Magic Prompt (Alpha)
Let’s be honest—prompt engineering is hard. You type “cool robot,” and the model gives you a tin can with legs. You want a “cyberpunk samurai robot with neon katana in a rain-drenched Tokyo alleyway,” but typing that out every time is exhausting. Enter Magic Prompt. 🪄 Think of Magic Prompt as your cool, artistically-gifted friend who knows exactly what to say to the AI to get it to produce its best work. You give it a rough sketch of an idea, and it translates that into a poetic, detailed instruction manual that the image generation models simply adore.Alpha Stage: Magic Prompt (and the whole Mynth app) is currently in early alpha. We are constantly tweaking the pipeline and adding support for more models. Your feedback on Discord is invaluable at this stage!
What is it?
Magic Prompt is an intelligent pipeline that intercepts your simple prompt and “remasters” it before it hits the image generation model. It’s not just appending “4k, trending on artstation” to the end of your string. It’s a sophisticated process involving normalization, inspiration hunting, and a smart rewriter that understands the specific “dialect” of the model you are using.Why use it?
1. Included in Service Fee
Magic Prompt does not add any extra charges to your generation. The compute power required is already covered by the fixed service fee for all image generations. You get better quality for the same price.2. It makes cheap models punch above their weight
Smaller, faster models (like the “Turbo” or “Eco” ones) can produce stunning results if you ask them nicely. Magic Prompt ensures you’re always asking nicely.3. You don’t need a PhD in Prompting
You shouldn’t have to memorize which keywords trigger specific art styles. Just say “oil painting style,” and Magic Prompt will expand that into “thick brushstrokes, textured canvas, chiaroscuro lighting…”Behavior & Control
By default, we always try to make your images look as good as possible. However, if you are a control freak (we love you too), you can manage this behavior using the structured prompt version of the API. Instead of passing a simple string, you can pass an object:Magic Logic:
- Default (not set): We use Magic Prompt if the model supports it. If it doesn’t, we just use your original prompt. No drama.
- Explicitly
true: We try to use Magic Prompt. If the model doesn’t support it yet, we’ll return an error so you know why your “magic” didn’t happen. - Explicitly
false: We skip the remastering entirely and send your prompt to the image model exactly as you wrote it.
Privacy & Your Data 🔒
You might be wondering: “If you use existing successful prompts to teach the AI, are you using my prompts?” By default: NO. We respect your privacy. Your prompts are yours. We do not add your generations to our public vector database for Magic Prompt training unless you explicitly opt-in to our community contribution program. So feel free to generate your secret projects; they stay secret.Supported Models
You might notice Magic Prompt is disabled or unavailable for certain models. This is intentional.- Quality First: We custom-tune the Magic Prompt pipeline for each specific model. It takes significant time and compute to “teach” our system how to prompt a specific model perfectly.
- Models with “Built-in Magic”: Some newer, highly advanced models (like Nano Banana Pro) already have internal prompt enhancement logic. Adding our Magic Prompt layer on top of these would be redundant—like putting a hat on a hat.
Example
You type:“A futuristic car in the desert”Magic Prompt converts it to:
Positive: “Wide angle shot of a sleek, aerodynamic concept vehicle speeding across a sun-scorched salt flat, heat haze, dust trails, metallic silver finish reflecting the harsh sunlight, futuristic design, syd mead, cinematic lighting, hyperrealistic, 8k resolution.” Negative: “rusty, old, broken, blurry, low quality, distorted wheels”Result: A wallpaper-worthy image instead of a generic clip-art car.
So go ahead, type “potato.” See what happens. We dare you.