Text to image: generate from a prompt

Text to image is the core loop of tendre.AI: you describe a picture in words, and Stable Diffusion XL renders it on your own NVIDIA GPU. Nothing you type or generate is uploaded anywhere.

1. Write your prompt

In the main panel, type a concrete description of the image you want: subject, style, light, framing. The more specific the prompt, the closer the result matches what you had in mind.

2. Pick a size

Choose portrait, square or landscape depending on how you plan to use the image. Larger sizes take longer to render and use more VRAM.

3. Generate

Press Generate. The first render on a new size warms up the model and takes a little longer; the following ones are faster. Each generation uses a random seed unless you lock one.

4. Review and save

Look at the result in the preview panel. If it is close but not quite right, adjust the prompt and generate again, or lock the seed to refine the same image instead of getting a new one each time. Save the image to disk or send it straight to the built-in gallery.

5. Iterate

Generating is fast enough to treat as a conversation: change one detail at a time (light, angle, outfit) and regenerate rather than rewriting the whole prompt from scratch.

To write stronger prompts, read the Prompting guide. To render several ideas unattended, see Batch and queue.