All posts

Thumbnails10 min read

Can Google Flow make YouTube thumbnails? What it does and what it leaves to you

Google Flow can make a still image with Google's Nano Banana models, from a prompt, your own photo or a frame saved from a video clip. Here is what Google's pages say it does, what a YouTube thumbnail still needs, and where WThumb fits.

Dark cover with white GOOGLE FLOW and blue THUMBNAILS? text. On the right, three small wide white outline frames in a gently curving row like a strip of film, the middle one lifted out of the row, larger and solid bright blue, under one soft blue light.

Yes, Google Flow can make YouTube thumbnails in the sense that it has an image generator: its prompt box has an Image mode that runs Google's Nano Banana models, and you can bring in your own photo or a frame saved from a video clip. What Google's Flow pages do not spell out is the thumbnail half of the job: the 16:9 shape YouTube recommends for videos, words that read at feed size, and a face that still looks like you at that size.

This post reads Google's Flow help pages and YouTube's thumbnail rules, both as of October 2026, and shows where our own thumbnail maker, WThumb, fits. New WThumb accounts get two free credits (1 credit = 1 thumbnail), no card needed.

How to use Google Flow's image generator

As of October 2026, Google's Create & edit images in Google Flow page gives these steps on a computer, and says an image can stand alone or become an ingredient for a video.

  1. Open a project. In Google Flow, click an existing project.
  2. Choose Image. Click the model name in the prompt box, then Image.
  3. Set it up. Pick the image aspect ratio, the number of outputs and the model.
  4. Generate. Describe the image in detail, add any ingredients and click Generate.

Four inputs matter for a thumbnail. Your own photo goes in as an ingredient: on a computer, drag it into the prompt box with the Agent switched on; in the Android app, tap Add, then Upload. Typing @me adds your avatar, and @ with a name adds a saved character. And Google's Edit videos & build scenes in Google Flow page, read the same day, says Save frame on a paused clip keeps that still as an image for later generations.

Editing is a prompt too. The images page says you drag an image into the prompt box with the Agent on and describe the change; earlier versions and their prompts stay in the History panel, and there is a Crop tool. The page's own Agent example asks for 20 variations of an image with different lighting.

So yes, Flow can make the picture. Whether that picture is a finished thumbnail is the rest of this post, the same question Nano Banana YouTube thumbnails asks about Gemini and can ChatGPT make YouTube thumbnails asks about ChatGPT.

What Google Flow is, and who can use it

As of October 2026, Google's Get started with Google Flow page calls Flow its AI filmmaking tool, used on a computer or in the Google Flow app. A Google blog post of May 19, 2026 says that since Flow's launch, Google has expanded it into an AI creative studio, with new capabilities in video and image generation and editing.

Google's models and supported features page, read the same day, lists three image models: Nano Banana Pro, the default for Google AI Ultra subscribers; Nano Banana 2 Lite, the default model available at no charge; and Nano Banana 2.1, the standard model. The images page names Nano Banana Pro as the prompt box default; check the model name before you generate.

Access has three conditions on the Get started page: your age verified as 18 or older, a supported region, and a Google AI Plus, Pro or Ultra subscription, with qualifying Google Workspace plans getting Flow at no extra charge and 50 daily credits. The Manage your Google Flow credits page reads more generously: with or without a subscription you receive 50 Google Flow credits a day, and without one you can still generate images. The two pages differ, so check what your own account shows.

The same day, the Where you can use Google Flow page listed more than 150 countries and territories, and it says product features may vary by country.

Every generated image carries a mark. The Get started page says all outputs made with Veo, Omni or Nano Banana include invisible SynthID watermarks. A visible one can be switched on under your profile picture, and it is applied automatically if you live in India, South Korea or Vietnam.

What a YouTube thumbnail still needs from you

Three jobs sit outside what Flow's pages describe. All three come from where a thumbnail ends up: a fixed shape, shown at fingernail size beside a title.

The right file

As of October 2026, YouTube's Add custom thumbnails page recommends 16:9 at 3840 x 2160 pixels for videos and 9:16 at 2160 x 3840 for Shorts, a minimum width of 640 pixels for videos, formats such as JPG or PNG, and file limits of 2 MB from a phone and 50 MB from a computer.

Flow's images page says you choose an image aspect ratio but lists neither the options nor an output size, and its models page says to check the active resolution in the prompt box settings. So check whether 16:9 and 9:16 are in the menu and what size the download is; the post on YouTube thumbnail dimensions has every number.

Words that read small

Google DeepMind's model card for Nano Banana 2.1, one of Flow's three image models, was published October 6, 2026 and lists Google Flow among the places the model is offered. It says the model is well suited to clear text for posters, and names poor text rendering in small text and long paragraphs among its known limitations.

YouTube's Thumbnail & title tips page, read the same month, says if you add text, make sure to use a font that's easy to read. So ask for two or three large words and read every letter before you upload; AI thumbnail text errors covers what to look for.

A face that is yours

Flow has three routes to your face. As of October 2026, Google's Create & use your avatar in Google Flow page describes an avatar with your own highly accurate visual and audio likeness, not available in the EEA, the UK or Switzerland. You can also add a photo as an ingredient, or build a character from one or two uploaded images, which Google's Flow projects, assets & collections page says keeps its face strictly consistent across generations.

The Nano Banana 2.1 model card above is more cautious: it says character consistency is not always perfect between input images and the output. The Get started page says Flow has precautions to protect uploaded photos of people and that certain queries might not be generated. Check the face at feed size.

Where Flow is the right tool for a thumbnail

Flow earns its place wherever the picture, not the layout, is the hard part of the thumbnail.

  • A scene or a background. A red Martian sky, a flooded subway tunnel, a forest at dusk: one prompt, several outputs. Your words and face can go on top afterward.
  • A still from your own clip. If the video was made in Flow, Save frame turns its best moment into an image, so the thumbnail shows what the video shows. A frame is only as sharp as its clip, so check its width against YouTube's 640-pixel minimum if it is the thumbnail itself.
  • A channel already working in Flow. The projects page above describes Publish to YouTube for single clips, with a title, description, visibility, tags and category. It lists no thumbnail field, so check for one when you publish, or add your thumbnail in YouTube Studio afterward.

Keep the scene honest. As of October 2026, YouTube's Thumbnails policy page does not allow a thumbnail that misleads viewers to think they're about to view something that's not in the video. A generated storm over a real city is a promise; a still from the short film you made is a fact.

A brief to paste

For a sci-fi short made in Flow, save the best frame of its lighthouse, download it, attach it in WThumb as @1, pick Faceless, and paste this:

  • Words. LAST SIGNAL on one line.
  • Changes. The lighthouse from @1 large on the right under a red sky, the words big on the left, no people.

Google Flow, WThumb and doing it by hand

JobGoogle FlowWThumbBy hand in an editor
Where you startA text prompt, with optional ingredients: an upload, a saved video frame, a character or your avatarA YouTube link or upload of a thumbnail you like, or the title and what happens in the videoA blank canvas or a template
Your faceYour avatar with @me (not in the EEA, the UK or Switzerland), a photo ingredient or a characterPick your persona, or FacelessCut out your own photo and place it
Words on the imageWritten into the prompt; check every letterThe words in their own field, instructions in the Changes boxTyped and set by you
16:9 and 9:16An aspect ratio menu; the pages list no options, so check for both16:9, 9:16 or both from one submissionYou set each canvas size
The fileNo image size on the pages; check the download against YouTube's numbersA Full HD+ JPEG, 2048 x 1152 or 1152 x 2048, ready to upload as it isWhatever you export
A change afterwardDescribe an edit; earlier versions stay in the History panelChange it under the thumbnail, one credit per revisionMove anything by hand
Free to start50 Google Flow credits a day with or without a subscription, per the credits page; the Get started page lists a Google AI plan as a requirementTwo free credits for new accounts, no cardDepends on the editor
Three routes to a YouTube thumbnail, job by job, per Google's Flow and YouTube help pages in October 2026; where Google's pages are silent, the Flow column says check. The WThumb column is our own product.

Flow fits where the picture is the job, an editor where exact placement is, and a thumbnail maker where the job is a finished file in YouTube's two shapes. The routes also combine: a Flow still can go into a WThumb run as an extra image.

How to do this in WThumb

This makes the lighthouse short's thumbnail in WThumb, with the Flow still as an extra image. The reference brings the idea; the design is new.

  1. Choose Recreate, then the reference. Paste the YouTube link of a thumbnail you like in your niche (only that video's public thumbnail is fetched) or upload a PNG, JPEG or WebP.
  2. Type the words. Only the words that should print, LAST SIGNAL here, or press No text. Instructions go in the Changes box below.
  3. Attach the Flow still. Add the saved frame as an extra image and type @1 in the Changes box to place it. Up to four work, @1 to @4, and each is a brief too.
  4. Pick your persona, or Faceless. With a persona, your photo goes into the design. With Faceless, WThumb asks the image model to remove the person in the reference and fill the space in the same style; check the finished image.
  5. Choose the formats beside Make. 16:9, 9:16 or both, and one or two designs. Each comes back as a Full HD+ JPEG, 2048 x 1152 or 1152 x 2048.
  6. Press Make. Sign in if asked and wait about thirty seconds. Your words are checked by a moderation service first, and a refused request costs nothing.

No reference? Start from a prompt

The other way in is Prompt: the video's title and what happens in it, no reference. The Studio's note says it works best for real-life footage, so for a Flow short, Recreate fits better.

Revise it in words

If the words came out too small, press Change it under the thumbnail and describe one change. Each revision uses one credit, and the new version can be revised again.

Flow can paint the scene. A thumbnail is that scene with words, in YouTube's shape, readable at feed size.

Common questions

Is Google Flow free to use?
Google's own pages give two answers. As of October 2026, the Manage your Google Flow credits page says you receive 50 Google Flow credits a day with or without a subscription, and that non-subscribers can still generate images; the models page names Nano Banana 2 Lite as the default model available at no charge. The Get started page lists a Google AI Plus, Pro or Ultra subscription among its requirements, and says qualifying Workspace plans get Flow at no extra charge with 50 daily credits. Check what your own account shows.
Which image model does Google Flow use?
Three, as of October 2026. Google's Flow models page lists Nano Banana Pro, built for complex designs and the default for Google AI Ultra subscribers; Nano Banana 2 Lite, the default model available at no charge; and Nano Banana 2.1, the standard model. The images page says the prompt box defaults to Nano Banana Pro. The model name sits in the prompt box, and the credits page says the number of credits a generation needs depends on the model you use.
Can I put my own face in a Google Flow image?
Yes, three ways. As of October 2026, Google's Create & use your avatar in Google Flow page describes an avatar with your own highly accurate likeness, added with @me and not available in the EEA, the UK or Switzerland. You can also add a photo as an ingredient, or build a character from one or two images, which Google's projects page says keeps its face strictly consistent. Google DeepMind's model card for Nano Banana 2.1, one of Flow's image models, published October 6, 2026, says character consistency is not always perfect, so check the face first.
Can Google Flow make a 16:9 thumbnail for YouTube?
Check the menu. As of October 2026, Flow's images page says you pick an image aspect ratio before you generate, but it lists neither the options nor an output size. YouTube's Add custom thumbnails page recommends 16:9 for videos and 9:16 for Shorts, 3840 x 2160 pixels for videos, a minimum width of 640 pixels and formats such as JPG or PNG. Check the download against those numbers and the size limit, 2 MB from a phone or 50 MB from a computer; the same page says uploading your own needs a verified account.
Can I use Google Flow images on a monetized YouTube channel?
Google's Flow help answers with its terms rather than a plain yes. As of October 2026, the Get started with Google Flow page says the Terms of Service govern how Flow may be used and, quoting them, that Google won't claim ownership over the content you generate. Read those terms for your own channel before a Flow image goes on it; the next answer covers what YouTube asks you to disclose.
Do I have to disclose a Google Flow thumbnail on YouTube?
Not for a normal thumbnail. As of October 2026, YouTube's Disclosing use of GenAI content page lists production assistance, like using generative AI tools to create or improve a thumbnail, among what creators don't need to disclose. Flow's Get started page says every Veo, Omni or Nano Banana output carries an invisible SynthID watermark, and a visible one can be added. YouTube's thumbnail policy still applies: no thumbnail that misleads viewers about what is in the video.
ShareX