six competitors filling the gap. Four of the five lost answers came from an engine running a model it had never run before, and both answers that produced competitor names were about a different product category entirely. Here is the whole chain, cell by cell — including the part of the movement that was real.
Buyers increasingly ask an AI engine which tool to use before they visit a single website. This is the open, reproducible record of how often the four major engines name TypelessForm when asked about voice form filling — from 33% presence in April 2026 to a peak of 100% on 11 July, easing to 92% on 13 August — and how to read that number honestly when you are choosing a form widget.
Start with 200 free fills. No credit card required.
Get Your Free API Key