Skip to content

How it works

From a video idea to an uploaded thumbnail

Six steps, the decisions you make in each, and what Thumbora does in the background — including what happens to your credits.

  1. 01

    Describe your video

    Start a project with the video title, a short idea for the image and the text you want on the thumbnail. A good idea names the subject, the emotion and the setting — for example: “surprised host next to a glowing PC case, dark room, cyan light”.

    • Keep the thumbnail text to three or four words — the title already carries the details.
    • Write in Ukrainian, English or Russian. The model renders the text on the image, so check the spelling and fix it with an AI edit or in the editor if needed.
    • Not sure what to show? Our thumbnail ideas guide lists patterns by video type.
  2. 02

    Add photos and references (optional)

    Choose an identity profile if you want to appear on the thumbnail, add reference thumbnails and pick what to take from them, or upload your current thumbnail to redesign it.

    • Face photos need your consent and are used only for your generations.
    • References guide the composition — Thumbora doesn’t copy them.
    • Files: JPG, PNG or WebP up to 15 MB each.
  3. 03

    Choose style, model and quality

    Pick a style, the model, the resolution (1K, 2K or 4K, depending on your plan and the model) and how many images to generate at once.

    The generator shows the exact cost before you start. With Nano Banana 2.1, one image currently costs 10 credits at the base resolution.

  4. 04

    Generate

    When you press Generate, the credits are reserved and the request joins the queue. Each image is a separate task, so you can keep working while it runs.

    Most images are ready within a few seconds to about two minutes, depending on the model and resolution. You’ll see a notification — and get an email if you enabled it — when the generation finishes.

  5. 05

    Compare, refine, edit

    Results appear as versions in your project. Compare them side by side, star the best ones and keep notes.

    • AI edit — change details with an instruction, e.g. “make the background darker”.
    • Variations — explore alternatives of a version you like.
    • Editor — adjust text and layers by hand and save a new version.
  6. 06

    Export and upload

    Download the final version as PNG, JPG or WebP at the original size, 1280×720 or 2560×1440. If a size is larger than the generated image, the file is upscaled and labeled as such.

    Upload it in YouTube Studio in the video’s details. Our size guide lists YouTube’s current requirements.

Behind the scenes

What happens after you click Generate

No magic — a pipeline designed to be reliable and fair with your credits.

Prompt builder

Your inputs are combined with versioned templates for the selected mode and style. The exact prompt sent to the model is stored with the generation, so results are reproducible and problems can be investigated.

Queue and provider

Generations run in a background queue, never inside your browser request. Images are produced by Google’s Nano Banana models via reAPI; our worker submits each task, checks its status and retries temporary errors.

Private storage

Results are copied from the provider to Thumbora’s private storage right away, because provider links expire. You open them through short-lived signed links issued only to your account.

Credit ledger

Every credit movement — grant, reservation, charge, refund — is written to an append-only ledger with an idempotency key, so a retry can never charge you twice.

Ready to try it?

Sign up and start with free credits.