Eight tools, one bench, one credit balance. One credit is one image, and everything else on this page is free to use.
Type a prompt and the model builds the whole scene. Upload an image and the bench remakes it. Paste a YouTube link and it grabs that video's thumbnail so you can remake that.
Add a headline, the words that go on the thumbnail, and the AI renders them into the image.
Big Face, Split Screen, Number Pop, Before / After, Neon Glitch, Clean Edu, Money Stack, Reaction. Each style is a layout the feed already understands.
Your headline is a field, not an afterthought. Three words, all caps, one accent, painted into the render.
A library of reference cards you can add to any render. Subject looks: 3D Pop-Out, Speed Blur, Neon Outline, Extreme Close-Up, Studio Portrait, Sticker Cut-Out.
Text effects: Fat 3D, Fire, Gold, Gradient Bevel, Ice, Neon Glitch. Rough preview on the right; the real thing is painted by the model.
Upload a selfie and the bench cuts it out automatically. Then generate expression versions of that face: Shocked, Surprised, Happy, Excited, Angry, Curious, Scream, Sad. One credit each.
Already have a result you like but the face is flat? Change the expression on the finished image: two options, two credits.
Any face saved in Face Lab can be put into a render. The treasure chest, the mansion, the trading floor: describe the scene, pick the face, pick the expression.
The model paints the face into the scene with its lighting and angle, so a whole series can carry the same face.
Open any result in the editor and add text layers with real fonts on top of the render. Move them, resize them, recolour them. No credits spent.
Region edit changes only a box you draw. Fix a hand, swap a prop, clean up a corner, and leave the rest untouched. Two options, two credits.
Tweak takes one of your results and uses it as the source for the next render. Remix takes any past render and runs it again with a new headline, style or face.
Every image keeps a version history, so a bad edit is one click from undone.
Every render is 2K: 2752 x 1536 for 16:9. Export HD gives you a 1280 x 720 file under 2 MB, ready for YouTube's upload limit. Export 2K gives you the full render.
Free accounts get a small watermark on exports. Every paid plan removes it and includes commercial use.
Photoshop, Canva and a freelance designer all make thumbnails. None of them puts your face in a new scene in about a minute.
| Thumbnail Bench | Photoshop | Canva | Freelance designer | |
|---|---|---|---|---|
| Time per thumbnail | About a minute | 1 to 2 h | 20 to 40 min | 1 to 3 days |
| Your face in a new scene | Send photos, wait | |||
| Headline painted into the image | Manual | Manual | Brief them | |
| Cost per 100 thumbnails | $29 (Studio, 140 images a month) | $23 + your time | $13 + your time | $1,500 to $5,000 |
| Options per render | 1 to 4, you choose | One at a time | One at a time | Extra fee |
| Learning curve | None | Steep | Moderate | None, but slow |
Not on the bench yet, and not for sale: a CTR rater that scores a cover before you upload, batch generation for split tests, a template library, and a channel style study that reads your past uploads. When one of them ships it shows up in the changelog first.
Guides on why viewers click, how titles and thumbnails work together, and how to test what you make. Free, no account.
What a cover has to do in the feed, the five parts, the eye path and the pre-upload checks.
Click psychologyCuriosity, emotion, stakes, specificity and novelty, and how each one shows up on a cover.
TestingYouTube's Test & Compare, the manual swap method, and how to read the result.
CTRWhat impressions click-through rate measures and what a good number looks like in context.
4 free credits, 4 more after your first render. No card. Your first options in about a minute.