Logos and Vectors

Make a simple mark or icon and deliver it as an SVG. The text-to-image models here return raster images (PNG or JPEG). A second step turns the chosen raster into an SVG.

Inputs to collect

• Brand name and what it does: one line.
• Mark type: symbol only, symbol plus name, or name only (wordmark).
• Style: flat, geometric, line, rounded, hand-drawn; number of colors and the colors (hex if the user has them).
• Use: app icon, favicon, website header, print. Small uses need fewer, thicker shapes.
• Existing raster logo, if the task is vectorize-only: a public https URL or a TokenLab result URL.

Models

<!-- tokenlab:models:start --> <!-- Generated from the live catalog by scripts/sync_models.py. Do not edit; change preferredModels in catalog.json. -->

Operation	Model	Pick	Use
text-to-image	recraft-v4.1	default	Mark or icon (raster)
text-to-image	recraft-v4.1-pro	alternative	Mark at a larger size (not live-tested)
text-to-image	ideogram-4-5	alternative	Mark with words
text-to-image	recraft-v4.1-flash	alternative	Low-cost draft
image-edit	recraft-vectorize	default	Raster to SVG
image-edit	picsart-image-vectorizer	alternative	Raster to SVG, alternative

This list refreshes daily from the live catalog. Preferred models come first; otherwise models are ranked by their 24-hour success rate. Before a call, confirm the model with GET /v1/models/{model} and read its price from GET /v1/models/{model}/pricing.

<!-- tokenlab:models:end -->

How to choose, by stage:

• Mark or icon (text-to-image): Recraft V4.1. Use the Pro tier for a larger size and the Flash tier for a low-cost draft. Use Ideogram 4.5 when the mark has words.
• Raster to SVG (image-edit): Recraft vectorize, or the Picsart vectorizer as an alternative.

All use /v1/images/generations. Exact fields come from each model's contract.

Workflow

Run each call with tokenlab-run (contract → estimate → call → poll → result). Commands below use its helper, ../tokenlab-run/scripts/tokenlab_run.py, a path relative to this skill's folder. If you installed this skill without tokenlab-run, send the same requests with curl. --wait stops after 100 s; a longer task keeps running, so continue with the resume command it prints.

1. Read the contract for each model. Each Recraft V4.1 text-to-image model lists one size value and n 1 only: 1024x1024 for the base and Flash tiers, 2048x2048 for the Pro tier.
2. Estimate the cost from GET /v1/models/{model}/pricing: marks times the price per image, plus one vectorize call per chosen mark.
3. Marks. Generate 3–4 options, one call each, with different shape ideas (see Technique). Send operation: "text-to-image", the model's one size value, and n: 1.
4. Read each response. A sync response has data[].url. A response with status: "pending" is a task: poll GET /v1/tasks/{id} every 5–10 s until completed or failed.
5. Let the user pick one mark. Vectorize only the pick.
6. SVG. Send {"model": "MODEL_ID", "operation": "image-edit", "image_url": "MARK_URL"}. Take MODEL_ID from the Models list above for the vectorize (image-edit) operation. Do not send output_format: "svg"; it returns 400 (the field accepts only png, jpeg, webp). Leave output_format out.
7. Download the result and check the file. In tests, both vectorizers returned real SVG (Content-Type: image/svg+xml) although the result URL ends in .png. Save it with a .svg name.
8. Return the SVG file or URL, the raster URL, the model per stage, and the task IDs.

Technique

• Say "flat vector logo mark" first. Recraft's own API example uses the style cue "flat vector" in the prompt.
• Few shapes, few colors. "A simple rounded bread loaf made of two solid colors, warm orange and dark brown." Simple shapes vectorize into clean paths; gradients, shadows, and texture turn into hundreds of small shapes.
• Plain white background, centered, no text. Add "No text, no gradients, no shadows" unless the user wants a wordmark.
• Words in a logo: make the wordmark with tokenlab-text-in-image (quoted text, letter style described), check the spelling, then vectorize that result.
• The background becomes a shape. Both vectorizers put the white background into the SVG as a full-size filled path. Delete that path (or set it to none) when the user needs a transparent mark.
• Check small sizes. Look at the mark at 32 px. If it turns into a blob, ask for fewer, thicker shapes.

Quality bar

• The mark reads clearly at 32 px and at full size.
• Only the colors the user asked for, as flat fills.
• The SVG opens in a browser, has a small number of paths, and has no raster image embedded.
• Any words are spelled right.
• The mark does not copy a known brand's logo.

Related skills

• tokenlab-run: the call, poll, and result mechanics.
• tokenlab-text-in-image: wordmarks and posters with exact text.
• tokenlab-edit-image: change an existing raster image.
• tokenlab-photoreal-stills: realistic photos.