Make pictures with AI, and change them one version at a time
Tinkerly makes pictures from a description and changes them when you ask. Describe what you want, pick an image model and a shape, and the picture is made in the background. Every change, by AI or by hand, is saved as a new version of the picture, so the one you started from is never lost.
Updated · By The Tinkerly team
How do you make a picture with AI?
Open AI images, describe the picture in plain words, choose the image model and the shape, and start it. The picture is made in the background while you keep working, and it appears with its prompt. From there you download it, or change it with the AI or by hand.
Step 1: Describe it
“A red bicycle leaning on a blue door, morning light, photo”. Say the style, the light and what matters most.
Step 2: Pick a model and a shape
The picker shows each image model’s price per picture. Shapes: Square, Portrait, Landscape and Wide.
Step 3: Look it up first, if it’s real
For a real place, product, species or building, turn on Web research so the prompt describes what it really looks like.
Step 4: Change it or download it
Describe a change, paint over the part to redo, or crop and adjust it yourself. Download the version you like.
How do you change only part of a picture?
Choose Paint area, brush over the part to change and describe what should be there. The model sees the whole picture with that area marked, and the app keeps its answer only inside the painted area. Everything outside it stays exactly as it was, pixel for pixel.
| Question | What it does | Cost |
|---|---|---|
| Describe | The AI changes the whole picture as you describe it | The picture price |
| Paint area | Only the part you painted changes | The picture price |
| Adjust | Crop, rotate, mirror, resize, brightness, contrast, saturation | Free |
| Upload | Start from your own PNG, JPEG, WebP or GIF | Free to upload |
Describe and Paint area work with image models that can edit pictures; the others only make new ones.
Which image models can you use?
The image models the site has switched on, each with its price per picture shown in the picker before you start. Some models can also edit pictures, using your picture as the starting point. Others make new pictures only. A model that can’t edit simply isn’t offered for changes.
Image models are separate from the chat models: picking a picture model doesn’t change which model answers your chats.
What does Web research do for a picture?
Before the picture is made, a chat model searches the web for what the subject really looks like and writes the prompt for the image model. The prompt it wrote and its sources are kept with the picture, so you can see what it found and why it looks that way.
Research is billed like a chat message with web search, on top of the picture. If it fails, the picture is made from your prompt as you wrote it, and only the searches that ran are charged.
Can you use the pictures elsewhere in the app?
Yes. Every picture you make is saved in your Library as your own file. Pictures are shown only to you: there’s no public gallery of what people make. You can put them on slides and in documents, or ask for one in a chat.
- From a slide, make or change a picture and put it straight back on the slide.
- Documents the AI writes can include pictures, made when you keep the document.
- In a chat, ask for “a picture of…” and it’s made with the image model you used last.
What does a picture cost?
Each picture costs its model’s price per picture, shown before you make it. The price comes from what the model costs to run, so a quick model costs less than a detailed one. A failed picture is never charged, and cropping or adjusting by hand is free.
Free accounts get 15 credits every day, which work on the image models marked for free credits. See every chat model’s price on the pricing page.
What can’t it do yet?
Pictures are between 256 and 2048 pixels a side, and a prompt is at most 4,000 characters. It makes pixel pictures only, not vector (SVG) files. One change runs on a picture at a time, and a few pictures at once per person.
Models running on your own computer answer chats, but they don’t make pictures: pictures always use the site’s image models.
Text inside pictures depends on the model: check spelling in signs and labels before you use one.
Keep reading
Make your first AI picture
Sign up free with no card, describe the picture, and change it until it’s right.