New to the Thumbnails feature? Start with the Thumbnails overview article first, then come back here.
Before You Click "Create": Set Your Brief
The single biggest factor in thumbnail quality isn't the tool, it's the prep work. Spend 60 seconds answering these questions before you generate:
What's the one idea this thumbnail needs to communicate? Thumbnails that try to say multiple things say nothing. Pick one promise (e.g., "I tried this diet for 30 days" or "This trick doubled my views").
What emotion do you want the viewer to feel? Shock, curiosity, delight; pick one. Then make sure your face, objects, and color choices point toward that emotion.
What text (3–5 words) will go on the thumbnail? Text should complement your title, not repeat it. Use bold, high-contrast text readable on a small screen.
Do you have the right assets ready?
A clear face photo (shoulders and above) if you want your face in the thumbnail
Any subject or product images you want included
2–3 reference thumbnails from creators whose style you like
These prep steps make generation results dramatically more on-brief and reduce the number of iterations and credits needed.
How to Write a Great Prompt
The Thumbnails feature accepts plain-language descriptions. The more specific and visual your prompt, the more accurate the result.
The Prompt Formula
A strong prompt covers four elements:
[Subject] + [Setting/Background] + [Emotion or Action] + [Text on the thumbnail]
Examples: Weak vs. Strong
❌ Weak Prompt | ✅ Strong Prompt |
"Me looking surprised" | "Close-up of a man looking shocked, mouth open, pointing at a large red $100 bill, dark background, bold white text: 'I Found This'" |
"Cooking video" | "Person pulling a golden roast chicken from an oven, warm kitchen background, steam rising, bold yellow text: 'Best Chicken Ever'" |
"Tech review" | "Smartphone held up against a clean white background, dramatic side lighting, person's eyebrows raised in surprise, text: 'Worth It?'" |
Prompt Tips
Name specific colors. "Bold red text" and "neon green background" give the AI something to work with.
Describe the composition. "Close-up of the subject's face on the left, product on the right" sets the layout.
Reference an emotion, not a pose. "Looking shocked" works better than "looking at the camera."
Include thumbnail text in quotes. The AI treats quoted text as literal copy to place on the thumbnail.
Don't overload it. One strong idea beats five competing ideas. If your prompt is more than 3–4 sentences, trim it.
Use the word “subject” (not “me”). If you write “me,” the AI can take it too literally and sometimes crops in weird ways. Instead, describe the person as the subject (e.g., “subject looks shocked,” “subject pointing at…”).
Use semicolons to separate ideas. This helps the AI read your prompt like clear instructions. Example: “wind-blown tree in the background; stormy sky; subject looks afraid; bold white text: ‘DON’T GO OUTSIDE’”
Getting the Best Results from Each Starting Point
Using a Video Upload
After generation, if the AI picked the wrong moment from your video, use a Reference Thumbnail alongside the video to steer the style.
Using Text-Only
Best when your video hasn't been filmed yet, or when you want to test concepts early.
Be extra descriptive, since the AI has no visual reference to fall back on.
Add a Reference Thumbnail to compensate; it anchors the style even without a video.
Using Reference Thumbnails
Pick 2–3 references that share a similar style or mood, not just the same topic.
References influence color palette, composition, and energy, not the specific content.
Use your own past thumbnails as references to maintain a consistent channel look.
How to Use the Attach (Likeness) Effectively
Attach (Likeness) helps put you front-and-center in your thumbnails. When you attach your photos before generating, vidIQ uses them as a reference so your subject photo appears consistently across the thumbnail options, without you needing to manually swap your subject photo in afterward.
Photo Tips for Best Likeness Results
Do ✅ | Avoid ❌ |
Shoulders-and-above framing | Full-body shots or heavily cropped subject photo |
Good lighting, the subject's face is fully visible | Sunglasses, masks, heavy filters, or blur |
Clear solo photo | Group photos or cluttered backgrounds |
If your likeness looks too stylized (for example, it changes your age, face shape, or skin texture), add a short instruction in your prompt like:
“Do not apply filters to the subject image; maintain the subject’s original facial features.” This often helps the results stay closer to the original photo.
Attach (Likeness) vs. Face Swap: Which to Use?
| Attach / Likeness | Face Swap |
When it runs | Before generation | After generation |
Best for | Getting multiple options already featuring you | Swapping your face into one specific design you like |
Use case | I want every option to include the subject. | I like this thumbnail; now replace the subject face in it. |
Refining After Generation
If your first result isn't quite right, don't regenerate from scratch, refine it.
Smart Commands: How to Direct Them
Smart Commands work best when they're specific and action-oriented:
❌ Vague | ✅ Specific |
"Make it better" | "Increase contrast and make the background darker" |
"Change the text" | "Replace the text with 'Don't Make This Mistake' in bold red" |
"Fix the face" | "Make the expression more surprised, raise the eyebrows" |
Save Prompts That Work
When you generate a thumbnail you like, click Copy Prompt to Clipboard and save it in a doc. Over time, you'll build a library of reliable prompts you can reuse or adapt, great for maintaining a consistent visual style across your channel.
Common Mistakes to Avoid
Uploading a full screenshot as a subject image. The AI needs a clearly cropped face or object, not a whole frame. Crop before uploading.
Writing prompts that describe what you don't want. Focus on what you do want; the AI responds better to positive direction.
Skipping reference thumbnails. Even one reference significantly improves style consistency.
Using the same Likeness photo repeatedly. Variety in your photo set produces a more natural-looking likeness.
Trying to fix everything in one Smart Command. Break large changes into smaller, sequential instructions for better accuracy.