All posts

Thumbnails9 min readUpdated

How to make a YouTube thumbnail from a video link with AI

Paste a YouTube video link into WThumb, bring in its public thumbnail as your reference, then rebuild the idea with your own words, photo, format, and direction. This complete workflow shows what the link imports, what it does not, and what to check.

Dark cover with white PASTE THE and red LINK text, a white chain-link icon, an arrow, and an empty glowing red thumbnail frame.

Making a YouTube thumbnail from a video link with AI removes one dull step: saving the source image before you can explain what you like about it. In WThumb, paste a supported YouTube URL and press Get thumbnail. The public thumbnail from that video becomes the reference for a new design. The link imports no video, title, description, script, comments, or audience data. You decide what the new image should promise.

A reference is a visual brief, not a locked template or permission to copy another creator's identity. Bring your own words, subject, colors, and optional photo. WThumb uses them to make a new 16:9 YouTube thumbnail, 9:16 Short, or both. Start in the AI YouTube thumbnail generator, or use the full AI thumbnail guide to understand every Studio choice.

What a YouTube link gives the AI

A YouTube link gives WThumb one thing: the public thumbnail attached to that video. It is a quicker route to a reference image. You do not have to save the thumbnail, find the file in Downloads, and upload it again. Paste the URL, press Get thumbnail, and check that the expected image appears before you continue.

The rest of the video does not come with it. WThumb does not listen to the spoken content, choose a promise from the title, or decide why someone should watch. If the reference says I QUIT but your video is a laptop repair, those words do not become part of your brief. You must supply the exact new words and explain what the new thumbnail is about.

Treat the link as a visual starting point. It can show that the source uses one large subject, a close reaction, a dark background, or a strong object on the opposite side. Those roles can guide a new design, but the result is generated rather than edited pixel by pixel. Do not expect every edge, pose, letter, or prop to land in the same place.

Reference imageReference thumbnail with a man holding colored banknotes and four percentages above them
WThumb outputNew thumbnail with a different person holding euro notes against a teal wall and a new headline
A real example transformation made with WThumb. The reference supplies the broad visual job; the output changes the person, words, currency, background, and design.

Small logos, prices, charts, and product labels need extra care. A generated result can redraw them. If one matters, attach a clean supporting image and name its role in the instructions, then inspect the result. If an asset must remain pixel exact, finish that placement in a layer-based image tool after generation.

Choose the visual job, not the creator

The safest place to begin is your own channel. An older thumbnail may already have the right amount of text, subject size, and energy while its topic or branding needs a refresh. Paste that video's link and use the earlier image as direction for the new one. You own the source, understand the original promise, and can judge what should remain familiar.

When the reference comes from another channel, ask what job the image performs. Do not choose it only because the video has many views or the creator is recognizable. A thumbnail built around a famous face may have little useful structure once that identity is removed. The reference selection guide shows how to look for useful roles instead of surface decoration.

  • First look. What catches the eye first: a face, product, result, number, or place?
  • Promise. What carries the idea: the expression, contrast, object, or relationship between two elements?
  • Identity. Which parts belong only to that creator: their face, words, logo, palette, set, or recurring character?
  • Fit. Will the visual job still make sense when your real topic replaces theirs?

A useful source might show a person on one side and a failed product on the other. That job can fit a repair, review, warning, or comparison. A weak source depends on a private joke, a recognizable set, a celebrity, or a branded character your video does not contain. Once those parts leave, there is no useful idea left.

As of September 2026, YouTube's Thumbnails policy says thumbnails must not mislead viewers about what is in the video. It also says unauthorized impersonation can include using AI to copy someone's likeness so a channel appears owned or authorized by that person. Use a source for its general visual job, then replace its identity with your own subject, words, and look.

Decide what changes before opening Studio

Write one sentence about the new video before touching the controls. For example: I repaired a slow laptop in ten minutes, and the thumbnail should show the broken state and my relief. That sentence becomes the filter for every choice that follows. If an object, word, or effect does not help it, remove it from the brief.

  • Use as direction. The broad visual job, energy, contrast, or relationship between the main elements.
  • Replace. The old headline, topic, person, brand colors, and channel identity.
  • Add. Your own photo, the real product, a screenshot, a logo you may use, or another detail that belongs to the new video.
  • Remove. Decorative arrows, badges, extra words, or background objects that do not help the new promise.

Choose the output frame at the same time. Pick YouTube for a 16:9 video, Shorts for a 9:16 image, or both. WThumb makes each selected format for its own frame, so the tall result is rebuilt rather than delivered as a center crop of the wide one. The subject and words stay useful in both, but their arrangement can change.

Then decide whether you belong in the thumbnail. Add your own photo if your identity or reaction helps explain the video. You can keep the expression already visible in your photo or take direction from the expression in the reference. If an object, screen, place, or result carries the story, continue without a photo and make a faceless design. The own-photo guide explains which source shots give a generated face the best chance to remain recognizable.

This planning step prevents the common failure: requesting another creator's image with only one noun changed. A new topic needs a new promise. The source helps express it, but should not decide it for you. If your one-sentence plan and the reference disagree, keep the plan and choose a better reference.

Write a brief for the new video

The Studio separates literal words from visual instructions. Keep them separate in your head too. The words field is only for text the viewer should read, with up to three lines. Do not put move this to the left in that box. If the design should have no overlaid headline, choose the no-headline option instead of leaving the decision unclear.

Additional instructions describe the image: main subject, background, position, scale, expression, colors, and supporting objects. One clear hierarchy is more useful than a long list of adjectives. Name the hero first, the supporting element second, and the background last. Viral, clickable, professional, and better do not describe anything the model can place in a frame.

If a real product, screen, document, or logo matters, attach it under Extra images. WThumb accepts up to four and labels them @1 through @4. Refer to each attachment directly, such as Put @1, the red camera, large on the right. Do not attach four images without giving each one a different role. An unused source is another instruction competing for attention.

A brief to paste

Words. FIXED IN / 10 MINUTES

Instructions. Make the damaged laptop the main subject on the right. Put me on the left looking relieved, with my face clearly separated from the background. Use a dark workshop with one warm light behind the laptop. Keep the design simple. Do not add other devices, logos, arrows, or words.

This replaces the source headline, names the new subject, assigns each side a job, and removes details the model might otherwise invent. It does not ask for an exact copy. For a faceless result, skip the photo and replace the person instruction with: Make the damaged laptop the main subject. Put a large warning symbol on the left where the person would otherwise be. Keep the rest of the frame quiet.

Check the result as a thumbnail

A generated image can look polished at full size and still fail as a thumbnail. Shrink it until it is close to the size of a suggested-video card, then check it in a fixed order. Start with the promise. Does the image represent something the video actually contains? Next, read every word, number, and punctuation mark. A strong picture cannot rescue the wrong claim.

  1. Main subject. Can you identify it immediately when the image is small?
  2. Identity. If you added your photo, do you still look recognizable and does the expression fit the video?
  3. Details. Are products, screens, hands, logos, and supporting objects accurate enough for the claim?
  4. Frame. Does the composition work in the selected 16:9 or 9:16 shape without important content at the edge?

As of September 2026, YouTube's Add custom thumbnails page recommends a 16:9 image for videos and a 9:16 image for Shorts. The same page says the account must be verified to upload a custom thumbnail, and that Shorts thumbnail uploads are currently available in YouTube Studio on a computer.

If one isolated part is wrong, revise it: press the pencil on the thumbnail and ask for one observable correction, such as make the laptop larger and leave everything else unchanged. The revision produces one new version and uses one credit (1 credit = 1 thumbnail). If the promise, subject, and direction are all wrong, return to the brief instead. The AI thumbnail error checklist covers the final accuracy pass in more detail.

How to do this in WThumb

  1. Choose the formats. Select YouTube 16:9, Shorts 9:16, or both. Each selected format becomes its own finished image.
  2. Paste the link. Choose YouTube link, paste the URL, and press Get thumbnail. Confirm that the preview belongs to the intended video.
  3. Make the photo decision. Add your own photo and choose its expression direction, or continue without one for a faceless result.
  4. Enter the exact words. Use up to three lines, or explicitly choose no headline. Put placement and visual requests in Additional instructions.
  5. Attach only useful sources. Add up to four supporting images and refer to them as @1, @2, @3, or @4 in the instructions.
  6. Confirm the request. Choose one or two designs per format. The Confirm view shows how many credits the complete request uses.
  7. Generate and inspect. Sign in when asked, start the request, and wait about thirty seconds. Download the Full HD+ JPEG after checking its words and details.

If the link does not load

Check that you pasted a complete YouTube video URL rather than a channel, playlist, search result, or YouTube Studio administration link. Standard watch pages, youtu.be links, Shorts links, and Live video paths work when they contain a valid video ID and a public thumbnail can be prepared. If the image is yours and the URL route still cannot prepare it, choose Upload image and add the reference file directly.

Start with one clear example

For the first run, choose a reference with one obvious subject and a short headline area. Replace its words and topic, then add your photo only if the new video needs you in the frame. A simple source makes it easier to tell whether the brief worked. For WThumb, new accounts get two free credits; no card is needed, and paid plans are available when you need more. Paste one link whose visual job fits your video, then make the result belong to your channel.

Common questions

Can AI make a new thumbnail from a YouTube link?
Yes. In WThumb, paste a supported YouTube video URL and press Get thumbnail. The public thumbnail from that video becomes the visual reference for a newly generated design. You still provide the new headline, instructions, output format, and optional personal photo. The link is a quicker way to supply a reference. It does not ask the AI to understand the video or automatically choose its marketing promise.
What part of the YouTube video does WThumb use?
Only the video's public thumbnail is fetched. WThumb does not import the footage, transcript, title, description, comments, or channel analytics from the link. The source image can guide the broad visual idea, but it cannot tell WThumb what your new video contains. State the new promise in your own words and instructions, and use only objects and claims that honestly represent the new video.
Can I use a YouTube Shorts or Live video link?
Yes, when the URL contains a valid public video ID. Supported forms include standard watch pages, youtu.be links, Shorts links, and Live video paths. A channel, playlist, search result, or YouTube Studio administration URL is not a single video reference. If a supported link still cannot prepare an image you own, download that thumbnail and use the reference image uploader instead.
Can I replace the person in the reference with my own photo?
Yes. Turn on the photo option and add your own image. You can keep the expression visible in your photo or take expression direction from the reference. Choose a clear source with the face visible, room around the head, and lighting that does not hide important features. If you skip the photo step, WThumb removes the person from the reference direction and rebuilds the space around the remaining subject.
Does WThumb copy the original thumbnail exactly?
No. The thumbnail from the link is a visual brief for a newly generated design, not a file WThumb edits pixel by pixel. The broad idea, energy, and roles of the elements can guide the result, but exact placement and detail are not guaranteed. Replace another creator's words, face, branding, and identity. Check generated text, numbers, products, and logos before uploading the finished image.
ShareX