Thumbnail Lab
How to Make Cinematic YouTube Thumbnails (With the Title Behind You)
YouTube ThumbnailsCinematicText Behind SubjectThumbnail LabCreator Tools
iReelz TeamSeptember 26, 20266 min read
iReelz TeamSeptember 26, 20266 min read
You know the look. A face nobody has, skin like wax, colour turned up until the sky glows, and a title that is almost spelled right, melted into the picture where you can't fix it. Viewers know it too. They have seen thousands of them, and they scroll past in the fraction of a second it takes to recognise one.
A cinematic thumbnail does the opposite. It looks like a still from a film: one light, real shadows, grain, a place you want to be, and a title set with care. It feels made by a person with taste, and that is what earns the click. This post is how to make one, step by step, and how the Thumbnail Lab does all of it from a single line.
A cinematic thumbnail does the opposite. It looks like a still from a film: one light, real shadows, grain, a place you want to be, and a title set with care. It feels made by a person with taste, and that is what earns the click. This post is how to make one, step by step, and how the Thumbnail Lab does all of it from a single line.
Why most generated thumbnails look fake
Ask a chatbot for a thumbnail in one prompt and you are asking it to be three people at once: the photographer, the designer and the typesetter. It does all three badly, in the same way every time.
• The picture defaults to glossy. Left alone, an image model reaches for the stock-photo look: everything sharp, everything saturated, a golden glow on every face.
• The letters are drawn, not typed. The model paints shapes that look like text. Some come out wrong, most come out soft, and none of them can be edited afterwards.
• Nobody decided where anything goes. The title lands on top of whatever happens to be there, so the face and the words fight for the same space.
The fix is not a better prompt. It is splitting the job back into the three jobs it always was.
• The picture defaults to glossy. Left alone, an image model reaches for the stock-photo look: everything sharp, everything saturated, a golden glow on every face.
• The letters are drawn, not typed. The model paints shapes that look like text. Some come out wrong, most come out soft, and none of them can be edited afterwards.
• Nobody decided where anything goes. The title lands on top of whatever happens to be there, so the face and the words fight for the same space.
The fix is not a better prompt. It is splitting the job back into the three jobs it always was.
A thumbnail is a film poster the size of a stamp. Direct the picture, grade it, then set the type.
— iReelz team
1. Write the picture like a film scene
Before anything is rendered, describe the frame the way a director would brief a camera operator. Five decisions do most of the work:
• A lens. 28mm for a place, 50mm for a person, 85mm for a face. Naming one stops the wide, distorted default.
• A film stock. Kodak Portra for warm, muted daylight; Fuji Pro 400H for pastel; Cinestill 800T for night and neon. Each one brings a whole palette with it.
• One light. Low sun from the right, a window, a street lamp. One source, real shadows. Flat, even light is what makes a picture look generated.
• A candid moment. Someone doing the thing the video is about, not posing at the lens. Hands busy, eyes on the work.
• A quiet area for the title. Decide now where the words go, and keep that part of the frame simple: sky, a plain wall, calm water.
• A lens. 28mm for a place, 50mm for a person, 85mm for a face. Naming one stops the wide, distorted default.
• A film stock. Kodak Portra for warm, muted daylight; Fuji Pro 400H for pastel; Cinestill 800T for night and neon. Each one brings a whole palette with it.
• One light. Low sun from the right, a window, a street lamp. One source, real shadows. Flat, even light is what makes a picture look generated.
• A candid moment. Someone doing the thing the video is about, not posing at the lens. Hands busy, eyes on the work.
• A quiet area for the title. Decide now where the words go, and keep that part of the frame simple: sky, a plain wall, calm water.
2. Grade it like film, not like an advert
The grade is what makes a thumbnail feel expensive. Aim for muted, slightly desaturated colour, blacks lifted a touch so the shadows are never pure black, fine grain, and a little halation where the highlights bloom. Leave out HDR, crushed contrast and the orange-and-teal slider pushed to the end. On a phone screen next to a feed full of neon, restraint is what stands out.
3. Set the words, never draw them
The title is type, and it should be typeset. Keep it short, one to five words, and give it a hierarchy: a small kicker above (a series name, a place, a day), the title, and a one-line tagline under it. Pick a type style that suits the video, and keep it for the whole series so your channel reads as one thing.
Typeset words are always sharp, always spelled the way you wrote them, and always yours to change. You can make three versions of the same picture with three different titles and test which one gets the clicks.
Typeset words are always sharp, always spelled the way you wrote them, and always yours to change. You can make three versions of the same picture with three different titles and test which one gets the clicks.
4. Put the title behind the subject
This is the move that makes a thumbnail look designed rather than generated. Cut the person out, set the title across the frame, then lay the person back on top, so they stand in front of the words. Film posters have done it for decades because it gives a flat picture depth.
It only works if the picture was planned for it. The person has to stand inside the band where the title goes, overlapping it, with plain sky or haze around them, which is why step one decides where the title sits before anything is rendered.
It only works if the picture was planned for it. The person has to stand inside the band where the title goes, overlapping it, with plain sky or haze around them, which is why step one decides where the title sits before anything is rendered.
5. Match the look to the video
Not every video wants the same treatment. These are the five looks we use most, and the kind of video each one suits:
• Epic — wide, glowing capitals and one small figure against a huge landscape. Travel, adventure, series openers.
• Cinematic — a tracked serif, a thin rule and a tagline over an establishing view. Places, drone footage, documentaries.
• Modern — a kicker, a clean sentence-case title and one word in your accent colour. Reviews, tutorials, gear, food.
• Editorial — one huge italic word, like a magazine cover. City walks, vlogs, a single idea.
• Poster — condensed capitals and a billing line. Action, sport, anything with momentum.
• Epic — wide, glowing capitals and one small figure against a huge landscape. Travel, adventure, series openers.
• Cinematic — a tracked serif, a thin rule and a tagline over an establishing view. Places, drone footage, documentaries.
• Modern — a kicker, a clean sentence-case title and one word in your accent colour. Reviews, tutorials, gear, food.
• Editorial — one huge italic word, like a magazine cover. City walks, vlogs, a single idea.
• Poster — condensed capitals and a billing line. Action, sport, anything with momentum.
Doing all of it from one line
Everything above is what the Thumbnail Lab does for you. Type what your video is about, the way you would say it to a friend, and it comes back with three concepts, each a different approach to the same idea: the wide place, the close moment, the person mid-action.
Directed, then rendered
Each concept is written as a scene first (lens, film stock, light, where you stand, where the title goes) and rendered at 2K with a film grade.
Words set in your font
Titles, kickers and taglines are typeset in your brand kit font if you have one, in a choice of styles, and free to change after the render.
Title behind the subject
The subject is cut out of every render, so one switch puts the title behind the person, the rider or the figure on the dune.
You in all three
Add a photo and the set puts you in every concept, looking like yourself: absorbed in the moment, never posing.
Start from a look
Pick Epic, Cinematic, Modern, Editorial, Poster or Night city, or press "Try this brief" on any example to start from that exact thumbnail.
YouTube and Shorts sizes
Export 16:9 for YouTube and 9:16 for Shorts from the same render, at its full resolution. No upscaling, no blur.
What it costs
You pay per set, and a set is three concepts at 2K. The $4.99 pack is 4 sets, $9.99 is 9 and $19.99 is 20, so about a dollar a set. On Fast, at 1080p, each pack goes twice as far. Everything after the render (the words, the type style, the title behind the subject, the exports) costs nothing, so you can try as many titles as you like on a picture you already have. Packs never expire, you can buy one right in the editor, and opening the Lab is free.
Common questions
Is it AI? Yes, art-directed. Every concept is planned like a scene before it is rendered, and the words are typeset afterwards rather than drawn, which is why the result reads as a film still instead of the glossy look viewers have learned to skip.
Can the text come out misspelled? No. The words are never part of the picture: the title is set in your browser from what you typed, and you can change it at any time.
Who owns the thumbnails? You do. Use them on any channel, commercially, with no attribution.
Can the text come out misspelled? No. The words are never part of the picture: the title is set in your browser from what you typed, and you can change it at any time.
Who owns the thumbnails? You do. Use them on any channel, commercially, with no attribution.
Make your next thumbnail cinematic
One line in, three film-grade concepts out, with the title behind you and sizes for YouTube and Shorts.
Open the Thumbnail Lab
