The two text fields
The dashboard gives you a main headline and a secondary line. Both are optional, and both are rendered into the image itself by the generator — not laid on top as an editable layer.
- Main headline — the words that get read at 168 pixels wide in a sidebar. "I LOST $10,000". Three to five words.
- Secondary line — a qualifier, a number, a date, a stakes line. Rendered smaller. Use it only when the headline genuinely needs a second beat.
- Both blank — the generator composes a clean image with no text, which is the right choice when you want to add your own text layers in the editor with exact control over font size, colour and position.
Writing a headline that survives the sidebar
A YouTube thumbnail is designed at 1280×720 and consumed at a fraction of that. Suggested videos render at roughly 360px wide, mobile home feed smaller again. That single fact drives every rule worth following:
- Cap it at five words. Six is usually one too many. If you need six, one of them belongs in the title instead.
- Do not repeat the video title. The title is right next to the thumbnail. Say the other half of the idea — title states the topic, thumbnail states the stakes.
- Lead with the number or the loss. "$10,000", "3 DAYS", "IT BROKE". Concrete nouns read faster than adjectives.
- Upper case for the headline. It is the one place where all-caps genuinely wins — it removes descenders and evens out the block shape at small sizes.
- Leave the face room. Text on one side, subject on the other. Text laid across a face costs you both.
Check it before you commit
Our free thumbnail preview tool shows your image at every size YouTube actually renders it, and the title checker shows where each surface truncates the title. Both run in your browser, no signup.
Generated text vs. text layers
There are two ways to get words onto a Thumblore thumbnail, and they are good at different things.
| Generated text (this page) | Text layers | |
|---|---|---|
| Looks like | Part of the artwork — lit, shadowed, integrated | A clean overlay on top |
| Control | The model decides placement and styling | Exact position, size, colour, rotation, stroke |
| Editable after | Via find & replace or AI rewrite | Yes, directly, any time, free |
| Costs credits | Part of the generation | No |
| Best for | Headlines that should feel painted into the scene | Series labels, episode numbers, exact brand type |
A common workflow for a series: generate with the headline in the image, then add a small text layer for the episode number so it can be updated per video without touching the artwork.
Fixing text that came out wrong
Generated text occasionally arrives misspelled or oddly kerned. You do not need to regenerate the thumbnail:
- Exact swap — Find & replace in the editor's Text tab. Type what is currently in the image and what it should say; the original font and placement are kept.
- Rewrite — the AI text rewrite in the same tab takes an instruction like "make the title say INSANE GAINS" and redraws it in place.
- Remove and replace — ask the custom edit chat to remove the text entirely, then add a text layer you control exactly.
When to ship with no text at all
Thumbnails with no text can outperform, particularly for vlogs, music and anything where the face or the object is the story. If you go text-free, the frame has to carry it: one clear subject, high contrast against the background, and an expression that implies a story. Generate with both text fields blank and a strong reaction.