Accessibility

Last updated: October 2, 2026

Thrifty AI is committed to making its website and services usable by as many people as possible, including people with disabilities. Artificial humans communicate through voice, video and facial expression, so we pay particular attention to people who are deaf or hard of hearing, blind or have low vision, have limited mobility, or have cognitive, speech or neurological disabilities.

1. Scope

This statement applies to the thriftyai.com website and to Thrifty Studio. It also describes the accessibility features of the artificial human conversation interface, including the embeddable widget, that we make available to customers. Customers who embed artificial humans in their own applications using the Thrifty AI APIs and SDKs, or through a managed deployment, control much of the surrounding experience and are responsible for the accessibility of their own applications.

2. Our target

We aim to conform to the Web Content Accessibility Guidelines (WCAG) 2.2 at Level AA for the thriftyai.com website and Thrifty Studio. WCAG 2.2 explains how to make web content more accessible to people with disabilities. We use it as our working standard when designing, building and testing new features, and we work to bring existing features into line with it.

3. Accessibility of artificial human conversations

The artificial human conversation interface is designed with the following measures:

Live captions and transcripts: the artificial human's speech can be shown as live captions during a conversation, and a text transcript is available where the customer has enabled it.
Text chat alternative: users who cannot or prefer not to speak can type to the artificial human instead of using voice, and can read its replies as text.
Camera optional: conversations can be held without turning on a camera, unless the customer's use case specifically requires it.
Keyboard operability: the widget and its controls can be reached and operated with a keyboard, with a visible focus indicator and a logical focus order.
Pause, mute and stop controls: users can pause or stop the artificial human, mute its voice, mute their microphone, and end the conversation at any time.
No unexpected autoplay: the artificial human does not start speaking with audio until the user has started or agreed to start a conversation.
Avoiding flashing content: we design avatar rendering, animations and generated canvas content to avoid flashing more than three times in any one second, and we respect operating-system settings to reduce motion where feasible.
Screen-reader support: controls are given accessible names and roles, and new captions and messages are announced to assistive technologies.
Adjustable presentation: captions and text support browser zoom and text resizing, and interface text and controls are designed to meet WCAG contrast requirements.
Multilingual support: artificial humans can converse, and display captions, in multiple languages, including Indian languages, so that users can communicate in the language they are most comfortable with.
Pace and clarity: customers can configure speaking rate and how long the artificial human waits before responding, which helps users who need more time to speak or process information.

4. Thrifty Studio

We design Thrifty Studio so that customers can build, configure and review artificial humans using a keyboard and assistive technologies. Recordings and scored reports from AI interviews, assessments and roleplays are accompanied by transcripts, so that their content is available in text form.

5. How we work towards accessibility

We consider accessibility during design and include accessibility checks in our development and testing process.
We test with automated tools and with manual review, including keyboard-only use and screen readers.
We provide guidance to customers on configuring their artificial humans accessibly, including enabling captions, offering text chat, and writing clear AI-interaction disclosures.
We review feedback from users and customers and use it to prioritize improvements.

6. Known limitations

Despite our efforts, some parts of our website and services may not yet fully meet our target. In general terms:

Live captions and transcripts are generated automatically and may contain errors, particularly with background noise, strong accents, specialist terminology or less widely supported languages.
The artificial human's facial expressions and gestures are visual, and not all of the meaning they convey is available as text.
Content generated live during a conversation, such as quizzes, games and videos created with Magic canvas, may not always include full text alternatives or captions.
Voice recognition may work less well for people with speech disabilities. The text chat alternative is available in these cases.
Some older pages, documents and third-party components, including meeting platforms such as Google Meet and Zoom when an artificial human joins a call, are outside our full control and may not meet our target.

We are working to address these limitations. If any of them prevents you from using our website or services, please contact us and we will try to provide the information or service you need in another way.

7. Feedback and accessible formats

If you have difficulty using any part of our website or services, would like information in an accessible format such as large print, plain text or a tagged document, or want to suggest an improvement, please contact accessibility@thriftyai.com. To help us respond, please tell us the page or feature concerned, the problem you encountered, and the browser, device and assistive technology you were using.

We aim to acknowledge accessibility requests within two business days and to provide a substantive response, or a timeline for one, within ten business days.

If you are an end user of an artificial human deployed by one of our customers, you may also contact that customer directly, as they control the application in which the artificial human appears. You can also reach us through our Contact page.

8. Changes to this statement

We will review and update this statement as we improve our website and services. The date at the top of this page shows when it was last updated.

1. Scope

2. Our target

3. Accessibility of artificial human conversations

The artificial human conversation interface is designed with the following measures:

Live captions and transcripts: the artificial human's speech can be shown as live captions during a conversation, and a text transcript is available where the customer has enabled it.

Text chat alternative: users who cannot or prefer not to speak can type to the artificial human instead of using voice, and can read its replies as text.

Camera optional: conversations can be held without turning on a camera, unless the customer's use case specifically requires it.

Keyboard operability: the widget and its controls can be reached and operated with a keyboard, with a visible focus indicator and a logical focus order.

Pause, mute and stop controls: users can pause or stop the artificial human, mute its voice, mute their microphone, and end the conversation at any time.

No unexpected autoplay: the artificial human does not start speaking with audio until the user has started or agreed to start a conversation.

Avoiding flashing content: we design avatar rendering, animations and generated canvas content to avoid flashing more than three times in any one second, and we respect operating-system settings to reduce motion where feasible.

Screen-reader support: controls are given accessible names and roles, and new captions and messages are announced to assistive technologies.

Adjustable presentation: captions and text support browser zoom and text resizing, and interface text and controls are designed to meet WCAG contrast requirements.

Multilingual support: artificial humans can converse, and display captions, in multiple languages, including Indian languages, so that users can communicate in the language they are most comfortable with.

Pace and clarity: customers can configure speaking rate and how long the artificial human waits before responding, which helps users who need more time to speak or process information.

5. How we work towards accessibility

We consider accessibility during design and include accessibility checks in our development and testing process.

We test with automated tools and with manual review, including keyboard-only use and screen readers.

We provide guidance to customers on configuring their artificial humans accessibly, including enabling captions, offering text chat, and writing clear AI-interaction disclosures.

We review feedback from users and customers and use it to prioritize improvements.

6. Known limitations

Despite our efforts, some parts of our website and services may not yet fully meet our target. In general terms:

Live captions and transcripts are generated automatically and may contain errors, particularly with background noise, strong accents, specialist terminology or less widely supported languages.

The artificial human's facial expressions and gestures are visual, and not all of the meaning they convey is available as text.

Content generated live during a conversation, such as quizzes, games and videos created with Magic canvas, may not always include full text alternatives or captions.

Voice recognition may work less well for people with speech disabilities. The text chat alternative is available in these cases.

Some older pages, documents and third-party components, including meeting platforms such as Google Meet and Zoom when an artificial human joins a call, are outside our full control and may not meet our target.

7. Feedback and accessible formats

We aim to acknowledge accessibility requests within two business days and to provide a substantive response, or a timeline for one, within ten business days.