HACKER Q&A
📣 andoando

Easiest way to create simple illustrations?


Id like to for example generate simple line drawn images of feet/rollerblading in various angles with arrows.

Or if I was making a tutorial on archery, generate various frames of a person holding a bow at different draw lengths, etc.

Now obviously theres a ton of ways to do this, what I am wondering if what's the laziest way I can do them?

LLMs sound great, but there's a couple issues:

1. They don't seem to be so great when you need something precise. When prompted to change something in particular, even if they get it right, they often mess up something else.

2. No consistency in models. This is something I've been wanting in general (use the same character model for example in a different setting for example). I assume it exists, just haven't seen it mainstream yet.

3. Need something vectorized/modifiable.

So I am thinking the approach would be something like.

1. Generate LLM image. 2. Vectorize it. 3. Edit it manually with some Adobe software

  👤 wruza Accepted Answer ✓
When prompted to change something in particular, even if they get it right, they often mess up something else.

I have no solid advice for you, but this particular issue is easily solved with “inpaint mask” mode. https://stable-diffusion-art.com/inpainting_basics/

Edit: If you’re fine with anime-ish looks, you can take some simpler anime model because it will be easier to vectorize. E.g. “visual novel” on civitai, or just use a detail lowering lora. You can img2img the result back with normal model later, if you want.

Also use controlnet and optionally 3d pose editor to create main poses.

👤 brudgers
Hiring someone is an easy way.
👤 Fricken
You can draw the images directly in Adobe illustrator. You can import a photo and trace over it, to get that airplane emergency instructions look.
👤 andrewfromx
Yeah, LLM-generated images aren’t quite there for precise and consistent modifications, and they definitely don’t produce clean, vectorized results. If you want the laziest approach with the most control, here are some options:

1. AI-Assisted Sketching (Fast & Adjustable) • Use an AI tool for rough generation, then clean up manually. • Lazy Workflow: 1. Use tools like DALL·E, Stable Diffusion (ControlNet), or Deep Floyd IF to generate rough sketches. 2. Convert the image to a vector format using Adobe Illustrator (Image Trace) or Inkscape (Trace Bitmap). 3. Manually tweak lines to make them precise.

Pros: Speeds up initial drawing, retains some AI assistance. Cons: Still requires manual cleanup.

2. Pose & Trace (Super Lazy & Customizable) • Use a 3D model as a reference, then auto-trace or hand-trace. • Lazy Workflow: 1. Use PoseMyArt, JustSketchMe, or Blender (with rigged models) to pose a 3D figure. 2. Screenshot or export as a line drawing. 3. If needed, vectorize in Illustrator/Inkscape.

Pros: Total control, precise angles, reusable. Cons: Minor learning curve with posing tools.

3. Pre-Made SVG Libraries (Even Lazier) • Find existing vector illustrations and modify them. • Lazy Workflow: 1. Download from Noun Project, Freepik, or SVG Repo. 2. Modify in Figma, Inkscape, or Illustrator.

Pros: No drawing required, quick edits. Cons: Limited flexibility, may not match style perfectly.

4. Procedural Line Art with Code (Automate It) • Generate line drawings programmatically using code. • Lazy Workflow: 1. Use Processing, Manim (Python for animations), or P5.js. 2. Script different poses and generate SVGs automatically.

Pros: Extreme consistency, great for tutorials. Cons: Requires coding skills.

5. Vectorize AI-Generated 3D Renders • Lazy Workflow: 1. Use RunwayML, Kaedim, or Blender to generate a 3D render. 2. Convert it into a line drawing with a Freestyle render (Blender) or vectorize it.

Pros: Consistent, reusable models. Cons: Needs a bit of setup.

👤 dapperdrake
Had good experiences with MetaPost (it comes with TeXLive in Debian, if I recall correctly).

Even used it to produce graphics live in a zoom call with a client. TeX level of elegance.

👤 pestatije
the obvious might not have come to mind: pay an illustrator to do it