Featured images make your help centre easier to scan: they show on category pages and article lists, and when someone shares an article on social media or in chat. Making one for every doc by hand takes a long time. PinkDocs → Tools → Featured images creates them for you with AI, in one consistent style, for as many docs as you choose.
Before you start
- An OpenAI API key. Featured images use OpenAI's image models. Add your key in PinkDocs → Settings → General → AI for the knowledge base. (If you use OpenRouter for the AI assistant, you still need an OpenAI key for images.)
- A verified OpenAI organisation. OpenAI asks you to verify your organisation before you can use its image models. Do this once under Settings → Organisation → General in your OpenAI account.
- Permission: you need to be an administrator.
How the screen is laid out
- Image style (top): the prompt, shape, quality, model and reference images. These apply to every image in the run, and PinkDocs remembers them for next time.
- Documents (below): pick which docs get an image, then Preview one or Generate with the buttons underneath the list.
Generate featured images
Open the tool
Go to PinkDocs → Tools → Featured images.
Write your prompt
Describe the image you want in the prompt box. To add something that changes for each doc, put the cursor where you want it and choose from the Insert fill-in menu:
| Fill-in | Becomes |
|---|---|
{title} | The doc's title |
{excerpt} | The doc's excerpt, or the start of its text |
{category} | The doc's first category |
{kb} | The knowledge base's name |
{site} | Your site name |
{brand_color} | Your brand colour, e.g. #ff26ba |
Choose shape, quality and model
Pick them from the three menus under the prompt. See Choosing the settings.
Add reference images (optional)
Click + Add next to Reference images and choose up to four images from your Media Library, or upload new ones. Good choices are your logo, an illustration in your brand style, or a featured image you already like. They're sent with every image, so the whole set looks like it belongs together. Hover over an image and click × to remove it.
Choose the documents
Pick a knowledge base, then narrow the list by category or by searching titles. Only without an image is on at first, so you don't replace images by mistake. Tick the docs you want, or tick Select all at the top of the list. Under the list you see how many images that is and roughly what it costs.
Preview one
Click Preview one. After 10–40 seconds you see the image for the first selected doc and the exact prompt that was sent. Click Use as featured image to keep it, or Discard to delete it and adjust your prompt. Repeat until you like the result.
Generate the rest
Click Generate, check the estimate and confirm. Docs are done one at a time; each row shows Generating…, Created, Skipped or Failed. Keep the page open until it finishes. You can Stop at any time, and Retry failed runs only the ones that didn't work.
Each image is saved to your Media Library (attached to its doc, with the doc's title as alt text) and set as the doc's featured image. Docs that already have one are skipped unless you tick Replace existing images.
Write a good prompt
Start from the default prompt and adjust it until the preview looks right. A good featured image prompt says:
- The style: flat illustration, 3D render, isometric, line art, photo.
- The colours: use
{brand_color}, or describe your palette. - What to show: usually an idea related to
{title}, not a screenshot. - What to avoid: always add "No text, letters or logos in the image." Image models often misspell words.
Choosing the settings
| Setting | Options | When to use |
|---|---|---|
| Shape | Landscape (1536 × 1024) / Square (1024 × 1024) / Portrait (1024 × 1536) | Landscape fits article headers and social previews best. |
| Quality | Low / Medium / High | Medium suits most help centres. Low is fast and cheap for testing; high for detailed images. |
| Model | GPT Image 1 / GPT Image 1 mini (cheaper) | GPT Image 1 gives the best detail. Mini is several times cheaper and fine for simple illustrations. |
What it costs
Approximate cost per image, charged by OpenAI to your account:
| Model | Low | Medium | High |
|---|---|---|---|
| GPT Image 1 (landscape) | ~$0.02 | ~$0.06 | ~$0.25 |
| GPT Image 1 (square) | ~$0.01 | ~$0.04 | ~$0.17 |
| GPT Image 1 mini (landscape) | ~$0.01 | ~$0.02 | ~$0.05 |
Reference images add a little to each image. OpenAI's prices can change: check their pricing page for current rates. The cost of every image you create is recorded in PinkDocs → Dashboard → Costs, under Featured images.
Troubleshooting
| Message | What to do |
|---|---|
| Add an OpenAI API key… | Add an OpenAI key in PinkDocs → Settings → General → AI for the knowledge base you chose. |
| OpenAI rejected the API key | The key is wrong, revoked or out of credit. Create a new key in your OpenAI account. |
| …verified organisation… | Verify your organisation in your OpenAI account settings, then try again (it can take a few minutes to apply). |
| Rate limit or a timeout | Wait a minute and click Retry failed. Large batches can hit OpenAI's limits. |
| The image has text in it | Add "No text, letters or logos" to the prompt, and avoid putting words in quotes other than {title}. |
Common questions
Yes. Select only that doc and click Preview one, then Use as featured image.
Up to 100 per run. Run it again for more.
Only the prompt (with the fill-ins replaced, so including the doc's title and, if you use {excerpt}, its excerpt) and your reference images. See AI privacy and data handling.
Featured images are normal Media Library images. To change one, open the doc and replace or remove its featured image. Discarded previews are deleted automatically.
They show on category pages, in article lists and as the image when an article is shared. Whether they show at the top of the article itself depends on your template.