Four versions of the click. Test the winner.
AI Thumbnail Generator
Describe the video and the words to show, pick a layout, and get 4 thumbnail variations with a bold headline you can read on a phone — 16:9 for the feed, 9:16 for Shorts covers.
8 credits for 4 variations, shown before you generate. Plans renew monthly until cancelled and can be cancelled online at any time — see pricing details.
See the full prompt
Video thumbnail for a budget meal-prep video — text: “5 MEALS £20”. One large hero object on the right, huge bold outlined headline text on the left, saturated colours, high contrast. Readable at small size, the headline spelled exactly as given, no real people, no logos, no watermark.
Opens the image studio on Qwen Image 3.0 with all of this set — press Generate there. 2 credits an image; the total is confirmed before anything is charged.
Start from a video like yours
Each card opens the studio in the Bold text + object layout with the headline already in the prompt. Swap the topic and the words.
a budget meal-prep video — text: “5 MEALS £20”
Open with this ideaDesk setupa minimalist desk setup tour — text: “MY $500 SETUP”
Open with this ideaTravel vloga week in the Scottish Highlands — text: “7 DAYS NO CAR”
Open with this ideaCoding tutoriala beginner Python tutorial — text: “PYTHON IN 1 HOUR”
Open with this ideaGardena balcony vegetable garden tour — text: “GROW 20 KG”
Open with this ideaPodcasta podcast episode about sleep science — text: “SLEEP BETTER TONIGHT”
Open with this ideaWhat makes a thumbnail get the click
A thumbnail is seen at the size of a fingernail, for less than a second, next to a dozen others. What wins is contrast and one idea: a single object big enough to recognise, three or four words that finish the title’s thought, and a colour that is not the colour of everything around it. The layouts in the starter are those rules written down — the fixed tail adds the rest.
a budget meal-prep video — text: “5 MEALS £20”
one large hero object on the right, huge bold outlined headline text on the left, saturated colours, high contrast
Readable at small size, the headline spelled exactly as given, no real people, no logos, no watermark.
From topic to test set in three steps
Topic and headline
What the video is about, then the words in quotation marks — “GROW 20 KG”. Two to four words read best.
Layout and shape
Bold text + object, before/after, clean tutorial, dramatic or minimal. 16:9 for the feed, 9:16 for vertical covers.
Four to compare
Pick two to test against each other and keep the others as fallbacks. The images land in My Creations.
Tips from the thumbnail playbooks
Don’t repeat the title. The headline on the image should add to the video title, not copy it — the title says what, the thumbnail says why it matters.
Check it at phone size. Zoom out until the image is two centimetres wide. If the headline or the object is gone, try the Minimal layout or fewer words.
Keep a series recognisable. Reuse the same layout and colour for every episode of a series; only the object and the words change. More on batch workflows in the thumbnail workflows guide and the variations playbook.
Settings the starter uses
| Setting | Detail | |
|---|---|---|
| Model | Qwen Image 3.0 | Text-rich, information-dense layouts |
| Shapes | 16:9 · 9:16 · 1:1 | Widescreen, Phone, Square |
| Quality | Fixed | One output size per shape |
| Price | 2 credits an image | 8 for the default 4 |
Real people, rights and labelling
The presets and prompts never show real people or brand logos, and prompts naming public figures are blocked — a thumbnail that puts a real person into a scene they were never in misleads the viewer. To use your own face, start from a photo you took in the image studio’s edit mode. Every image is AI-generated and carries a signed C2PA provenance mark saying so, in line with Article 50 of the EU AI Act; anyone can read it on the verification page, and the Content & Safety Policy sets out what may not be created. Plans renew monthly until cancelled and can be cancelled online at any time; the full terms are on pricing details.
Frequently asked questions
How do I make a thumbnail with AI?
Describe the video and the exact words you want on it — for example, a budget meal-prep video with the text “5 MEALS £20” — tap a layout and press Create. The image studio opens on Qwen Image 3.0 in 16:9 and renders 4 variations, so you can compare them before you upload.
Can the thumbnail include readable text?
Yes — that is why the starter uses Qwen Image 3.0, the studio's text-layout model. Keep the headline to two to four words, in capitals, inside quotation marks in your idea; short text is what reads on a phone and what the model renders most reliably. Always check the spelling before you publish.
What size is the thumbnail?
16:9 by default, the shape video platforms show in search and on the home feed, with 9:16 for vertical Shorts and Reels covers and 1:1 for square posts. Each runs at the model's output size for that shape; the image upscaler can take it larger if a platform asks for more pixels.
Can I put a real person or a famous face on the thumbnail?
No. The presets and the prompt tail rule out real people and logos, and prompts naming public figures are blocked. A thumbnail showing a real person in a scene they were never in misleads viewers — for your own face, use a photo you took and add the text layout around it in the image studio's edit mode.
Why generate four variations?
Because thumbnails are tested, not guessed. Four versions of the same idea — different colours, a different object, a different crop — give you a real choice and a second option to swap in if the first one under-performs.
How much does it cost?
Qwen Image 3.0 is 2 credits an image, so four variations are 8 credits. The total is on the Generate button before anything is charged, and failed images are refunded.
Everything around the video
One credit balance across image, video and music.
AI Image Studio
Edit a thumbnail or start from your own photo.
OpenAI Video Generator
The clip the thumbnail sells.
OpenAI Prompt Generator
Turn a rough topic into a sharp prompt.
OpenImage Upscaler
More pixels when a platform asks for them.
OpenAI Music Generator
An intro sting for the channel.
OpenAI Poster Generator
Event posters with the same bold type.
OpenVerify AI content
Read the C2PA provenance mark on any file.
Open