Use a dedicated provider for images (Vision)

What it is

AISA can write image alt-text and titles with AI, and it does this best when the model can actually see the image — this is called Vision. Because image work can use a different account or model than your text work, AISA lets you set a dedicated image provider: one provider can handle your SEO text/meta while another handles images. With Vision on, the AI looks at the real picture and describes what is shown; with Vision off, alt-text is inferred only from context (filename, title, page).

Where to find it

AISA → Settings → API Keys, in the Images & Vision box. There you choose the dedicated image provider (only Vision-capable providers are listed), pick its vision model, and toggle Vision active on or off. Image generation itself is triggered from 1-Click SEO ("Images (alt-text)"), the Images tab in Bulk, or the floating panel.

Requirements & tier

Requires a Vision-capable provider: Claude, OpenAI or Gemini. DeepSeek, Mistral, Grok and Perplexity have no vision models and are excluded from the image-provider list.
A valid key for the chosen image provider. It can be a different account from your text provider.
Works on BYOK tiers; on Cloud the server handles vision behind credits.

How to use

Add and enable a Vision-capable provider (Claude, OpenAI or Gemini) in Settings → API Keys.
Open the Images & Vision box and choose that provider as the dedicated image provider.
Pick its vision model (a cheaper vision model is usually fine for alt-text).
Leave Vision active on so the AI truly analyses each image; turn it off only if you want context-only alt-text.
Save.
Generate alt-text from 1-Click SEO (tick "Images (alt-text)"), the Bulk Images tab, or the editor panel. AISA writes alt and title on the attachment and on the in-content <img> tags, and sets a social (OG/Twitter) image from the featured image if one is missing.

Questions people ask

Why can't I pick DeepSeek for images? DeepSeek has no vision model, so it is excluded from the image-provider list. Use Claude, OpenAI or Gemini.
What does Vision off do? Alt-text is deduced only from context (filename, title, parent page), not from the picture itself — less accurate but still useful and cheaper.
What does Vision cost? Roughly €0.0001–0.004 per image depending on provider and model; a cheap vision model keeps it at the low end.
Can my text and image providers differ? Yes — that is the point of the dedicated image provider: e.g. DeepSeek for text, Gemini for images.
What language is the alt-text in? AISA detects the language from the image's caption/title, then the parent page, then Brand Context, then the site locale — so it matches your content's language.
Does it fix images already inside posts? Yes — it writes alt/title both on the media attachment and on the <img> tags inside the content, which is what Google actually reads.

Troubleshooting

If the image provider dropdown is empty, you have no Vision-capable provider configured — add Claude, OpenAI or Gemini. If alt-text seems generic or ignores the picture, check that Vision active is on and that the chosen provider supports it (the provider card shows a green "SEO images" badge when it does). Very large images may be rejected by vision models; AISA prefers a sized-down version, but extremely heavy originals can still fail — regenerate a standard image size. If alt-text comes out in the wrong language, set the page's language correctly, since detection follows the content.

Supported AI providers and which one to choose
Choosing the model and understanding costs
Connect an AI provider and add your API key
Bring your own key (BYOK) vs AISA Cloud credits