Cover Image Generator

Renders the branded images for a conversation: the video thumbnail, the LinkedIn header, and quote cards for social sharing.

At a glance

Needs	The generated title (and, for quote cards, the quotes)
Produces	Thumbnail, LinkedIn header, quote card images
Waits when	The draft hasn't been generated yet
Re-running	Safe and cheap — re-render freely after the content changes

What it does

Renders images from HTML templates using the conversation's title, quotes and participant photos — not from an image-generating AI. Templates are why every episode looks like it belongs to the same show: same layout, same type, same treatment, only the content differs.

Each journey picks a template pack, and the pack supplies all three templates. Changing the pack on a journey step changes the design of everything that journey renders from then on.

Some packs ship their own image library and pull artwork from it at render time, so their cards differ between runs and between sibling cards in the same run. Draws come from a shuffled deck, so a batch exhausts the library before repeating an image. Because the choice happens during rendering, it isn't recorded with the result — if you want a specific image, re-render until you get one you like, or use a pack without a library.

Packs also differ in whether they show participants' faces. The step always looks up each participant's photo and hands it to the pack; whether the card draws it is the pack's choice. The default pack sets the names as a list instead, so a conversation with no portraits on file looks exactly like one where everybody has a good headshot.

How it behaves

• It waits for the title. No title means no card to render, so the step waits rather than failing.
• Partial success counts as success. If one card fails to render, it's skipped and logged; the step only fails outright if nothing rendered.
• Re-rendering is expected. Changed the title or the quotes? Re-render. The images are cheap to make.
• Pushing a regenerated thumbnail to an already-uploaded video is a separate step — see YouTube publishing.

Settings

Set on the step: the template pack, whether to generate each of the thumbnail / LinkedIn header / quote cards, and the maximum number of quote cards.

To judge whether a card actually looks right, use the card preview route in a browser (?pack=<name>&conversation=<id>) — automated checks can only confirm that a card rendered, not that it's good.

Troubleshooting

Symptom	Likely cause	What to do
Waiting with no error	The draft hasn't been generated yet	Normal. Check the Content Generator step.
No cards at all, step failed	The rendering browser isn't installed or can't start on the server	An administrator installs it; then re-run
Some quote cards missing	Those individual renders failed, or there were fewer quotes than the card limit	Check the quotes in the Publishing panel; re-render
Cards show the old title	They were rendered before the content changed	Re-render this step
Video still shows the old thumbnail	The thumbnail is pushed to YouTube by a separate step	Re-run the thumbnail sync step
Participant photos missing from cards	Those People have no photo on record	Add photos to the Person records, then re-render
Wrong branding	The journey step is using a different template pack	Change the pack on the step, then re-render
Old image files piling up	Each re-render writes new files	Harmless; an administrator can clean up orphaned image files

Technical reference

Step type	cover_image_generator
Runs after	content_generator
Feeds	youtube_thumbnail_sync (and any other consumer of the generated images)
Reads	fields.title, fields.subtitle, fields.instagram_quotes, participant Person.photo; step settings template_pack, generate_thumbnail, generate_linkedin_header, generate_quote_cards, quote_card_limit
Writes	artifacts.thumbnail, artifacts.linkedin_header, artifacts.instagram_quote_cards, records.cover_images
Needs on the server	A headless Chromium for rendering