For the dumb/smart prompt research component of Project 2.
Two steps: write your prompt with an LLM, then run it through an image generator.
These are language tools, not image generators. Paste your creature description in and ask it to build a detailed text-to-image prompt. For the smart prompt, include all your worldbuilding Q&A answers too.
All of these have usable free tiers. Start here if you don't want to pay anything.
| Tool | Best For | Free Tier |
|---|---|---|
|
Google Gemini recommended |
Top-tier quality, huge free allowance. Most students should start here. | ~100 images/day |
|
Leonardo AI recommended |
Best presets for character and concept art. Realtime Canvas for sketch-to-image. | 150 tokens/day |
| ChatGPT (GPT-4o) | Write and generate in one conversation. Refine iteratively in chat. | ~3 images/day |
| Ideogram | Strong stylized illustration. Best if your design involves readable text or labels. | ~40 images/week |
| SeaArt | 200,000+ community models. Strong for anime, fantasy, and creature styles. | 150 stamina/day |
| Kling | Multi-image reference blending — combine traits from up to 10 reference images. | ~66 credits/day |
| Krea | Real-time sketch-to-image. Draw a rough shape and watch it generate instantly. | Free tier available |
| Reve Image | Best prompt adherence for long, complex, multi-element descriptions. | Free tier available |
| Adobe Firefly | IP-safe (trained on licensed content only). Integrates with Photoshop. | Free tier available |
These require a subscription but produce noticeably higher quality output.
| Tool | Best For | Cost |
|---|---|---|
|
Midjourney highest quality |
Best overall artistic output. Niji 7 model is tuned specifically for stylized characters. | From $10/mo |
| Runway | Primarily a video tool — useful for animating a creature you designed elsewhere. | From $15/mo |
| Pika | Image-to-video animation. Take a still creature design and bring it to life. | From $8/mo |
These run on your own machine or require more technical setup. Higher ceiling, steeper learning curve.
| Tool | Best For | Cost |
|---|---|---|
| Stable Diffusion | Fully open source. Thousands of community fine-tuned models for fantasy/creature work. Runs locally via ComfyUI or Automatic1111. | Free (local) |
| Flux | Exceptional prompt accuracy. Kontext model edits one part of a character while keeping everything else consistent. | Free open-weight / pay-per-image API |
Craiyon (formerly DALL-E mini) and similar low-quality generators are not accepted for this assignment.
The dumb vs. smart prompt exercise only works if your tool produces meaningfully different results based on prompt quality and detail. Low-quality generators flatten everything into the same muddy output regardless of how specific or researched your prompt is — which defeats the entire point of the exercise.
If you are unsure whether a tool is on this list or not, ask before submitting.