Pricing
from $150.00 / 1,000 thumbnail generateds
AI Thumbnail Generator
Under maintenanceTurn any title or story into a scroll-stopping thumbnail: an AI emotional close-up face with a bold burned-in hook headline. Auto-detects emotion and writes the hook, or set your own. 9:16/16:9/1:1. For YouTube, Shorts, TikTok/Reels covers, ads, course covers, blog headers โ any content.
Pricing
from $150.00 / 1,000 thumbnail generateds
Rating
0.0
(0)
Developer
Actor stats
0
Bookmarked
2
Total users
1
Monthly active users
a day ago
Last modified
Categories
Share
Turn a title or a story into a thumbnail people actually click: an AI-generated emotional close-up face with a bold hook headline burned across the lower third. Built for YouTube, Shorts and Reels covers, TikTok, and ad creative, where the thumbnail does most of the work and you need a new one for every video.
How it works
It scans your title and story for emotional cues (betrayal, rage, shock, heartbreak, and a few others), picks the strongest match, and writes a short all-caps hook from the same text. That feeds a prompt to gpt-image-1, which renders a hyper-realistic face with deliberately empty space at the bottom. The image comes back text-free on purpose, then the hook is overlaid afterward with a real font so the type stays crisp instead of being mangled by the image model.
Input
You can run it with nothing but a title. Everything else is an override or a knob.
| Field | Required | Notes |
|---|---|---|
title | No* | The video title or story headline. Drives both the detected emotion and the hook. AITA/TIFU shorthand gets expanded automatically. |
story | No* | Full script or story text. Gives the emotion detection and hook extraction more to work with. |
aspectRatio | No | 9:16 (Shorts/Reels/TikTok, default), 16:9 (YouTube long-form), or 1:1 (square). |
emotion | No | Force the facial emotion instead of auto-detecting. One of betrayal, rage, shock, disgust, heartbreak, fear, guilt, defiance. |
hookText | No | Force the burned-in headline. Leave empty to derive it from the title or story. |
gender | No | Bias the face, e.g. feminine or masculine. Empty uses the model default. |
burnText | No | Overlay the hook on the image. Default true. Turn off to get the raw face only. |
openaiApiKey | No | Your OpenAI key for gpt-image-1. Bring your own. Marked secret. |
imageQuality | No | gpt-image-1 quality: low, medium, or high. Default high. This is what moves your OpenAI bill. |
*At least one of title or story is needed, otherwise there's nothing to work from.
Output
One PNG per run, saved to the key-value store, with a record pushed to the dataset. The record includes output.imageUrl and output.imageKey for the file, plus emotion, hookText, aspectRatio, and size so you can see what it decided. If the text overlay fails for any reason, you still get the face image back without the headline rather than an error.
Example
{"title":"AITA for refusing to pay for my sister's wedding?","aspectRatio":"9:16","imageQuality":"high","openaiApiKey":"sk-..."}
Pricing
$0.15 per thumbnail, pay per result, no subscription. The image generation runs on your own OpenAI key, so that usage is billed to you separately by OpenAI.
Notes
It needs an OpenAI key with gpt-image-1 access. Emotion detection is keyword-based, so it lands well on emotionally loaded story titles (relationship drama, AITA-style posts) and is less sure on dry or abstract ones. When auto-detection guesses wrong, set emotion and hookText directly.
