Most AI “vector” tools still start with pixels and trace them afterward, which is how you end up with a logo made of a few hundred wobbly anchor points. Ideogram’s new vector model, announced October 22, is pitched as generating the SVG itself, with separate shapes, editable colors, and smooth gradients you can open in a vector editor and keep working on.

What changed

Ideogram Vector (the API calls it ideogram-vector-1, and coverage is calling it Vector 1.0) does two jobs:

  • Text to SVG. Describe what you want, optionally add reference images, and it generates SVG artwork. Ideogram suggests logos, prints, and illustrations as uses.
  • Image to SVG. Upload a JPEG, PNG, or WEBP (up to 50 MB) and it converts it to a vector, no prompt needed. There are two modes: one that keeps more fine detail and a “simple” one that draws fewer, simpler shapes.

Ideogram says you can scale the results without losing sharpness, change their colors, and select and adjust individual shapes in a vector editor. It’s live in the Ideogram app and the API, and the API has a dry run option that quotes a price without generating anything, which is handy for checking costs before you wire it into anything.

Why it matters for designers

The useful part is getting something closer to a working file rather than a picture of one. If the shapes really do arrive separated, with real color values on them, then recoloring a mark for a brand palette or pulling one element out of an illustration is a normal vector edit instead of a cleanup job.

The image-to-SVG side might matter just as much day to day. Plenty of jobs start with a client’s low-res logo or a scanned sketch that needs to become a clean vector, and Illustrator’s Image Trace has been the default answer for years. This is another option to compare it against, especially with the simple mode for flat marks.

What to watch out for

Editability is Ideogram’s claim, not a tested result yet. One early analysis, from the blog of the AI model gateway company OrcaRouter, points out that nobody outside Ideogram has published a structured test of what’s actually in the files, and that Ideogram hasn’t said whether type comes through as live text or outlines, or how many anchor points simple shapes use. Open a few outputs and check before you build a workflow on it.

Layer grouping on conversions isn’t there yet. The vectorizer’s API docs list an option to group shapes into layers, then say plainly that grouping is not yet available, so expect to organize converted files yourself.

There’s no editing step. You can generate a new SVG or convert an image, but you can’t ask the model to change an SVG it already made. Revisions happen in your own vector tool, or by generating again.

And for logos specifically, a generated mark still needs the usual checks: that it’s actually distinctive, that it isn’t too close to an existing mark, and that you’re comfortable with the rights. The sidebar covers what we could and couldn’t confirm.