Choosing a model
Which image model should I start with?
Which image model should I start with?
Flux — the widest range in Pixio, strong prompt adherence, and the best LoRA support. Compare against Stable Diffusion and OpenAI on the same prompt before settling.
I need text inside the image
I need text inside the image
Try Ideogram — it’s the family best known for legible text in images.
I need a logo or something that scales
I need a logo or something that scales
You want vector output, not a raster image. Recraft is the main source of SVG output in Pixio.
I want to edit an existing image, not make a new one
I want to edit an existing image, not make a new one
Why do two models with the same capability behave so differently?
Why do two models with the same capability behave so differently?
Same capability doesn’t mean same controls. Two text-to-image models can accept completely different inputs. Check the tooltips on each input in Generate rather than assuming a setting exists.
Quality and prompts
My results don't match my prompt
My results don't match my prompt
Usually the model, not the prompt. Prompt adherence varies a lot between families — try the same prompt on a Flux model before rewriting it. If one model consistently ignores you, it’s the wrong model for that job.
How do I get consistent results across images?
How do I get consistent results across images?
Fix the seed if the model exposes one — same seed and same prompt gives you a repeatable starting point. For a consistent subject or character across many images, train a LoRA.
Can I get a specific person or product to look right every time?
Can I get a specific person or product to look right every time?
That’s what LoRA training is for. Train on 15–30 varied images of the subject, then trigger it with its trigger word. See Model Training.
The image is right but too small or soft
The image is right but too small or soft
Don’t regenerate — upscale. Use the Upscale or Image Upscale chained action, or Magnific directly.
Costs
How much does an image cost?
How much does an image cost?
Depends on the model — from a couple of credits up to a few dozen. The cost is shown in the picker and again before you submit. See Credits & Plans.
How do I not waste credits?
How do I not waste credits?
Draft on a cheap model at low resolution to get composition and prompt right, then rerun the keeper on a better model. Iterating on the expensive model is the most common way people burn a balance.
I was charged for an image I don't like
I was charged for an image I don't like
Successful generations cost credits even if the output isn’t what you wanted. Only failed runs are free. That’s the reason to test cheap first.
Inputs and outputs
How do I use my own image as a reference?
How do I use my own image as a reference?
Upload it in the generation form, or pick an existing file from your assets. Uploads are reusable — they land in the Uploads tab of My Assets.
Where did my image go?
Where did my image go?
My Assets → Generated, and in the history column in Generate. The prompt is in Prompt History.
Can I remove the background?
Can I remove the background?
Yes — the Remove BG chained action, or a background-removal model in 3D & utility.