Voice AI for websites — every question, one page.

Yes — in three main ways: a chatbot that handles typed questions, a text-to-speech reader that narrates pages, or a full voice assistant like Spelo that talks back, searches your content, and takes actions. Voice assistants convert better because they answer faster and feel personal.

The fastest path is a hosted widget you embed with one script tag — Spelo takes about two minutes. Building from scratch requires a speech-to-text model, a large language model, a text-to-speech model, and realtime orchestration — typically weeks of engineering.

For plain read-aloud, browsers include the SpeechSynthesis API for free — but it sounds robotic and does not answer questions. For natural voice plus conversation, use an AI voice widget like Spelo.

The Web Speech API supports voice input in Chrome, Edge, and Safari with a few lines of JavaScript. It handles raw transcription but not conversation. Spelo handles transcription, response, and page actions in one drop-in widget.

Free: browser-native SpeechSynthesis (for read-aloud) and Web Speech API (for input). Those get you robot-voice narration and transcription but no intelligence. For full voice conversation, Spelo starts at $0 for small sites with monthly plans above that.

Yes. Voice input is a major accessibility win for visitors with motor or vision impairments. Spelo also falls back to a text chat for users who prefer it, and respects prefers-reduced-motion for the orb animation.

Yes. The widget runs on iOS Safari, Android Chrome, and every desktop browser. Mobile users benefit most — voice is faster than typing on a small keyboard.

ChatGPT does not know your specific website, products, or policies. Spelo reads your live site content and your connected databases on every conversation, so answers are grounded in your data — not generic AI knowledge.

The widget pauses voice and shows visitors a text-chat fallback until your minutes reset next cycle — or you upgrade in one click. No surprise bills, no silent overage charges.

No. Audio streams directly between the visitor's browser and OpenAI over WebRTC. Our backend only issues ephemeral tokens and proxies database queries you explicitly configure.

It does both. The widget can scroll, open pages, fill filters, highlight elements, and submit forms. You decide which actions are allowed per site.

Thirteen adapters out of the box: Shopify, WooCommerce, BigCommerce, Airtable, Google Sheets, Notion, Supabase, Postgres, MySQL, MongoDB, REST APIs, GraphQL, and CSV. Custom adapters take ~50 lines of JavaScript.

Chatbots wait for typing. Spelo is full-duplex voice. Visitors interrupt, clarify, and the agent keeps the page in sync with the conversation. No dead chat bubble in the corner.

The widget gracefully shows a "chat only" fallback until the next month or you upgrade. It never silently fails or charges you without consent.

A minute is any 60 seconds during which the agent is actively streaming audio to or from a visitor. Idle time, pauses, and text-only messages do not count against your allotment.

No. Your plan includes voice usage through our managed OpenAI infrastructure. Enterprise customers can optionally bring their own key for volume pricing.

Yes. Cancel from the dashboard in one click. You keep access until the end of the current billing period.

Yes. 40% off Pro for registered nonprofits and accredited schools. Email hello@spelo.ai from your institutional address.

Two months free when you pay annually, on Starter and Pro.

Dedicated infrastructure, custom voice training, SSO/SAML, 99.99% SLA, SOC 2 Type II reports, and a named customer success manager. Pricing is scoped to your volume and compliance needs — contact sales@spelo.ai.