This overview presents the most important generative AIs for text and image creation. This includes ChatGPT, Google Bard, Perplexity, Cohere, Claude, Midjourney and many others.
Summary / TL;DR
- For text AIs, ChatGPT is the most used solution worldwide. Google Bard is slowly catching up. Also very relevant: Bing Chat. Other models aimed at companies, such as Cohere, Claude 2 (Anthropic), Luminous (Aleph Alpha) are much less sought after worldwide.
- For image AIs Midjourney is the clear leader. The strongest competitors worldwide are Adobe Firefly and Stable Diffusion.
Part A: The best AIs for text generation
ChatGPT
ChatGPT from OpenAI is the most powerful and widely used text AI. Shortly after its launch in 11/2022, it became very successful and is used by millions of people for personal and professional tasks. The Plus subscription provides access to GPT-4, over 100 useful plugins that enable practical additional functions. In addition, there is Code Interpreter, which allows you to upload your own files, run code, visualize data and export and download your results in the desired format.
- Provider: OpenAI
- Target group: All
- Special Features: Best quality results, many plugins available, API available, powerful code interpreter, internet access available
- Costs: GPT-3.5 (free to use), GPT-4 (only for Plus customers, about 20€/month)
- Website: https://openai.com/chatgpt
- App: iOS, Android
Bing Chat / Microsoft Bing
Microsoft offers “Bing Chat”, a chatbot similar to the popular ChatGPT. Here you can even use the powerful GPT-4 model for free. With Bing Chat you can create texts, answer questions and even create images. Microsoft recommends using the in-house browser “Microsoft Edge” (Windows, macOS) or installing the Bing app (Android, iOS). However, Bing Chat now also runs directly in Chrome and soon in Safari.
- Provider: Microsoft
- Target audience: everyone
- Special Features: Best quality results, Internet access available, easily integrated into browser
- Costs: GPT-4 (free to use)
- Website: https://chat.bing.com
- Mobile: Bing Download: iOS, Android
Microsoft Copilot
Microsoft offers its leading chatbot “Copilot” in various versions, so that all users can use it appropriately. Private users can use it for free, but they only get a simple version (without integration into Office tools). Corporate and private customers pay monthly, but in return, they get direct integration into Word, Excel, PowerPoint. This integration significantly increases productivity. Moreover, Microsoft Copilot is based on OpenAI’s latest GPT language model, making it superior in response quality to many other chatbots.
- Provider: Microsoft
- Target Audience: General Public & Businesses
- Special Features: Best result quality, chatbot can access the internet, best integration with Office tools (Word, Excel, PowerPoint, Microsoft Teams, etc.)
- Cost:
- Copilot: Free (for Everyone)
- Copilot Pro: $20/Month (for Private Customers)
- Copilot for Microsoft 365: $30/Month (for Corporate Customers)
- Website:
Google Bard
Google’s text AI Bard launched in late 2023. It is based on Google’s AI model “PaLM 2” and will be increasingly integrated into Google Search in 2024, reaching significantly more users than any other provider. Google Bard can create texts of all kinds, generate code in over 20 programming languages (e.g. Python, Java, Javascript, Typescript, C , Go), analyze data and much more.
The results of analyses can be output directly to a Google Sheet or Google Doc. Data analysts can even import generated Python code into Google Colab with a single click. Interesting for those who need search engine optimization for their websites: Google Bard also provides information on keyword research and approximate search volume (prompt: “Perform a keyword research on the topic ___. Add information about search volume and search trends”).
- Provider: Google
- Target group: All
- Specifics: Sufficiently good text generation (worse than ChatGPT), good coding skills, current Internet data usable, Integrated with other Google products
- Website: https://bard.google.com/
Perplexity
Perplexity is a combination of chatbot and web search. The quality and speed are convincing. It even cites sources from the web, so hallucinations can be avoided. The base of the Free Version is OpenAI’s GPT-3.5 model, combined with the company’s independent LLM, which includes natural language processing (NLP) capabilities. Perplexity Pro has access to GPT-4, Claude 2, and other models available on the market. Perplexity was founded in August 2022 and is led by CEO Aravind Srinivas, who previously worked at OpenAI.
- Provider: Perplexity
- Target Audience: Everyone & Corporate Clients
- Special Features: Brilliant combination of chatbot and search engine (Google search, GPT), API available, fast and good results
- Cost:
- Free Version: Free
- Pro Version: $20/Month
- Website: https://perplexity.ai
Cohere Command
The generative language model “Command” from Cohere addresses the needs of companies. Cohere therefore offers a helpful customer service, so that companies can integrate the AI models as seamlessly and easily as possible into their own applications in the company. The new AI assistant “Coral” uses Command and can connect to various data sources, is trained on company-specific documents and thus gives very appropriate results with a low error rate (hardly any hallucinations). In addition, it is always traceable where the result came from (e.g. “Source: Annual Report 2023, page 8”). Good for data protection and corporate security: Cohere can be installed on in-house servers or clouds. Data is then not sent to Cohere and remains DSGVO-compliant in the company’s own country.
- Provider: Cohere
- Target group: Enterprises
- References: Spotify, Oracle, others
- Special features: Learn your own data, privacy compliant, company knowledge is not shared with outside companies, API available
- URL: https://cohere.com/models/command
Luminous
Luminous is the large language model (LLM) of the German provider “Aleph Alpha” and is aimed at companies and organizations. Luminous can be set up and operated on its own infrastructure and is ready for use to analyze and process text and image data. Aleph Alpha offer their product with a clear focus on “Made in Europe” to ensure data protection and other standards for their corporate customers. Due to the proximity to the university, scientific findings flow into the further development. Other advantages according to the provider are scalability, secure and data protection-compliant hosting in Germany and ease of use.
- Provider: Aleph Alpha GmbH (Heidelberg)
- Target group: Companies
- Special features: EU-compliant data protection and other EU standards, API available
- Customers: City of Heidelberg, BWI GmbH, TU Darmstadt
- URL: https://www.aleph-alpha.com/luminous
LLaMA 2
Meta has released LLaMA 2, an open source language model that has received wide attention.
- Provider: Meta (formerly Facebook)
- Target group: Companies
- Special Features: Open source language model
- URL: https://ai.meta.com/llama/
Claude 2
Anthropic’s generative language model Claude 2 is not expected in Europe until late 2023.
- Provider: Anthropic
- Target group: companies
- Special features: to follow
- URL: https://www.anthropic.com/index/claude-2
Part B: The best AIs for image creation
Midjourney
Midjourney generates images with artistic qualities, by using , You can request diverse image styles, such as photorealism, anime, comic, art eras, specific artists (e.g. “Bart Simpson in the style of Picasso”), oil paintings and much more. Getting started is a bit complicated, as you need a Discord account to log in. After logging in, you can see a confusing number of other users’ chats and learn from their prompts. However, you can make your pictures private so no one else can see them. The learning curve is worth it: because Midjourney is considered the best image AI by many professional users.
- Provider: Midjourney (a U.S. research institute on computer art)
- Target group: All
- Special features: Best quality results, most used image AI in the world
- Cost: from $8 / month for about 200 generated images (“3.3 hours with near-CPU usage”)
- URL: https://www.midjourney.com/
Stable Diffusion
Stability AI’s sophisticated “Stable Diffusion XL” delivers very good results that can match and in some cases exceed Midjourney. Benefits include: Compelling photorealistic image generation, rich image styles with appealing aesthetics. Realistic face generation. You can give prompts for image composition (“Show a vase of roses on a mahogany table. Now move the vase to the left”). Even short prompts are enough to produce good results. You can also have readable text generated on the images.
- Provider: Stability AI
- Target group: All
- Special features: Open source model, very good results
- Costs: Free (few images per month), about 8 $ / month for 999 images (via provider StableDiffusionAPI)
- URL: https://stability.ai/stablediffusion
Our cover image was generated with Stable Diffusion XL. The prompt: “In Cyberspace a beautiful AI woman is creating texts and images, neon aesthetics”. For this we chose a graphical image style:
DALL-E 2
DALL-E 2 is the image AI from OpenAI – the company behind ChatGPT. It delivers amazing results, but sometimes makes mistakes (e.g. 6 finger problem or unclean renderings especially of faces). Even simple tools are included: you can mark areas in the image and have them replaced by prompt (e.g. “insert cola bottle here”)
- Provider: OpenAI
- Target group: All
- Costs: 15 credits for free per month (= approx. 15 images), you can buy credit packages (= approx. 10 cents per image)
- Special features: Very easy to use, simple tools for image manipulation
- URL: https://openai.com/dall-e-2
Adobe Firefly
The generative AI “Adobe Firefly” solves design tasks in images, videos and 3D. It is integrated with Adobe’s tools (e.g. Photoshop, Adobe Express and others). Firefly has a novel approach that is especially helpful to designers and image creators. This is because the works created with it are generated exclusively from opensource images and the gigantic Adobe photo pool. The origin of the image material can be traced at any time. This is how copyrights are supposed to be protected. Adobe’s big advantage is the integration of the AI into the sophisticated professional tools of the Creative Cloud for designers, motion designers, layouters and more. This allows you to iteratively change the work you’re designing and stay in one tool instead of having to switch. This will put other image AIs to shame.
- Vendor: Adobe
- Target group: Companies, freelancers
- Cost: no figure available yet
- Special features: Professional editing of image/3D/video possible via prompts, copyright protection or AI reuse by artists possible (“Do not use for training” tag)
- URL: https://www.adobe.com/de/sensei/generative-ai/firefly.html
Example: Select area, enter prompt, click Generate
The result: the sign has been inserted to match the lighting, and the arrow even points correctly to the door. However, the shadow is missing, but can be added by a follow-up prompt.
You can use a command to transfer an image to a completely new atmosphere and add to it (“Show the landscape at night. Add snow. Add a family in winter clothes.”). Firefly can also handle 3D, so even non-professionals can edit 3D models without having to learn a complex 3D tool. You can even change the weather and tone, adjust the density of rain, fog and snow by simply describing it via prompt.
Adobe’s impressive Firefly product videos show more:
More overviews of text and image AIs
Here you can find more comparisons of the most important LLMs.
- 12 Best Large Language Models (LLMs) in 2023 – good overview with technical info and benchmark results
- 14 Best AI Image Generators in 2023 – comparison of many image AIs and tools (e.g. Nightcafe, which allows you to use multiple image AIs comfortably)
