Location

Online seminar. (Directly from your computer or smartphone)

Price

Free

Introduction by Jean-Luc Lasquellec Wallcrypt

Here you go! Jean-Luc Lasquellec thanks speaker Samuel Dumas from ACADEE for speaking at this Master Class on Image Generative AI.

He invites Samuel to introduce himself and his training center

He also thanks all the people present who are participating in this webinar who are members of the Wallcrypt Club, the Web3 and AI business club whose objective is to democratize these technologies and implement business relationship.

We also talk about the lottery that Wallcrypt has set up and the prizes that Samuel is offering to the winners.

Subject: Image Generative AI

SUMMARY:

1. 1. Generative AI (GAI) as a source of inspiration

   2. Campaign examples made by image generative AI

   3. Learn the art of the prompt + GIFT

Speaker starts taking speech (01:43)

Samuel Dumas:

Samuel begins by explaining that he is here to talk about creativity augmented by artificial intelligence, with an emphasis on concrete examples and practical applications. He presents his center of training, Acadee, which specializes in web 3 and offers training in blockchain development He also mentions the IA Web 3 to 3 fair that they successfully organized, as well as the training they offer on software like. Midjourney for creating images, suitable for communication professionals and individuals.

Next, he addresses the subject of image-generative AI and highlights its potential as a source of inspiration, but also the importance of intervention human in the creative process He promises to show concrete examples of communication campaigns using Lia, as well as training techniques that they offer at Acadee He also announces a gift for members of the Wallcrypt community at the end of the. presentation.

Example Comparison of real and Midjourney-created advertising images (5:52 – 9:42)

The image on the right seems to be more appealing to the majority of people, although it is generated by MidJourney (Image Generative AI), a tool with a monthly subscription of €10, tandispute that the image on the left is an official advertising campaign which would have cost around €500,000. This demonstration highlights the efficiency and speed of image generation with tools like Midjourney, while highlighting the legal and ethical issues linked to the use of such technologies. It also shows how new technologies can compete with traditional methods of marketing and advertising creation.

According to the text, the majority of participants seem to prefer the image on the right when purchasing the perfume. However, it is emphasized that the image on the right is a creation made in Midjourney (Image Generative AI) with a monthly subscription of €10, while the image on the left is an official advertising campaign that would have cost around €500,000. Despite this, the right-wing image appears to rival professional campaigns and raises questions about the ethical and legal use of such image-generating technologies.

KREA SKETCH TO IMAGE (9:42 – 10:36)
The tool discussed in the text is KREA. It is a tool that allows users to draw with the mouse or use a video stream to create images. In the example given, the user quickly draws a car with a palm tree and adds text describing the scene The tool Image Generative AI then generates an image matching the description. This allows images to be quickly created without the need to make a photo shoot This is an example of the use of generative tools for creating images
.

MIDJOURNEY: LA ROLLS DU TEXT TO IMAGE (10:37 – 11:40)

In this passage, the speaker discusses Midjournée, a tool he describes as “Rolls from text to image”. He also mentions other similar tools like Leonardo and Dali 3. He explains that he created two fake advertisements to illustrate Midjournée’s capabilities, one for North Face and one for Marineland, adding a little Photoshop for a more advertising look.D. He points out that the images generated by the Image Generative AI can be used on a large scale with sufficient resolution for printing on large formats. Next, he discusses using Midjournée for logo creation, where users can use prompts to generate a series of logos. This shows the versatility of the tool in different creative contexts.

LOGO CREATION (11:41)

The speaker explains that Midjourney Image Generative AI can be used to create a series of different logos using prompts. He gives the example of a request to create a logo for a store of toys He begins by creating a small character representing a fox, described as smart, clever and smiling. Then, he uses this graphic style and this fox to create a logo corresponding to the request. He emphasizes that the results are usable and professional. , thus demonstrating Midjourney’s ability to generate consistent and quality logos.

PRODUCT DESIGN (12:16 – 14:19)

In this passage, the speaker talks about the use of generative tools (Image Generative AI) as a source of inspiration for product design. He shows some examples, such as a jellyfish lamp generated by Alexis, a trainer, and a bottle of perfume generated on the right He also mentions the possibility of imaginary collaborations, like between Ikea and Patagonia on an armchair He compares this approach to that of a Chanel designer who passes his sketches to the little hands who make the dresses. thus highlighting the potential of generative tools to stimulate creativity and serve as a basis for the actual making of objects. In response to a question about the use of Kenzo images in Midjourney, he explains that the algorithm behind Midjourney has been trained. on millions of images, but it also raises legal questions about the use of such images. He concludes by saying that generative tools (image generative AI) offer a great opportunity for inspiration for designers and the manufacturing of real objects.

In this passage, the speaker describes a real-life application of Midjourney. He explains that on Discord, they use prompts to generate dress styles, then pass it off to stylists to make them. Then, the dresses are presented in an official fashion show with real models wearing the creations designed by Midjourney. He then answered a question about Midjourney’s ability to produce images of famous people. He explains that Midjourney has no restrictions in this regard and can create images in response to requests such as “Macron at the beach”. However, it highlights ethical issues related to depicting people with images generated by theImage Generative AI and advocates for the introduction of a Lia-generated label to distinguish generated images from real images. This would allow consumers to know if an image is fake or not.

3. VIRTUAL INFLUENCE (15:54 – 16:55)

The speaker raises the topic of virtual influence, highlighting cases such as Aitana Lopez, a fictional muse with pink hair who has hundreds of thousands of followers on LinkedIn and runs lucrative campaigns. He also mentions the example of Lil Miquela, another virtual figure with 2 million followers on Instagram. He talks about the influencer project, where fictitious models are created for fashion shows and advertising campaigns. In the upper left corner, it shows fictional photorealistic characters created for luxury products. He points out Midjourney’s limitations, including that it generates variations every time, requiring you to juggle different options to achieve the desired result. Finally, he mentions the use of the image generative AI Midjourney to produce packshots, thus making it possible to test different product packaging.

4. PACKSHOT (16:58 – 17:35)

The author highlights the importance of pack shots for testing products and presenting them in an attractive way. They also mention the Maker app as a tool for generating high-quality renders in different atmospheres,which is particularly useful for e-commerce and the cosmetics industry. Finally, the author discusses the possibility of working on pack shots to obtain professional results, emphasizing their importance in the field of cosmetics.

5. ICONS (17:35 – 18:09)

The speaker highlights the importance of pack shots as an essential element in producing icon galleries for various applications such as video games and the web. They also mention appreciation for Pierre Chevalier’s work in image generation, highlighting the possibility of creating video games from these elements. They also reference a ChatGPT-generated version of Angry Birds for Halloween, highlighting the technology’s potential to facilitate video game creation.

6. WEB DESIGN (18:10 – 18:38)

The speaker shares his experience in the field of web design, explaining how it is possible to ask an artificial intelligence like ChatGPT to create website structures according to specific objectives. They mention the idea of “hallucinations”, perhaps referring to creative suggestions or ideas that image-generative AI can generate. They also describe how it can help visualize the structure of a website, even if the images are not perfect, giving an idea of the size and layout of the site.

7. STORYBOARD (18:39 – 19:28)

It seems that the speaker is discussing the current trend of big influencers moving towards storyboarding, a common practice in the field of audiovisual and cinema production. Storyboarding involves creating visual plans to prepare for filming, whether for photo shoots or films. With the advent of version 6 of a specific tool (probably software or application), it is possible to experiment with different angles, lights, lenses and focuses to achieve the desired result. Although this versionmay not yet be perfect, it offers precise management of the visual parameters necessary to create effective boards for filming.

8.CHARACTER DESIGN (19:29 – 19:56)
The speaker touches on another interesting topic: character design, a crucial area in visual creation. They mention having launched a GPT (image generative AI) specifically designed for this task This means that you can configure this tool to generate characters for various uses, whether for cartoons, illustrations or other forms of visual communication. This offers a practical solution for creating characters suitable for different projects. having to design them manually, which can be time-consuming and require specific design skills.

9. SETTINGS (19:57 – 20:16)

The speaker emphasizes the importance of being able to reuse generated characters in different positions and contexts. These characters can be integrated into various settings, whether film sets, filming, video games or even promotional brochures, such as that of a Resort in Bali, even if such an establishment does not actually exist. These fictional settings offer creative possibilities for developing scenes, brochures and other materials. visuals, allowing you to explore different ideas and test concepts before implementing them in larger projects.

10. STYLING (20:17 – 21:04)

The speaker further expands the scope of the characters and settings generated by the Image generative AI, considering them in fields such as architecture, fashion design, and even science fiction or special effects For example, characters and settings can serve as the basis for architectural projects, styling concepts for advertising campaigns or fashion clothing.re used in cinematic contexts, whether to create film posters, costumes for science fiction productions or fantasy environments in the style of “Lord of the Rings”. The speaker highlights that everything that has been presented so far has been generated in seconds by the tool they are using, demonstrating the speed and versatility of this approach in visual creation.

11. MOVIE POSTER (9:05 p.m. – 10:05 p.m.)

The speaker shares a variety of projects they have worked on, including in collaboration with the Ministry of the Armed Forces. They wrote a scenario called “Operation Cyber Shadow”, which involves a hack of the Department of the Armed Forces headquarters in Ballard Balard. Additionally, they had the opportunity to create posters for this project, drawing inspiration from the style of Christopher Nolan. They also mention the creation of characters for NFTs used as tickets to a web show, offering different benefits depending on the level of ownership.

Apart from these projects, the speaker wrote a new adventure for his son entitled “Le Prince au pied qui pu”, illustrated by an image that could serve as the basis for a children’s film. They emphasize the importance of mood boards in their communication work, thus illustrating the diversity of visual supports they use in their projects.

It seems that the speaker also works for Palmolive and helps create mood boards for fragrances, perhaps drawing inspiration of the work done by Alexis for watchmaking, where different tones of colors and planes are used to create a specific atmosphere. Additionally, they mention creating moodboards for sportswear, which allows them to keep up with current trends in this area. Mood boards are essential tools in visual creation, bringing together inspiration, colors and textures to guide the design process.

13. SOCIAL NETWORKSAUX (22:30-23:00)

The speaker also discusses the importance of visual creation on social networks. They mention having produced an image generated by the image-generative AI Midjourney for their company’s wishes for the year 2024, with the message “Let’s connect our intelligences”. The idea for this image came from their own imagination, but to make it a reality, they thought of two people taking a selfie with a robot. This image was used for communication on social networks to present the company’s wishes for the current year. This shows how creativity can be used to convey messages on digital platforms and engage audiences.

What technical difference does it have? between Midjourney and Dallee? (23:01- 24:25)
It appears that “Midjourney” is a tool or platform used for image generation, probably based on artificial intelligence algorithms (Image Generative AI). The speaker explains that this tool was trained on millions of images, many of which were taken from film image banks and cinematic references. It appears to have recently been revealed, via a leaked internal email from a developer at Midjourney, how the algorithm was trained. It is mentioned that the algorithm is particularly strong at producing photorealistic renderings, thus approaching the image quality found in advertising.

In addition, the speaker mentions a new version, “D3”, which has been integrated directly into ChatGPT. This version appears to be focused on image generation and allows users to directly create images from prompts. However, the speaker notes that this version has not yet reached the photorealistic level of quality, but it is good enough to generate non-photorealistic illustrations and creations.

Finally, the speaker discusses the future evolution of the tool, avec the planned arrival of version 4, 5, etc., and suggests that ultimately, these tools could offer different results and perhaps even come closer to photorealism.

ADVERTISING CAMPAIGNS:

Undiz and Stradivarius brand campaign (24:45- 26:12)

The speaker highlights the diversity of uses of image generation tools, noting that advertising campaigns have already been successfully carried out using this technology. They cite as an example the brand “Undiz ” which ran a campaign for its swimsuits this summer. In this campaign, the image of the model used to promote the swimsuits was generated by a Image Generative AI, as shown in the top left of the image. The “Cowboy” agency which created this campaign openly recognizes the use of this technology, emphasizing transparency in advertising.

They also highlight the quality and detail of the images generated, noting that the resolution is not limited to 512 by 512 pixels, and that the poses of the models can be diverse, such as that of a person taking an underwater selfie with a GoPro, for example. Additionally, they emphasize the importance of integrating real products into the generated images, which can be technically challenging but achievable, as the. demonstrates their collaboration with Martino.

Finally, they mention another ad campaign run by Stradivarius on Instagram, where flowers were superimposed on their products. This shows how companies are leveraging the creativity of image generation tools to create innovative and visually striking campaigns.

Campaign for the Château de Chambord (26:23 – 27:39)

It seems that the Château de Chambord also used images generated by Midjournée Image Generative AI as part of their advertising campaigns The speaker noticed a similarity between the posters of this campaign and the. style of Midjournée By reading the legal notices and examining the communications of the communications agency responsible for the campaign, it was confirmed that Midjournée was used to create these images, particularly those present in the metro, such as the. staircase in the shape of DNA and salamanders, symbols of Francis the First.

This highlights the emerging role of the “prompt artist”, a new profession that involves translating a concept or idea into text to generate a corresponding image. The speaker questions whether prompts are more effective in English than in French This is because the algorithms were mainly trained in English, which means that results may be slightly better when using English prompts on platforms like Chat GPT. only understands English, so if you use this platform you must communicate in English On the other hand, with D3 de Chat GPT you can use French to generate images, although their quality may be slightly lower because. initial training focused on English.

The communication campaign for Chamonix Mont-Blanc: “Reconnect with nature” (27:40 – 29:23)

The communications campaign for Chamonix Mont-Blanc, focused on the theme “Reconnect with nature”, appears to have attracted a mixed reaction. The featured image, showing people doing the racket with the SER on the right, was generated by Lia Although some participants found the image beautiful and impressive, others expressed disappointment and frustration.

Some have considered that using an image generated by a Image generative AI to represent Chamonix was a mistake. They emphasized that what people look for in Chamonix are the natural and authentic views of the valley and surrounding landscapes. The fact that the image did not represent local reality was seen as a disconnect with the very essence of the destination.

Ultimately, this use of an image generated by an image generative AI resulted in “bad buzz” for Chamonix Mont Blanc. This raises questions about the authenticity of representations used in tourism communications campaigns and serves as a reminder of the importance of capturing the true essence of a destination in promotional images.

LEARN THE ART OF PROMPT (29:26 – 30:42)
The speaker shares information about the art of the prompt, mentioning that they work with a trainer named Alexy Chon, who specializes in generative artificial intelligence. This trainer, from the field of blockchain and specialized in AI for a decade, became a trainer in 2023.

The speaker presents an example poster, where on the left we can see what they describe as an advertisement for a luxury perfume called “Electricity Fragrance by Midjourney”, generated from of a prompt. They specify that the characters presented in this advertisement do not exist in reality and that the creation of this image took approximately 5 minutes thanks to the prompt. On the other hand, the creation of the product packaging, in the shape of a d. The light bulb, took about 2 hours, as it was a complex process. Then, in Photoshop, they combined the three main elements of the poster: the text, the product packaging and the embracing couple.

ACADEI TRAINING: Midjourney Deep Dive (MDD) (30:43 – 32:22)

The speaker explains that academic training on the use of Midjourney Image generative AI takes place over one day and is carried out in groups, with a maximum of five people. This training includes introspection on past experiences and the limits encountered. and setting up Midjourney servers, as this can be complex Training takes place on Discord and covers the use of prompts, zooms, pans and all the commands needed to generate the desired images. tools to bring their ideas to life and are often amazed by the results obtained.

The training offers complex prompts for different types of renderings, including a manga style. Participants also have the choice to work on the consistency of recurring characters, human and non-human archetypes, etc. There is a strong emphasis on practical work, with support for creating prompts and images, as well as advice on optimizing them.

Participants leave with a wealth of content, including sample prompts, an image gallery, and tips for optimizing their use of Midjourney. A next session is planned for January 25 , or tailor-made training courses can be organized in the company to train several employees.

POSSIBLE FUNDING? CONTACT YOUR HR DEPARTMENT (32:23 – 33:02)

The speaker mentions that the training is eligible for funding, because their organization is Quailiopi certified, which includes Atlas, particularly for those working in the digital and financial fields. Then, he announces a surprise for Wallcrypt members: a promo code “wallcrypt24” which offers a 10% reduction. To stay in the Christmas spirit, he shares an image generated this morning representing a Christmas tree created by Midjourney Image Generative AI . He concludes by thanking the audience for their attention.

COPYRIGHT & SUBSCRIPTION QUESTIONS (33:07 – 36:04)

When you have an article to write, a blog article to illustrate, well there you go, it takes 2 minutes and then it’s good. When you have an image, you own the rights . If you have a paid subscription to Mjourney, you own the rights. For example, with Phil, we made t-shirts, he does freestyle scootering. Well, we have a whole series of, it’s a. e-commerce store for scooters, finally, t-shirts, polo shirts, etc. The illustration base, a bit in the freestyle hip-hop style, that we do with Midjourney. I also have the rights to use and commercially exploit these images and how can you prove it? So, you have the rights, there is a certificate, how is that? That’s a very good question. So Midjourney, the SA, would be the one who attacks me because he’s the owner. So he knows that I have my paid account and he sees the images. that I generated. So he has this image, finally, this information. And the other idea, and this is Alexi who is launching this, is to do it in NFT. And so, each time we generate it, we can run it through Photoshop, a small Photoshop plugin, there we say ‘OK, generate an NFT’. So here, we are clinging to Web 3 and the blockchain. And there, we have the title, so a bit like a solo envelope, it’s such a day, such a time, well it was recorded for the first time, that’s it. And so it dates the owner with my private key and û, how much Mjourney subscription. It starts at €10 per month, after that, in fact, it will go up to €30. This is also the speed at which the images will be generated. Either it takes, that’s it, you have to wait a little bit, 30 seconds for the image to be created if you have a €30 subscription, it’s almost instantaneous, that’s it. And then that’s how many images you can generate too. So it’s offers, but hey, there’s plenty to do with a lot of things. Yes, it’s accessible to everyone. Ah yes, yes, it’s really accessible, even everyone can indeed subscribe.”

DISCOVERY OF AIVERT, LITTLE WHATSAPP HACK

“I discovered something called Ibert. Ibert, I discovered this and this is what I use the most. That’s it, you can go to Discord, install, configure. That’s for people who want to go into expertise. Abert, it’s a group, finally, it’s a friend, it’s a new WhatsApp friend and in your WhatsApp, you have Midjourney, you know how to send a WhatsApp, you know how to make Midjourney prompts? Oh yes, that’s excellent! So here, I’m sharing my screen with you, right now there’s snow here, I don’t know if you also have snow where you live in the south, but there you go, I wanted to make snow landscapes. DSo here are the images that it generated. So you have the prompt at the bottom so snow and landscape French country. So here is a French snow landscape, golden hour. So that’s the sun setting a little. 16:9 format, version 6 that I use and so it generates this kind of images. Yeah, do you still give very specific instructions? Ah, that’s it, there you go. So that’s knowing how to Prompt but you, I write my prompts, you see in my WhatsApp there you go. So here I have my WhatsApp feed and so I write my prompts and I have the images and I get the images directly into my feed. And there it’s the same, it’s paid, it’s with Midjourney’s paid offer that you can. So in fact, there is no need to subscribe to Midjourney, you have to subscribe to Abert. Oh okay and what is the price, Ibert? And even well it’s really c I took for 30 € but for 30 € you have 300 images and in addition your credit is recharged so there I must be at 2000 images I don’t know why but I bought 200 on 3000 images so I think I will use it for you but what is interesting is that it is per image it is not per month so if there are months where you do not use it unlike Midjournée there you will pay your subscription ultimately for nothing you are on vacation you are at Christmas and everything is not useful you continue to pay your subscription there it is in the image so that makes me I find I prefer this version there when we want here we are n not all the functionalities of retouching an area of the image that’s doable in Discord it’s not feasible there but overall for the uses we have of it it’s really very very good there we redid our catalog brochure for web developer training 3 yes so all these images if that’s it so that was the image we had in the introduction to find blocks voà these images of blockchain blocks is that that’s the prompts that I put here B2B photos so there the B2B photo and to have magnificent images which are very very good on covers on brochures and which allow me to work on that I did it from my WhatsApp . So for PR, should we favor a series of keywords or write a sentence? So especially not a sentence, in no no in yeah that’s it in Midjournée on Chat GPT he will understand it, will interpret, he will do it to you but on Midjournée see that’s the way we’re going to Prompt it’s really words separated by commas so here it is road dramatic landscape Ocean View huge wave high angle golden hour panoramic and we’re going to cut these words into small pieces and that’s how we’re going to see the best results. There is the order of the words, yes quite indeed the words at the beginning are more important than the words at the end.”

ETHICAL ISSUES AND DEEP FAKE (47:24 – 49:07)

“Then the ethical questions, the fake news also that we can spread with that. So this whole false information part is really the real question of what can we do with it. I will end with this because it is also important to mention it. The universe is yes, it’s interesting but it’s not all rosy either, heh. There are people who also use it for bad purposes, look here so you must surely have seen Pap Enoudoun Balantiaga pass by, that doesn’t exist the false arrest of Donald Trump the false fire also of the Pentagon during the demonstrations against pensions so elderly person mistreated by the police it was debunked by the fp does not exist and then voice manipulations there too we can do fake Voice or clone voicing I know it since I I cloned it voice and then we can make anyone say anything from these more general tools there too here is the Deep fake I replaced my head finally the presenter from here that we see from France Info with my head and with my voice there it is so it’s just I clonedmy voice and it’s me who becomes presenter of France Info there are tools that allow you to do that to avoid testing in any case to see is it true or is it not true fake hor note A hor note which allows you to do that to send images or audios and it will tell you if it was fake or not ah there are already tools which allow you to do that effectively and well of course or given that that goes so far that humans are incapable of detecting that it will tell you if it is true or not and then what is this tool for example the one you use www.aiornot.com”

ENDING WORDS FROM JEAN-LUC LASQUELLEC (55:48 – 57:03)

Okay, listen, we thank you very much, thank you, thank you everyone for following and then I also remind you that for people who want to win and many training courses, you go to lotterie.wallcrypt. com since there is currently a lottery which has been launched and therefore the draw will be done at cryptosphere at the cryptosphere show so on Saturday and there are training courses to be won and Acadee is winning five training courses so that’s when not even anything and that’s it, well it’s actually the opportunity to be able to discover this training so participation is completely, I repeat, completely free on lottery..com that’s good very good and well thank you all this hour has really passed happened very very very quickly eh so it will actually be put on YouTube during the day thank you all thank you Samuel and then we will reschedule I imagine probably I would say another Masterclass perhaps on a subject perhaps a little more precise why not on the “videos” finally we will see we will do a little survey to find out where there is the most interest so that’s it, thank you and then see you soon and very soon.”