Walkthrough

How Thumbly works

From a single reference photo to a finished YouTube cover in four steps.

STEP 01

Upload a reference image

Provide a photo of yourself, a product, or a scene (PNG, JPEG, or WEBP). The image is stored in your private Convex cloud storage and passed as a visual conditioning input to the generation model.

Note: A sharp subject with clear lighting produces the most consistent style transfer.

STEP 02

Select a preset and add optional text

Choose from six preset styles: Cinematic, Bold, Minimal, 3D Render, Anime, or Neon Glow. You can also supply a custom prompt string to describe specific elements (like props, background lighting, or expressions).

Note: The preset prompt template and your custom text are merged into a single prompt.

STEP 03

Asynchronous synthesis

The generation job is queued on our Convex backend, which calls the image model via OpenRouter at native 16:9 aspect ratio. Synthesis typically finishes in 5 to 30 seconds. One credit is deducted from your balance upon dispatch.

Note: If the upstream model returns an error, the credit is immediately refunded.

STEP 04

Download the 16:9 file

The completed image is saved in your gallery and available for direct download. The aspect ratio matches YouTube Studio's 16:9 upload requirement, so no manual cropping or framing is required.

Note: All past generations remain in your gallery history.

Request beta access