Image SEO Optimizer runs on most modern WordPress hosts. Two versions are hard requirements — WordPress 6.1 and PHP 8.0 — and Pro adds one of its own: WordPress 6.5. Everything else on this page is either an AI provider requirement or practical guidance for large media libraries.
Enforced requirements
| Requirement | Minimum | Recommended |
|---|---|---|
| WordPress | 6.1 | 6.5 or newer — Pro requires 6.5 |
| PHP | 8.0 | 8.1 or newer |
| PHP OpenSSL extension | Enabled | Enabled — API keys cannot be saved without it |
The version numbers come from the plugin headers Requires at least: 6.1 and Requires PHP: 8.0. WordPress refuses to activate a plugin whose PHP requirement your server does not meet, and the plugin’s own activation hook re-checks the WordPress version and stops below 6.1. Pro carries Requires at least: 6.5 plus Requires Plugins: infy-ai-image-seo-optimizer, so it does not run without the free plugin under it.
What activation does not check
Activation checks the WordPress and PHP versions and nothing else — no database version, no memory limit, no extension probe. The rows below are practical guidance, not gates. Nothing here blocks the plugin from running.
Practical guidance, not enforced checks
| Feature | What to know |
|---|---|
| MySQL / MariaDB No separate check exists. |
The plugin creates its tables (infyp_changelog, infyp_api_usage, and infyp_rest_log since 1.0.6) through WordPress dbDelta, so it inherits whatever database version your WordPress install already runs on. No database version is read or compared at any point.
|
| PHP memory_limit Not measured by the plugin. |
No minimum is enforced and none is measured. The Debug page prints the current memory_limit and the state of the cURL, JSON, OpenSSL, mbstring, GD and EXIF extensions, so you can read what your host allows rather than guess.
|
| max_execution_time Has to outlast one AI call. |
A single AI request waits up to the plugin’s request timeout, which defaults to 60 seconds and accepts values from 60 to 120. A shorter PHP max_execution_time cuts the request off mid-flight. REST optimize requests raise their own limit to 120 seconds when set_time_limit() is available.
|
OpenSSL is not optional
API keys are encrypted with AES-256-GCM before they reach the database, under a key derived from your site’s AUTH_KEY and SECURE_AUTH_KEY. Without the OpenSSL extension the encryption call returns false, the key is never stored, and the plugin shows an admin notice instead. Rotating those two salts makes stored keys unreadable — you re-enter them once and carry on.
The mbstring extension sits in the same category. AI output arrives in whichever of the 133 locales the language setting offers, and the optimization path truncates prompt text with the multibyte-safe mb_substr() rather than a byte-counting equivalent. The Debug page reports whether the extension is loaded.
Browsers: the admin assets are built for the last two versions of Chrome, Firefox, Safari and Edge, which is the browserslist target the plugin ships. Older browsers are not tested, and Internet Explorer is not supported at all.
AI provider accounts
You need an account with one AI provider — one is enough, and the plugin only calls the provider whose model you selected. On WordPress 7.0 and later you can skip provider accounts entirely and route requests through the AI connectors configured in WordPress itself.
Account: an Anthropic Console account. Generate a key at console.anthropic.com/settings/keys.
Models: Claude 4.5 Haiku, priced at $2.00 per 1M tokens blended. Pro adds Claude 4.5 Sonnet at $9.00 per 1M tokens.
Without a key: on WordPress 7.0+ you can select WordPress AI (Connectors) as the connection method, and core answers the request through a configured connector instead — see AI Connectors.
Account: a Google AI Studio key from aistudio.google.com/app/apikey.
Models: Gemini 3.1 Flash-Lite at $0.56 per 1M tokens blended — the default model in a fresh install — and Gemini 3.5 Flash-Lite at $0.85, added in 1.0.6. Pro adds Gemini 3.1 Pro at $2.50 per 1M tokens. Both 3.1 models carry preview model IDs (gemini-3.1-flash-lite-preview, gemini-3.1-pro-preview), so they change or retire on Google’s schedule, not yours.
Without a key: on WordPress 7.0+ you can select WordPress AI (Connectors) as the connection method, and core answers the request through a configured connector instead — see AI Connectors.
Account: an OpenAI Platform account, which is separate from a ChatGPT subscription. Generate a key at platform.openai.com/api-keys.
Models: GPT-4.1 mini at $0.70 per 1M tokens blended. Pro adds GPT-5 mini at $1.20 per 1M tokens.
Without a key: on WordPress 7.0+ you can select WordPress AI (Connectors) as the connection method, and core answers the request through a configured connector instead — see AI Connectors.
These prices are estimates, not a bill
Each figure is the blended per-million-token price the plugin uses to estimate spend on the API Usage page — one number mixing input and output rates, not the provider’s rate card. Checked against the model registry on 2026-08-12. Provider pricing changes; the number that matters is the one on your provider invoice.
Outbound network access
Your server must reach the AI provider you configured, plus infypress.com when Pro is active for license validation. On managed hosts with strict outbound firewalls, allowlist the endpoints below or AI calls fail with connection timeouts. The plugin only contacts the provider whose model you selected — it does not call the others.
Endpoints to allowlist
All connections are HTTPS on port 443. Image data goes from your server to the AI provider and nowhere else; the only call to infypress.com carries the license key, and only while Pro is active.
api.anthropic.com
generativelanguage.googleapis.com
api.openai.com
infypress.com
Connectors move the call out of the plugin
With WordPress AI (Connectors) selected as the connection method, WordPress core makes the provider request, not the plugin. The hosts above stop applying — allowlist whatever your configured connectors talk to instead, and keep infypress.com if you run Pro. The Test Key buttons still use the three provider hosts, because they test plugin keys.