AI Art Direction Studio
You already know what
you want the image to look like.
Writing the prompt is the hard part.
Every AI image model has a completely different way it wants to be spoken to. Midjourney likes short, comma-separated keywords with parameter flags at the end. Flux responds better to long, descriptive prose written in full sentences. DALL-E prefers something concise because it rewrites your prompt anyway.
Learning all of that takes months of trial and error that most people just do not have. FrameBrief handles the translation. You answer questions about what you want, and we write the prompt in the right language for each model.
5 free prompts to start. No credit card needed.
Your subject
Premium product photograph on brushed aluminium surface, octagonal softbox key light, shallow depth of field with creamy bokeh --ar 4:5 --s 100 --no text watermark
What most people type into Midjourney:
professional photo of wireless earbuds white background
What FrameBrief writes for the same subject:
Premium wireless earbuds on brushed aluminium surface, octagonal softbox key light at 45 degrees, shallow depth of field, warm neutral palette --ar 4:5 --s 100 --no text watermark logo
0+
vocabulary entries mapped
What we actually do
We have already figured out what each model responds well to. You just tell us what you want to make.
Over the past year, our team has run thousands of tests across every major AI image model. We tracked what phrasing actually produces good results, what parameters matter, and what you can safely leave out. All of that research lives inside FrameBrief as a vocabulary of over 1,200 option-to-model translations.
When you answer questions in FrameBrief, every option you tap maps to a specific phrase we know works for Midjourney, a different phrase for Flux, and yet another for DALL-E. We assemble those pieces, then send it to Claude to add two or three sensory details that make the image feel specific and intentional.
0+
Vocab entries
0s
Average time
0+
AI models
Watch it happen
Here is what actually happens when you use FrameBrief
The demo on the right is running a real example in real time. You can see how the questions adapt to the subject you typed, and how the final prompt reads nothing like what you would have written yourself.
You describe your subject in plain language
Just write what you are making. No special format required. Something like "wireless earbuds for an Instagram lifestyle post" is completely enough. The AI reads that and figures out what kind of creative project this is.
The AI asks you questions that actually make sense for that subject
If you typed a product shot, it asks about surface material and lighting. If you typed a logo, it asks about typography and brand personality. It never asks about camera angle for a logo or typography for a product photo. Every question is chosen based on what you typed and every answer before it.
You get a prompt written in the right syntax for each model you picked
Midjourney gets keywords with parameter flags. Flux gets a paragraph of natural prose. DALL-E gets a concise description. Same creative intent, three completely different formats. You copy whichever one you need.
Supported Models
One brief, written correctly for every model you use
Each of these models has real differences in how prompts should be written. We handle all of them so you do not have to learn every one separately.
Midjourney
Image modelMidjourney works best with short, precise keywords strung together with commas, followed by parameter flags like --ar for aspect ratio and --s for how stylised you want it to look. We write exactly that format, with the right flags for whatever platform you picked.
Flux
Image modelFlux rewards longer prompts written like a paragraph of prose. We write 80 to 150 words of natural, descriptive language that Flux actually responds well to. Short keyword prompts tend to produce mediocre results in Flux.
DALL-E / GPT Image
Image modelDALL-E runs your prompt through its own rewriting system before it generates anything. So over-specifying actually works against you. We keep it to 20 to 80 words of clear, intentional description and let DALL-E fill in the rest.
Stable Diffusion
Image modelStable Diffusion uses weighted comma-separated tags and always needs both a positive prompt and a negative prompt. We handle both, including the quality tags you need at the start and the negatives that prevent the most common artifacts.
Ideogram
Image modelIdeogram is the best model available right now for anything with text in the image, like logos, posters, or social graphics. We format prompts to take advantage of its text rendering, which the other models handle poorly.
Runway Gen-4
Video modelVideo prompt writing is different from image prompt writing. For Runway, we lead with camera movement because that is what the model pays the most attention to. Dolly, orbit, crane, handheld. Then the subject and environment.
What is included
The details that make it actually work in practice
These are the things you notice after using FrameBrief for a few weeks, when you realise how much thought went into the pieces that seem small.
Questions that change based on what you are making
We do not show you the same questions every time. The AI reads your subject and picks what to ask. A logo gets questions about typeface and brand personality. A product shot gets questions about surface texture and lighting angle. A portrait gets questions about pose and expression. The questions are always relevant to your specific project.
Aspect ratio handled automatically for each platform
When you pick Instagram Feed, every model gets the right aspect ratio in its own format. Midjourney gets --ar 1:1. Flux gets a note in the prose about square framing. DALL-E gets the 1024x1024 pixel dimension. You pick the platform once, and it flows into everything.
Brand colours and fonts coming in the next version
You will be able to save your brand colours as hex codes, note your preferred fonts, and describe your visual voice. FrameBrief will then pull those into every prompt, translating hex codes to model-friendly colour descriptions where the model does not support them directly.
The copy button works on every device and browser
It sounds obvious, but a lot of tools get this wrong on mobile. We use the modern Clipboard API where it is available, with a textarea-based fallback for older browsers and iOS Safari. The copy button is the most important button in FrameBrief, so we made sure it never fails.
You still get a prompt even if the AI connection fails
The prompt assembly engine runs in two stages. The first stage is purely deterministic and does not need a network connection. It assembles your answers into a template using our vocabulary library. The AI adds finishing touches in the second stage. If the AI is unavailable, the first stage output is still a solid, usable prompt.
Every prompt you generate is saved to your account
Your full history stays in your dashboard with the subject you typed, the models you targeted, and the complete prompts. You can go back, re-copy something from six months ago, or use a past brief as a starting point for a new one.
How the questioning works
The questions you see depend entirely on what you typed in the first step.
Take two very different subjects: a logo for a coffee shop, and a product photo of wireless earbuds. Those two projects need completely different information to produce a good image. A logo needs to know about typeface personality, colour symbolism, and whether it should feel vintage or modern. A product photo needs to know about the surface it sits on, the quality of light, and the depth of field.
FrameBrief reads your subject before deciding what to ask. And then each answer you give shapes what the next question will be. If you say you want a warm, artisanal feeling, the next question focuses on texture and material. If you say you want something bold and modern, it goes in a completely different direction. You can see this playing out in the animation on the right.
See it for yourselfPricing
Start with the free plan and see if it changes how you work
Every new account gets 14 days of the Creator plan automatically, so you can see what it is like with full access before choosing anything.
Free
A good way to see whether FrameBrief actually improves your output before you commit to anything.
- ✓5 prompts per month
- ✓Midjourney, Flux, and DALL-E
- ✓Full prompt history
- ✓Watermark text on each prompt
Creator
For people who use AI images regularly and need prompts that are consistently good across multiple models.
- ✓Unlimited prompts every month
- ✓All 10 models including video
- ✓10 saved brand presets
- ✓No watermark on prompts
- ✓Faster generation priority
Pro
For small teams who share a brand identity and need everyone generating consistent, on-brand prompts.
- ✓Everything in Creator
- ✓3 team member seats
- ✓Side-by-side prompt comparison
- ✓Curated prompt template library
- ✓Early access to new models
Your first five prompts are completely free, and you do not need to add a credit card to try it.
Sign in with Google, describe your first subject, answer a few questions, and see what comes out the other side. If the prompt is better than what you would have written yourself, you will know immediately.
Get started, it is freeFrameBrief is built for marketers, brand designers, social media managers, and content creators