Question 1

Why do images need alt text?

Accepted Answer

Alt text makes images accessible to screen readers (used by people with vision impairment), provides a fallback when images fail to load, and helps search engines understand image content. WCAG 2.1 considers alt text a legal accessibility requirement on most public websites.

Question 2

How does this generator work?

Accepted Answer

Your image is sent to Google's Gemini vision model with a strict system prompt that follows WCAG best practices. The model returns a concise alt text under 125 characters, an optional longer description for complex images, and detects whether the image is decorative (and should use alt='') or contains readable text.

Question 3

Is my image uploaded or stored?

Accepted Answer

The image is sent to Google's Gemini API for analysis but is not persisted by Tinkr. We don't log images or alt text. Only an IP-based rate-limit counter is kept. See our privacy policy. If your image contains sensitive PII, generate alt text on private copies first.

Question 4

What's the maximum file size?

Accepted Answer

4 MB. Most photos are under that even at high resolution. If your image is larger, downscale it or convert to a more efficient format like AVIF or WebP. The vision model doesn't need pixel-perfect input — a 1024px wide version captures the same content for alt-text purposes.

Question 5

Why under 125 characters?

Accepted Answer

Most screen readers stop reading alt text around 125 characters. Anything beyond is silently truncated for users who actually need accessibility, defeating the purpose. If an image needs more, use a longer description (returned separately) in surrounding text or in a `figcaption`.

Question 6

What is a decorative image?

Accepted Answer

An image that adds no information beyond visual decoration — abstract patterns, divider lines, repeated icons, ornamental flourishes. These should use `alt=''` (empty string) so screen readers skip them entirely. Don't omit the alt attribute — that lets some screen readers fall back to reading the filename.

Question 7

Should I always use the AI's suggestion verbatim?

Accepted Answer

Treat it as a starting point. The model can't know context that isn't visually verifiable — who the people are, what's significant about the moment, brand-specific terminology. Use the 'context' field to give it that information, then edit the output to match your voice and copy guidelines.

Question 8

Can it transcribe text in screenshots?

Accepted Answer

Yes — the `containsText` field returns any readable text the vision model detects in the image (signs, labels, screenshot UI text). Useful when you want to provide a text alternative for an image that's primarily textual content. Watermarks and pure logo text are intentionally not transcribed.

AI Alt-Text Generator

Frequently asked questions

Related tools