YouTube thumbnails that are clear at phone size

Two thumbnail variants for one video, each readable at 160 pixels wide, that match what the video really delivers.

Best for: Creators and small brands who publish video and want thumbnails that look deliberate, not improvised.

• The video title and a one-sentence description of what a viewer gets from it.
• A photo of a real person or object from the video, if you plan to use one (you need the rights to it).
• An editor that can set text and export a 1280 × 720 image, such as Canva, Figma or Photoshop.

1. 1

   Pick one idea per thumbnail

   A thumbnail has about a second to be understood. Decide the single idea: a result, a question, a contrast or a moment. If you need two ideas, you need two videos or two thumbnails to test.

   ✦Tip. Say the idea out loud in under six words. If you cannot, it is still two ideas.

2. 2

   Choose one subject

   Use one face or one object, large, with clear edges. Crowded frames turn to noise on a phone. Keep the subject away from the bottom-right corner, where the video duration label sits.

3. 3

   Generate or shoot the background separately

   Ask an image model for a background with an empty area for text and a clear place for your subject. Place your real photo on top, so the face or product is never an AI approximation of itself. Match the direction of the light.

4. 4

   Add at most three words

   Text should add something the title does not say. Use a heavy, high-contrast typeface at a size you can read when the whole image is 160 pixels wide. Keep text inside a safe margin of about 5% on all sides.

5. 5

   Test at real size

   Export at 1280 × 720 pixels, then shrink the preview to roughly 160 pixels wide and look at it next to other videos. If the idea, the subject or the words are not clear in one glance, simplify.

6. 6

   Compare two versions honestly

   If the platform offers thumbnail testing, run it. Otherwise change one thing at a time and keep notes on which version you used and for how long. Never judge a result from a tiny number of views, and never promise viewers something the video does not show.

Replace anything in [square brackets]. Prompts are starting points: results vary between tools and versions, so review every output.

Prompts are written in English because image, video and most text models follow English instructions best. To get answers in your language, add “Reply in English.” to a prompt for a text assistant.

I make videos about [topic] for [audience]. Here is my video title: "[title]". Give me 5 thumbnail concepts. For each: the single idea in one sentence, what is in the frame (one face OR one object), at most 3 words of on-image text, and the emotion it should trigger. Do not promise anything the video does not deliver.

Cinematic close-up of [one subject], strong rim light in [colour] against a dark [colour] background, shallow depth of field, large empty area on the right third for text, high contrast, vivid saturated colour, 16:9, no text, no logos, no watermark.

Here are two thumbnail variants for the same video: A) [describe] B) [describe]. Act as a critical viewer scrolling a phone feed. Which is clearer at 160 pixels wide, and why? Point out anything that would be unreadable or misleading. Do not flatter either option.

• ✕Packing several ideas, faces and text blocks into one frame.
• ✕Using thin or low-contrast type that disappears on a phone.
• ✕Letting an image model redraw a real person’s face or your product.
• ✕Promising in the thumbnail something the video never shows.