> ## Documentation Index
> Fetch the complete documentation index at: https://docs.talkcut.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Generating images

> Create, edit, enhance, and transform images using TalkCut's AI image generation tools.

TalkCut provides several ways to create and modify images through the AI agent. All results appear as new images on your canvas — originals are never overwritten.

## Text-to-image

Describe what you want, and the AI generates an image. The agent selects the best model for your request automatically.

Examples:

* *"A serene Japanese garden in autumn with a stone path and red maple trees"*
* *"Product photo of a white sneaker on a marble surface, studio lighting"*
* *"Concept art of an underwater city with bioluminescent buildings"*

You can be as brief or detailed as you want. Including style references, lighting preferences, and composition details helps the AI produce more targeted results.

<Tip>
  For even better results, upload a reference image to the canvas and ask the AI to match its style. The [vision engine](/core-concepts/vision-engine) analyzes your reference and guides the generation.
</Tip>

## Editing images

Modify an existing image on your canvas with text instructions. The AI understands what's in the image and applies your changes while preserving the rest.

Examples:

* *"Change the sky to a dramatic sunset"*
* *"Remove the person in the background"*
* *"Add snow to the ground and rooftops"*
* *"Make the lighting warmer and more golden"*

The edited image appears on your canvas — the original stays untouched.

## Remove background

Remove the background from any image with a one-click operation. This is useful for:

* Isolating characters or objects for compositing
* Creating transparent assets for use in other tools
* Cleaning up generated images for further editing

Ask the AI: *"Remove the background from this image"*

## Upscale

Double the resolution of any image on the canvas. Upscaling works well for:

* Preparing images for high-resolution output
* Enhancing details in generated images
* Improving quality of compressed or low-resolution references

Ask the AI: *"Upscale this image"*

## Output format

Generated images are delivered as PNG files. TalkCut supports 10 aspect ratios:

| Aspect ratio | Use case                   |
| ------------ | -------------------------- |
| 21:9         | Ultra-wide cinematic       |
| 16:9         | Widescreen / YouTube       |
| 3:2          | Classic photography        |
| 4:3          | Traditional broadcast      |
| 5:4          | Print / social media       |
| 1:1          | Square / Instagram         |
| 4:5          | Portrait / Instagram       |
| 3:4          | Portrait photography       |
| 2:3          | Tall portrait              |
| 9:16         | Vertical / Stories / Reels |

Resolution varies by model — TalkCut automatically selects the best model for your request. Output ranges from 1K to 3K resolution depending on the task.

## Downloading images

Click any image on the canvas and use the **Download** button in the toolbar to save it as a PNG file to your device.

## Credit costs

| Operation         | Credits |
| ----------------- | ------- |
| Generate an image | 20–150  |
| Edit an image     | 20–150  |
| Remove background | 1       |
| Upscale (2x)      | 100     |

Credit cost varies by model. TalkCut automatically selects a model based on your request — more complex generations may use models that cost more credits.