higgsfield-youtube-thumbnail

Provide the video topic, an optional face photo, and a ratio, and the skill builds a truthful information-gap concept, then renders it with Nano Banana Pro at 4K through the higgsfield CLI. You can request a single concept or a variant set of about four.

Request a headline overlay and the skill applies one of five presets (Beast, Fire, Neon Lime, Clean Glass, Marker) using an HTML/CSS preview and a canvas bake at native resolution. Text is kept off the face, limited to 2-4 words, and baked only when explicitly requested.

Yes, attach up to three face photos as --image references and the prompt enforces an identity lock preserving bone structure, skin tone, and hairline. The skill never beautifies or restyles referenced faces and verifies the match after rendering.

It supports 16:9 for standard YouTube thumbnails, 9:16 for Shorts covers, and 4:5 for Instagram video covers. Note that Seedream edits lack a 4:5 mode, so edits on 4:5 renders require approving a change to 3:4.

CLI versions through 1.1.20 can mislabel a nano_banana_pro job reference as nano_banana_pro_job, causing the edit rejection. Download the picked result_url to a local file and retry the same edit with the local image path instead of the job ID.

Do not use it to generate the video itself, product catalog photos, or marketplace listing cards, as dedicated skills cover those workflows. It also cannot run on unlimited allowances through the CLI, which operates on credits.