YouTube thumbnails: three hooks, three credits
Seedance 2 team · Updated
Three thumbnail concepts for one video made with Nano Banana 2.1 on Seedance 2, one credit each at 1K: a reaction shot with "ONE PROMPT. FULL AD.", a before/after split with "10 SECONDS", and a sceptical face with "WORTH IT?". Each took under a minute and all the words came out spelled as asked.
The result



The other two prompts
- YouTube thumbnail, 16:9: a split screen, left side a plain product photo of a sneaker on white labelled BEFORE, right side the same sneaker splashing through a neon wet street labelled AFTER, a bold yellow arrow pointing from left to right, big text 10 SECONDS at the top, high contrast, fictional sneaker, no real logos.
- YouTube thumbnail, 16:9: a close-up of a creator's face looking sceptical, next to a glowing video play button and a price tag reading 33 CREDITS?, big bold text WORTH IT? on the right, dark background with lime green accents, high contrast, no real logos.
The prompt
YouTube thumbnail, 16:9: a surprised young creator at a desk pointing at a laptop that shows a cinematic sneaker video frame, bright rim light, big bold white text with black outline reading ONE PROMPT. FULL AD. on the left side, high contrast, clean background, fictional sneaker, no real logos.
Settings: text to image, 1K, 16:9, one run per concept.
What it cost
| Model | Settings | Credits | Time to video |
|---|---|---|---|
| Nano Banana 2.1 | Hook 1, 1K, 16:9 | 1 | 54 s |
| Nano Banana 2.1 | Hook 2, 1K, 16:9 | 1 | 39 s |
| Nano Banana 2.1 | Hook 3, 1K, 16:9 | 1 | 26 s |
What came out
Every headline came out spelled as written, large and readable at thumbnail size.
The neon signs in the "after" half carry made-up lettering; check backgrounds for invented text before you publish.
The people are AI-generated, not real creators.
Make your own
- Add Seedance 2 as a custom connector in Claude or ChatGPT with the URL https://www.seedance2.net/mcp, or open the generator on https://www.seedance2.net/text-to-video.
- Paste the prompt and pick the same model and settings.
- Results vary from run to run with the same prompt; generate twice if the first take misses.
- Write three different hooks (result, transformation, question) and make one thumbnail for each.
- Keep the on-image text to three words or fewer.
- Nano Banana 2.1 costs 1 credit per image at 1K and 2 at 2K, so try several versions before you animate one.
How this was made
Made on 2026-11-20 through the Seedance 2 MCP server (https://www.seedance2.net/mcp), the same way Claude, ChatGPT or Codex would call it, on a paid Seedance 2 account, so there is no watermark. Credits and times are what each run actually charged and took.
Try the prompt
YouTube thumbnails: three hooks, three credits
Text to videoYouTube thumbnail, 16:9: a surprised young creator at a desk pointing at a laptop that shows a cinematic sneaker video frame, bright rim light, big bold white text with black outline reading ONE PROMPT. FULL AD. on the left side, high contrast, clean background, fictional sneaker, no real logos.
Questions
- Can AI make YouTube thumbnails with text?
- Yes. Nano Banana 2.1 spelled all three short headlines correctly here. For your exact brand fonts, add the text in a design tool instead.
- How big should a YouTube thumbnail be?
- YouTube uses 16:9; 1280×720 is the common size. Make it at 16:9 and resize if needed.