Firefly: make a cover image with room for a title

Your deliverable

Create a 4:3 cover for an article called “A slow weekend coffee.” Put a cup on the right and leave the left clear for the title. Deliver the source image, a layout with the real title added, and the prompts from two attempts.

Follow the five steps

1. Sketch the layout first

   Mark a title area on the left of a 4:3 canvas and a cup area on the right. Describe one subject, its surroundings and the light. Keep extra props out of the first attempt.

   Check: The sketch makes the title and subject positions obvious.

2. Generate a baseline

   On the Firefly homepage choose Image → Generate image. Select a model in Model, choose 4:3 in Aspect ratio and enter the example in Prompt. Choose Photo if the selected model offers that content type, then Generate.

   Check: Save a version A with an intact subject and useful viewing angle.

3. Refine the composition

   If your model offers Composition Reference, upload your own sketch and compare its effect at a moderate Strength. Otherwise choose a supporting Firefly model or state the subject position more clearly in your text.

   Check: The title area is free of the cup, steam and distracting high-contrast objects.

4. Change the light once

   Keep the model, subject and composition. Change only “soft morning light” to “warm side light” and generate version B. Compare the cup edges, background and readability of the title area.

   Check: You can explain which version is more useful for the intended cover.

5. Add the real title before deciding

   Download your preferred image and add the actual title in your layout tool. Check it at desktop size and as a small mobile thumbnail; adjust the crop if needed.

   Check: The title stays legible, the cup remains intact and there is no unwanted generated lettering.

Try this original example

These examples were written by FindGoodAI for practice; adapt them to your own material.

An original cover-image prompt

A ceramic coffee cup on the right side of a wooden table, a little steam rising from the rim, a quiet window-side setting, soft morning light from the right, realistic photography, a large clean light-colored area on the left for a title, no text or logos in the image.

When the result needs work

There is still no space for the title

Reduce the number of objects and state the subject position. Add your own composition sketch where supported, then place the actual title to judge the result.

The image looks good but makes a poor cover

Compare it inside the real layout, at the required aspect ratio. Do not choose only from the generation grid.

You cannot tell which changes helped

Hold the model and composition constant. Change just one item, such as the light or a subject detail, and keep a record.

Before you finish

• The cup rim and handle have no obvious defects.
• The real title has a clear, readable place.
• The subject remains recognizable on a phone thumbnail.
• The source image, final layout and useful prompt are saved.

Match the video to the current interface

The video calls the entry Text to image. Adobe’s current guide uses Image → Generate image. The available controls and number of generated results depend on the model; do not expect every model to reproduce the video’s four-result layout. The official player provides English captions.

The official player’s chapters include prompt entry (0:17), aspect ratio and content type (0:48), composition reference (2:55) and effects (3:54).

Video by Adobe Learn · Jesús Ramirez. The written steps, examples and checklist are an editorial practice guide based on the linked official documentation. Practice time is an estimate; generation queues may add time.

Sources & references

Adobe Learn: Generate images from text ↗

Adobe official video player ↗

Adobe: Current text-to-image workflow ↗