Removing text from a picture is easy when it sits over a plain wall or sky. It becomes harder when letters cover hair, product edges, fabric, or repeating patterns.
A useful workflow starts with a close selection, checks the repaired texture at full size, and repeats only the area that needs attention.
Key takeaways
- Preserve the original before editing.
- Select the letters and a narrow margin, not the whole object.
- Use separate regions for distant marks.
- Inspect edges and repeating patterns at full size.
1. Start with the original image
Use the highest-quality copy available. Screenshots and repeatedly compressed images contain block artifacts that make reconstruction less predictable.
2. Select text without swallowing the background
- 1
Upload the image
Choose JPG, PNG, or WebP up to 500MB.
- 2
Choose a removal mode
Use the subtitle option for text, or subtitle plus watermark when the mark includes a logo or larger graphic.
- 3
Resize the region
Leave a small margin around the characters but avoid unrelated edges.
3. Inspect texture and edges
Zoom in around straight product edges, hair, tiles, brick, and repeating fabric. These details make an imperfect fill easier to spot.
For two marks far apart, use two small boxes rather than one large box across the image.
When to make a second pass
- A faint letter edge remains.
- The fill is correct but slightly soft.
- One part worked while another needs a tighter box.
- A logo contains both text and a graphic shape.
Frequently asked questions
Which images are easiest to clean?
Text over sky, walls, soft gradients, and other simple backgrounds is usually easiest.
Can I remove several text areas?
Yes. Add separate regions, up to five of each type.
Should I select a large area around the text?
Usually no. A narrow margin gives the model more surrounding detail to work from.
