Products

From Adobe Firefly to Imagen 3: a short guide to generating images with text

Using the best text-to-image through easy, fast interfaces. And even free. Or pay for more? .

27 October 2024

3' min read

Use the best image generators, with artificial intelligence, through easy, fast interfaces. And even for free. Or pay for something more: for more professional use for example.

These are the possibilities offered by current AI image generators. A panorama recently enriched by the arrival of Google's powerful Imagen 3 model within Gemini, which can also be used for free.

There are many alternatives and almost all of them allow you to do at least something without paying. It is therefore possible to try out different services to see which one suits us best, for that specific image or activity.

Imagen 3 by Google - Gemini

Imagen, now in its third version, was already available in ImageFx, also by Google, but not to the Italian public. It is therefore, above all, good news for us that it has landed on the Gemini and Gemini Advanced chatbots. The former is free, the latter is by subscription, EUR 21.99 per month. Only the Advanced generates images with people's faces. An interesting advantage, since it is not possible to do the same with its main rival: Dall-E3 by OpenAI. Recall that Google had put on hold the possibility of creating images with Gemini after a scandal: it generated faces of black Nazis and other nonsense, daughters of gender and ethnic bias.

The latest tests say that Imagen 3 is superior to rivals in image quality, detail; in photorealism of people, animals, scenes and the ability to interpret different artistic styles.

Dall-E3 - Microsoft Image Creator and Copilot

Open AI's Dall-E3 is among the first and most popular services of its kind, but it only allows two generations a day for free (less than rivals). Unlimited, it requires a Chatgpt Plus subscription of $20 per month. The same model is accessible for free, however, via Microsoft's Image Creator, which is also integrated in Copilot. Dall-E3 is distinguished by a less photorealistic style, but sometimes succeeds better in interpreting the user's wishes expressed in the prompt.

Gemini, Copilot and Chatgpt have this in common: they are chatbots that allow the use of image creation within the same text generation interface.

Midjourney

For more special purposes, however, it is good to delve into alternative services. Such as Midjourney, capable of creating amazing digital artwork and images of people, objects, scenery of the highest quality. In the past, it was penalised by being not very user-friendly (only via Discord), whereas it has now been made accessible via the site as well. Main negative note: it is only pay-as-you-go, $10 per month.

Stable Diffusion

Stability Ai's Stable Diffusion model is also more specialised and open source. The tool can be downloaded for free, but requires specialised skills. However, it can also be accessed via the simplified Dreamstudio interface of Stability AI itself. It is characterised by the high possibility of customising images by selecting 'negative prompts' (where we indicate what we do not want to see). After the free trial it costs one dollar for 100 credits, with which you can make 50 images. Stable Diffusion is also accessible via various third-party sites, free of charge but with long queues as usual, and it does not have many of the typical pitfalls found on other services (violence, sex).

Adobe Firefly and Getty Images

Two tools appreciated by image professionals (photographers, creatives, marketers) are Adobe's Firefly and Getty Images' Generative AI. Both ensure copyright compliance and protect the creators of the images used to train models.

Firefly is free and is only a few days into version 3.0. One of its most interesting features is the new style engine, with a wide range of styles, colours, backgrounds and poses. Also particular is the possibility of generating images that mirror characteristics of other images, used as a reference for structure and style. The user therefore has more control over the end result. For example, he can feed a quick sketch or a company logo to a more accomplished image that reflects them. Adobe Firefly has been trained on Adobe Stock images, openly licensed content and public domain content. Similarly, Getty Images takes advantage of its extensive library. These are therefore two services favoured by companies that want to avoid any possible copyright issues. Getty Images, however, charges a fee, starting at 45 euros for 25 generations.