TL:DR
Using a random spread of AI Image generators, evaluate the sign-up process, from the offer versus reality and eventual output. Demonstrate quality degrees using the same prompts, contrast against amendment flexibility, what each is best at and credit cost (real value).
“Creative interpretation of the human mind”
Elements indicating continuous learning, course listings and interactive learning tools, and seamless student and faculty portals.
⸺ Image prompt ⸺
Visualising without Frontiers
Conceptualising
Specific Aesthetic
The core benefit is generating a high volume of diverse visual concepts quickly. Designers can use the AI platforms to visualise in seconds, saving hours of manual sketching or searching for stock.
Rapid Concepting
The principle benefit is generating a high volume of diverse visual concepts quickly. Designers can use the AI platforms to visualise initial ideas, mood boards, or multiple design variations in seconds, saving hours of manual sketching or searching for stock imagery.
Style and Aesthetic Discovery
By using the same prompt across 10 different platforms, designers can discover the unique style bias of each AI model. This helps them learn which platform is best suited for a abstract art, for hyper-realism or digital landscapes.
Abstract Art Example Prompt is specifically designed for this.
Pushing Creative Boundaries
Use highly abstract and metaphorical prompts;
- “Learner evolving and expanding their skills“
- “Linking neurons and grids“
These encourage the AI to interpret ideas creatively, often generating visuals a human designer might not have initially considered. This can breakdown creative blocks and lead to novel solutions.
Technical Proficiency
prompt skills
adherence
workflow
This experiment directly tests and refines prompt engineering skills. Designers learn what specific keywords, style descriptors, and technical instructions (i.e aspect ratio, specific themes like 3D flavours and mobile payment) yield the best results on different models.
Platform Benchmarking
Image Quality
Which free platform offers the sharpest, most professional-looking output?
Prompt Adherence
For complex instructions
The two prompts requiring a complex sign object with no words and a transparent background are perfect for testing adherence to specific technical requirements
A budget prioritise tools at different project stages.
Credit/Cost Efficiency
For free tiers, understanding the credit limits;
- Canva’s 30-day limit
- NightCafe’s 20 max credits
- OpenArt’s 40 max credits
Workflow Integration
Identifying which platform is the fastest and most reliable for specific tasks (like generating icons, textures, or full scenes) allows a designer to streamline their workflow, using AI tools for repetitive tasks and dedicating more time to high-value creative work.
Prompt Mastery
The experiment directly tests and refines prompt engineering skills. Designers learn what specific keywords, style descriptors, and technical instructions (like aspect ratio: 9:16, or specific themes like 3D flavours and mobile payment) yield the best results on different models.
Testing the AI
prompt skills
adherence
workflow
Testing for the practical. Each use case involved some specific design asset generation. The principle focus was:
- Backgrounds – Using the 9:16 aspect ratio for backgrounds.
- Icons/Objects – Generating complex 3D icons for app or web design with transparent backgrounds
- Branding/Mood – Creating a visual aesthetic for a lab/studio environment using specific color and style constraints.
⸺ Human Prompt #1 ⸺
“Creative interpretation of the human mind, elements indicating continuous learning, course listings and interactive learning tools, and seamless student and faculty portals”
Microsoft Designer / Bing Image Creator
Powered by DALL-E 3
Microsoft tools are generally free to use with an account and provide high-quality image generation directly from text prompts
AI examples
Variation and offering suggestions (often within the prompts) allows the designer to evaluate and compare
Adobe Firefly
AI examples
Creative Interpretation: Very symbolic-led, although clean and communicative, expetataed AI visuals already have 3D element – which these visuals are lacking.
Ethically sourced training data
Adobe’s AI tool offers free access and monthly generative credits for creating images and text effects from text.
It’s known for its focus on ethically sourced training data.
Leonardo AI
Token model
This platform is known for its generous free plan, offering a significant number of tokens (credits) daily to generate images.
It also provides various models and fine-tuning options.
AI examples
Technical Adherence: The extreme detail adds value at every level.
Craiyon
AI examples
Aesthetic Quality: Again pushing the boundaries. Very foundational art.
Token model
Formerly DALL-E mini
A completely free and easy-to-use AI image generator. While the quality might not always match the top-tier paid options, it’s excellent for experimentation and quickly generating a wide variety of images.
DreamStudio
Token model
Developed by Stability AI
Using the Stable Diffusion model, this text-to-image diffusion model is targeted for everyone, from hobbyists to professionals. Typically adopting the credit-based system the complexity/size of the image determines the credit cost per generation.
AI examples
Aesthetic Quality: Again pushing the boundaries. Very foundational art.
⸺ Human Prompt #2 ⸺
“Slight emphasis on the colour orange, art imagery and experimental typography, gradient overlays, art installations, and subtle animations to create a cutting-edge aesthetic that reflects a lab studio’s innovative approach to digital design.”
Freepik AI Image Generator
Token model
Freepik provides a free tier with a certain number of credits for AI image generation, along with a vast library of other design assets.
It also provides various models and fine-tuning options.
AI examples
Technical Adherence: The extreme detail adds value at every level.
ImageFX (Google)
AI examples
Aesthetic Quality: AI cannot cope with text. Therefore, eliminate any word preferences, (unless your purposely looking for an alien’ language aesthetic)
Token model
Google’s standalone AI image generator, which uses Imagen 3.
It’s free to use with a Google account and offers “expressive chips” to easily modify prompts. You can also access Imagen 3 through Google Gemini.
⸺ Human Prompt #3 ⸺
“Represent a creative lab environment for experimenting with a gallery of images, with data visualisation, AI tools and immersive typography through to animated brand identities, tinted to pop art orange and candy red.”
NightCafe
Powered by DALL-E
This platform offers a robust free tier with daily credits for generating AI art. It supports various AI models (including Stable Diffusion and DALL-E 3) and has a strong community aspect with challenges and sharing.
20 max credits
AI examples
Variation and offering suggestions (often within the prompts) allows the designer to evaluate and compare
OpenArt
AI examples
Variation and offering suggestions (often within the prompts) allows the designer to evaluate and compare
Powered by DALL-E
OpenArt provides a free plan with daily credits and access to various AI models, including Stable Diffusion XL.
It also offers advanced features like image-to-video conversion and character consistency.
40 max credits
Comparison matrix
Over a set period with my designer head on, I logged my results. Initially focusing on what the tool is best for, other dependents became important criteria (like cost and flexibility).
| – | Model/Engine | Aesthetic Quality (Image Quality) | Creative Interpretation (Prompt Adherence) | UX/Efficiency & Cost | Workflow Integrations (Specific Tasks) | Overall Best For |
|---|---|---|---|---|---|---|
| Microsoft Designer | DALL-E 3 | (Rating 1-5) & Comments | Adherence: (Y/N) on “No Words” & “Transparent BG” | (Rating 1-5) & Style Notes | Speed/Credit cost (e.g., Fast/Free) | Depth of 3D believe-ability. This AI represented every word from my prompt – it also added that something extra. |
| Adobe Firefly | Firefly | Speed/Credit cost (e.g., Medium/Generous Free) | ||||
| Canva | (Magic Media/DREAM) | Speed/Credit cost (e.g., Fast/30-Day Limit) | ||||
| Leonardo AI | (Specify model used) | Speed/Credit cost (e.g., Medium/Generous Daily Tokens) | Very detailed which contributed to the believe-ability factor. Also succeeded in spiking my imagination, and in-turn fixed my creative vision. So a double-win | |||
| Craiyon | (DALL-E mini variant) | Speed/Credit cost (e.g., Slow/Unlimited Free) | Now I was iterating on the iterated prompts, and understanding the ‘…more you give’ concept. However, despite lacking a glossy 3D checklist layer, the tool was successful in meeting the deeper, conceptual, or foundational need of kick-starting your initial brainstorming mode. | |||
| DreamStudio | Stable Diffusion | Best of the bunch. In depth and appropriately style creative mind visuals, with variations and within the specified colour space. Enough crest flexibility to iterate, to amend and achieve a desirable result. My Top Pick. | ||||
| Freepik | (Internal/Stable Diffusion) | Speed/Credit cost (e.g., Medium/Limited Free) | Creative Interpretations: Although variation on ta them, the initiation first stage output did pump those creative juices. Realistic and sometimes) very abstract representations of prompt. A close second place. | |||
| ImageFX (Google) | Imagen 3 | Speed/Credit cost (e.g., Fast/Free) | Similar to Freepik, this engine also competed in the Top 3. Followed direction, with a familiar UI, provided flexibility to iterate with enough credits and delivered an appropriate level of creativity. | |||
| NightCafe | Pollo | Speed/Credit cost (e.g., Slow/20 Max Credits) | Boss of Bosses: The secret is to employ an AI that employs an AI and NightCafe has that to perfection. The big takeaway for this engine is the credits enable a simple experiment with image-to-video conversion. Regarded as Premium functionality, this elevates their offer to my Top 3. | |||
| OpenArt | (Specify model used) | Speed/Credit cost (e.g., Medium/40 Max Credits) | This engine needs a differentiator. Similar to Freepik and NightCafe, with a familiar interface as an end user I have to ask what is the benefit FOR ME. Currently, I don’t see it. | |||
| Platform 10 | (Name) | Speed/Credit cost |
Conclusion
Image Quality
Prompt Adherence
Diversity
For grading or evaluating AI image generators I used Image Quality, Prompt Adherence, and Diversity.
Quantitative metrics
Image Quality
Aesthetic and technical excellence
Realism/Fidelity
3/10
Detail and Clarity
6/10
Artifacts
2/10
Prompt Adherence
Text-to-Image Alignment
Content Accuracy
3/10
Attribute Accuracy
6/10
Composition
2/10
Diversity
Aesthetic and technical excellence
Mode Coverage
3/10
A full spectrum of visual concepts the AI was trained on, avoiding repetition.
Stylistic Range
6/10
Novelty
2/10