Text to visual

Skill PicsArt/gen-ai-skills/skills/text-to-visual

Shared skills, prompts, and tooling for Picsart's gen-ai CLI and generative-AI workflows

Install

npx -y skills add PicsArt/gen-ai-skills --skill text-to-visual

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

4 stars4 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Generate matching visuals from text via Picsart gen-ai.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

3.6 KB, as published. Nobody here has run it

Text to Visual

A single skill that covers every "given text, produce a matching image" workflow Picsart's gen-ai CLI supports. Use when the user has any kind of written content — a paragraph, a blog draft, a URL — and needs visuals generated from it. Replaces three narrower skills with one entry point and three mode references.

Input: text (paragraph, article, or URL). Output: one or more images matched to the text's tone, topic, and target placement.

When to Use

Mode	Trigger phrases	Reference
single	"match a visual to this paragraph", "inline image for this section", "one visual for this content"	`references/modes/single.md`
article-set	"illustrate this blog post", "hero + inline visuals for an article", "full visual set for a draft"	`references/modes/article-set.md`
og	"OG image for this URL", "open graph preview", "Twitter card image", "dynamic meta image"	`references/modes/og.md`

If the user wants product photos transformed, that's product-photo-studio. If they want video, that's gen-ai-use.

Prerequisites

Picsart gen-ai CLI installed and authenticated:

curl -fsSL https://picsart.com/gen-ai-cli/install.sh | bash
gen-ai login
gen-ai whoami

Per-mode setup (caching, serverless endpoints, font choices) is documented inside each mode reference.

How to Run

Identify the mode from the user's request using the table in When to Use.
Load the corresponding mode reference: Read references/modes/<mode>.md.
Follow the procedure described there — extract text signals, build prompt, generate.
Return to this SKILL.md only when switching modes mid-task.

Quick Reference

# Single image from a prompt
gen-ai generate --model <model> --prompt "<derived from text>"

# Estimate cost first
gen-ai pricing --model <model> --count <N>

# Browse available models
gen-ai models

Prompt-construction patterns (how to derive a prompt from a paragraph, an article, or a URL's metadata) live in the individual mode references.

Procedure

Shared outer loop:

Extract signals — pull subject, tone, palette hints, and target dimensions from the input text.
Build prompt — translate signals into a gen-ai prompt; each mode has its own template.
Estimate — gen-ai pricing before committing for multi-image runs.
Generate — invoke gen-ai generate. Stream progress.
Place — drop into the right slot: PDP, blog frontmatter, OG meta tag, social variant.

Pitfalls

Don't generate from raw text. Always extract signals first; raw paragraphs produce literal, lifeless images.
Match aspect ratio to placement. OG = 1200×630, blog hero = 16:9, social = varies.
Cache OG images. Don't regenerate on every page view — see references/modes/og.md.
Mode-specific pitfalls live inside the individual mode references.

Verification

# Confirm output exists and matches expected dimensions
gen-ai inspect outputs/<run>/<image>.png

# Spot-check the visual matches the source text by re-reading both side by side