Tripodos, number 55 | 2024| OnlineFirst

ISSN: 1138-3305

Received: 29/11/23

Accepted: 03/03/24

OnlineFirst: 10/06/24

Publicación final:

Doi.org/10.51698/tripodos.2024.55.05

The Revolution Will Be Artificial. An Analysis
of AI-generated Audio-Visual Creation

Silvia Magro-Vela

Rey Juan Carlos University (Spain)

Pablo Sánchez-López

Rey Juan Carlos University (Spain)

Nuria Navarro-Sierra

Rey Juan Carlos University (Spain)

Generative artificial intelligence (GAI or Generative AI) is rapidly penetrating fields and disciplines that it had hitherto unexplored. This is a clear warning about the depth and breadth of the transformation that is taking place. This study examines the status of products within the audio-visual industry, focusing on those that used GAI in its image development process. The rapid adoption of these applications in recent times has led to a proliferation of audio-visual products, resulting in a fleeting and unfathomable universe that is constantly changing and expanding. Given the nature of the phenomenon, the study period spans from the first products created in 2016 to the creations produced in the summer of 2023. The findings show a constantly evolving process in which products are created using different tools, with multiple styles and motivations, while balancing creativity and technology. However, it is still early days to identify a clear set of features or claim that it has become a fully established trend.

Keywords: artificial intelligence, generative models, audio-visual production, creative industries.

Few topics have generated both as much controversy and uncertainty as the rise of Artificial intelligence (AI) and its application to the different areas of the audio-visual environment. While a European Regulation on Artificial intelligence is being drafted in Brussels (Ministry of Digital Transformation, 2023), the business world has not hesitated to use AI to develop its content products, be it video clips, fiction series or, above all, advertisements. Moreover, the various mass media industries could be among the most likely to benefit from the potential use of AI capabilities and applications to their business models in the future (Accenture, 2023). While this remains an unknown quantity, it is also business opportunity that will continue to grow in the coming decades, as illustrated by initiatives such as Amazon’s creation of a Generative AI tool for advertisers (Ruiz Cano, 2023).

Artificial intelligence refers to cognitive capabilities displayed by any type of software or equivalent technology (Russell and Norvig, 2020). As a field of knowledge, AI initially belonged to computer science studies aimed at inventing, creating or developing intelligent ‘machines’ capable of interacting with human beings, either through text, voice, image, audio or a combination of them. The concept of AI arose in 1956 at the initiative of Professor John McCarthy, during one of the sessions of a working group on the subject held at the University of Dartmouth, in order to differentiate AI itself from other, possibly confusing, similar terms, such as cybernetics (Crevier, 1993). For McCarthy, AI was a part of computer science ‘devoted to the design of machines that were capable of simulating some human behaviours usually labelled as intelligent’ (Gross, 1992, p. 73).

The already numerous definitions of the term in the academic literature continue to grow with AI’s growing presence and extent of use in society. For Ransbotham, Kiron, Gerbert and Reeves (2017), AI is both the theoretical work and the practical development of any kind of computer software capable of carrying out functions or tasks that we commonly associate with the human intellect such as pattern recognition in images, voices and sounds, particularly with its own decision-making ability, resulting in a recognisable output (text, voice, image, etc.). As its capabilities have advanced in the 2020s across the globe, AI is increasingly seen as a tool at the service of businesses or individuals, whether this is as a voice assistant (Alexa), an image generator (Midjourney), or the manager of an online advertising campaign (as the algorithms in Google or Meta are already operating). For the public, the first substantial advances in AI took place in the 1990s, with the highly publicised victory of the Deep Blue computer against chess champion Garry Kasparov (Kasparov, 2017). At the time Deep Blue was considered a ‘supercomputer’, a term that is no longer in use. In the 2020s, algorithms and incipient AI programmes are fed not only by data provided by users themselves, but also by companies. This was made possible thanks to the development of the omnipresent Internet, which permeates our homes, our jobs and even our pockets, with gadgets such as smartphones and smartwatches. Within the technological-social change in which we are immersed, AI is an inextricable part of other scientific developments such as augmented reality, the so-called Internet of things and virtual reality.

While the original chatbots could only respond to pre-programmed scripts, machine learning AI systems are based on algorithms that allow them, with minimal human intervention, to find patterns in data, even without being specifically designed for it. On the other hand, deep learning, which is understood as an evolution of the former, employs what are known as artificial neural networks (ANN) —which mimic the neural connections of the human brain— with a mode of learning through the processing of a vast amount of data (Franganillo, 2022).

THE INTEGRATION OF DIGITAL TECHNOLOGIES IN AUDIO-VISUAL PRODUCTS

The history of audio-visual products is the history of technological advancement applied to image manipulation. Manipulation not in the ideological sense per se, but rather in the aesthetic sense, with a view to altering images and sounds to better serve the vision of content creators. In fact, altered photographs were initially not intended to deceive the public but, rather, to achieve a more accurate resemblance between an image and reality (Tucher, 2017). Photography as a still image and its applications, whether in commercial advertising or photojournalism, have leveraged the ability to digitally manipulate images since the 1960s (Cerdán and Padilla, 2019), reducing the cost of manipulation processes and their ability to reach a wider audience, which is now benefiting from the new opportunities afforded by algorithmic images (Gómez Gómez and Rubio Tamayo, 2023).

In audio-visuals, a path can be traced from the manipulated moving image to the artificial creation of images, with or without sound/audio. In the digital era, technological applications can be found both in studios (chroma) and in terms of visual effects, ‘digital manipulations and enhancements of the footage that] happen primarily during post-production (Dinur, 2017). Within visual effects, the creation of partially or fully digital images is associated with the term CGI (computer-generated imagery), a ubiquitous concept in the Hollywood blockbusters of the 21st century. Some CGI creations have their own characteristics. This is the case of morphing, where patterns are identified between two images or frames to create a moving transition leading from image A to image B (Ivakhiv, 2016). One of the earliest and best known uses of morphing in popular audio-visual culture is Michael Jackson’s music video Black or White, released in 1991. Another case would be that of warping, the digital alteration or distortion of an image usually used to create comic effects by, for example, exaggerating the features of a human face (Boté-Vericad and Vállez, 2022).

The logical evolution of these manipulations is deepfake, in which the alteration is no longer aimed at exaggeration but at recreating an image —usually the face— of a real person. In line with Chaikin and Barty’s (2023) definition, it is a synthetic medium that replaces the image and/or audio of one person with that of another; hence they are called deepface and/or deepvoice. This technique has been used for years in Hollywood to rejuvenate actors and actresses in certain films, for example, Robert De Niro in The Irishman (Martin Scorsese, 2019).

The evolution of Generative AIs, capable of creating their own images (always within the parameters of the basic algorithm) is a creative and business opportunity within the audio-visual world, whether in the broadcasting of advertisements such as that by Cruzcampo spot that ‘recreated’ Spanish singer Lola Flores (Palomo-Domínguez, 2021) or the optimisation of film post-production (Caballero, 2023).

In this new configuration of society, where the integration of AI increasingly permeates more areas, its role as an assistant or as a tool that helps individuals in the most artistic disciplines to improve and reduce creative processes is beginning to be seen. But it also has raised public concern about the positive and negative aspects of the use of AI to create or alter images. This phenomenon contributes to a new reality about which debate is already taking place from different perspectives that address everything about technical aspects, as well as issues from an ethical, deontological, and legal point of view (Franganillo, 2022; Magro-Vela and Navarro-Sierra, 2024).

Having contextualized the current situation regarding AI and its tangential position with respect to the audio-visual medium, this research aims to make an approach to the first official audio-visual products that have integrated GAI in their production processes. This will allow us to understand the emerging reality around the use of this technology in professional environments.

Thus, the following questions were used to formulate a starting point for the study:

Q1. What GAI technology or tools are currently used and how are they used in the creation of products within the cultural industry?

Q2. Does the inclusion of GAI remain on the surface through purely technological use, or does it permeate deeper levels by reaching into the layers of narrative discourse?

Q3. What is the degree of suitability of images created with GAI in relation to the brand, within the analysed pieces?

Q4. What are the reasons behind the use of applications and, therefore, the benefits of GAI?

METHODOLOGY

The present study uses a qualitative and descriptive methodology to approach the entry of GAI technologies into cultural industry, including graphic or audio-visual productions, whether in cinema, video, or advertising. This is an exploratory study because it is an early-stage area of research that deals with pioneering uses of different types of GAI for the creation of audio-visual products.

In order to structure the research, we have opted for a qualitative content analysis from the perspective of audio-visual analysis, taking into account the variables inherent in the use of GAI. A non-probabilistic convenience sampling technique was used in the analysis, which, as defined by Wimmer and Dominick (1996, pp. 70-71), is a filter intended to determine whether an item is eligible or not to be part of the sample. 

To answer Q2, this analysis is complemented by a narrative approach (Chatman, 1990) in which, besides studying the structure, explicit mention of AI in the unit of analysis is considered.

Regarding Q3, one of the limitations of this research is the lack of information in relation to the motivation for the use of GAI. Therefore, this has been addressed through the consultation of media in which both brands and creators have been interviewed about the use of AI in the creation of these products. To complement this, an interpretation based on the comments from the brands or creators themselves on platforms and social networks regarding the launch or premiere of the mentioned campaigns has been carried out.

Several parameters were established to delimit the items to be analysed. Only official productions were included, and therefore all those that had arisen as a result of creations by users or by authors other than those within the company or brand itself were discarded. Despite appearing real, the well-known Taco Tornado and Peperoni Hug videos were nothing more than AI-created videos from non-existent companies, which is why they have not been included. The Croix-Rouge Française campaign was also excluded; while at first it seems to have been created using GAI, it mentions that sadly, the photographs were not AI-generated, with reference to the crude but true images displayed.

Products using deepfake technology have not been taken into account either, as it is considered to be an earlier tool which aims to alter the natural characteristics of an image in order to obscure, hide or modify an existing reality. The ‘Visit Denmark’ campaign aptly illustrates this point by promoting tourism in Denmark by bringing to life the characters of famous works of art such as La Gioconda (Leonardo da Vinci, 1503-1519) or The Girl with the Pearl Earring (Johannes Vermeer, 1665-1667), by the use of deepfake technology.

Similarly, only finalised products were included; therefore, products that have the ability to interact with the consumer and can be personalised were excluded. Examples include the Galician company Emotional Films, which creates films with famous characters, and Vespa’s campaign to generate soundtracks for different cities.

In addition, there is the need for a visual component that may or may not be accompanied by sound. This characteristic is determined by the image (whether static or in motion) being the raw material of analysis. This leaves out those products where GAI input is limited to script writing, such as the 2018 Lexus campaign, and music creation. Along the same lines, units of analysis were required to be complete pieces, which caused creations that function as illustrations of news items or parts of a larger message to be discarded. For example, this excludes the image published in April 2023 on the front page of El Mundo in which Pablo Iglesias and Yolanda Díaz were portrayed walking arm in arm. 

A time period was also established, which spans from the first pieces that acknowledged having used GAI for their creation in 2016, until July 2023, when the study phase of this research began. The filtering was conducted by using Internet search engines and entering keywords. The terms used, both in Spanish and English, were divided into two categories. One of the categories referred to products: ‘audio-visual’, ‘production’, ‘graphics’, ‘posters’, ‘advertising campaign’, ‘music’, ‘video clip’, ‘film’, ‘short film’; ‘television’; ‘series’; and the other referred to artificial intelligence, including the names of the most well-known software: ‘AI’, ‘artificial intelligence’, ‘Midjourney’, ‘DALL-E’, ‘Stable Diffusion’.

The combination of these led to identify products and websites that alluded to other products which in turn became part of the sample. Having considered all these aspects, 29 items1 were analysed. Although some of them were composed of several pieces, they were analysed as a single campaign or product by following the same principles: all the variables were taken into account in the different pieces. The classification obtained was as follows:

Figure 1. Classification by product type

Source: Developed by the authors.

Additionally, there were 16 still image products and 13 video products2. The first category includes the magazines and seven of the advertising campaigns, whereas the second category includes the video clips, the openings and the other five campaigns.

Ten of the products had an international scope. These included the Coca-Cola campaign, the opening of the Secret Invasion series (Kile Bradstreet, Disney+, 2023), and video clips of internationally known bands such as Linkin Park, as opposed to other nationally-based products (n19)3, such as the Burger King campaign for Germany and Colombia, the covers of Vogue magazine in Singapore, and the video clips of little known bands (although the use of GAI in their production may have boosted their impact outside their countries).

All data were collected on a work sheet specifically designed for the study. This encompasses variables that dealt with basic information about each piece: V1 and V2 referred to the name of the product and the brand or company to which it belonged; V3 listed the date of publication or release of the product; V4 identified the type of product (advertising campaign, magazine, video clip or fiction series); V5 included the sector to which the brand belonged after the product analysed, including publishing (magazines with different themes, such as fashion and trends, economy, design, etc.); food, advertising, music, and television (the opening of a TV series), etc.; V6 detailed the nature of the image in relation to the movement of the image (still or video); the authors of the pieces were reflected in variable V7, and this first block ended with the geographical scope of the piece, in variable V8.

There was also a second block of variables about GAI itself: V9: Application used; V10: Type of GAI; V11: GAI function; V12: inclusion; V13: Contribution of GAI to the product. 

Variable V14 was added to analyse the narrative. This is concerned with the discourse of the piece and whether GAI was introduced into the narrative of the piece, and then moves on to a block on the visual design of the piece. The variables contained in this section were: V15: aesthetics, that is, the style of the image with respect to the existing ones (illustration, futuristic, realistic photography...); V16 analysed whether there was a connection between aesthetics and narrative; V17 measured the level of realism by taking the degree of iconicity of the scale proposed by Villafañe and Mínguez (2006, p. 39); V18 and V19 dealt with the presence of errors in the image and their type; and this block ended with V20, which was only completed in the case of a moving image to determine whether the piece was a succession of still images (similar to the Stop Motion technique) or an animated image.

The last block focuses on elucidating whether there was a different way of making the same piece without using GAI (V21); determining what this would be (V22); and concluding by explaining the reason for using GAI in this audio-visual product —including cost, opportunity, virality, curiosity, time, etc. (V23).

The coding of the 29 items was carried out independently by the three researchers in August 2023. Inter-coder reliability was ensured by pooling the results in order to reach a degree of agreement of 95%. For those variables where the discrepancy was found to be greater, adjustments were made and re-coding was carried out in order to achieve reliability of the coding criteria.

RESULTS

The data obtained from the analysis were grouped according to the same organisational criteria used in the design of the worksheet.

GAI Technology Options

Today’s generative artificial intelligence (GAI) image applications or tools have a common origin: Open AI. The pioneering company conducts research into GAI implementation and is the creator of ChatGPT. It is open source, which means that many applications have emerged that make use of its advances. This is the case for DALL-E and its updates. 

Stability AI emerged at a later stage. This is open source and is accessible to everyone. It is behind Stable Diffusion in all its versions and other applications, such as Dream Studio and Kaiber, the latter having been used in the Linkin Park video clip that has been analysed.

It was impossible to find this information for give of the 29 units analysed. In the remainder, there was a remarkably more common use of Midjourney (n14), compared to a more even use of DALL-E (n5) and Stable Diffusion (n2). 

A mix of technologies was also detected in Yorokobu and Vogue Singapore. In the former, Ángel Alonso and Ruth Falquina used ΔIko and DALL-E 2 to create the cover for the publication. In the words of these creatives, the two of them together with ΔIko (a multidisciplinary form of AI) formed a collective called AI Gang, where ‘Humans-AI, AI-Humans influence each other and together we explore new territories of artistic expression’. It is a new era of the algorithmic renaissance, a new artistic current called AIart’ (Quoted in Lago, 2022).

For Vogue Singapore’s special issue in March 2023, titled Roots, Varun Gupta (creative director of Mumbai-based content agency We Create Films) designed avatars that represented a blend of tradition and innovation by working with Midjourney and DALL-E.

Image 01. GAI-created covers for Vogue Singapore

Source: Vogue Singapore.

Likewise, in the Paintress video clip by The Streechers and in the opening of The Secret Invasion, creators used customised artificial intelligence-based tools that were personalised for the project on an ad hoc basis.

Therefore, there was no dominant trend in the use of GAI applications linked to the type of piece according to the sector to which it belonged, but rather they were used interchangeably. However, Midjourney did stand out as the most widely used technology in the corpus of analysis, even though it is closed source, with limited use through Discord and a paid subscription.

As far as the origin of the generated images is concerned, they were sometimes produced based on an image (they belong to the GAI category of Image2Image), either still or moving, as in Lost, the Linkin Park music video, or ‘Masterpiece’, the Coca-Cola campaign. However, most of the case studies shared text as a starting point for image creation (Text2Image).

In the ‘Unbottling Martini’ campaign, as Jiménez (2023) explained, images were generated to ‘visualise the ingredients, processes and flavours that come together in a bottle of vermouth’. In this way, as the author argued, ‘the creators introduced keywords such as botanicals, floral, petals, flowers or chamomile’ and this resulted in evocative images that were the result of Midjourney and ASMR (Autonomous Sensory Meridian Response) (Creative Brief, 2023).

Image 02. Images from the ‘Unbottling Martini’ campaign

Source: AVM BBDO.

Contrary to what it may seem, designing prompts is not a simple task. The music video for Disturbed’s Bad Man (directed by Tristan Holmes) required a thirty-day period of inputting prompts that resulted in the generation of more than 10,000 frames, as the American metal band pointed out on their YouTube channel (Disturbed, 2022).

Karen X. Cheng, the digital creator behind the cover of Cosmopolitan, explained on her Instagram account that it is not a matter of magic, nor does it just take the insertion of a few words. The artist discussed her work process for the publication:

For something like this, there was a TON of human involvement and decision making. While each attempt takes only 20 seconds to generate, it took hundreds of attempts. Hours and hours of prompt generating and refining before getting the perfect image. Not to mention many zoom calls, emails, text message threads to bounce around ideas (Cheng, 2022).

While it is true that most of the elements studied used GAI to create complete images, there were three cases in which the outpainting tool was used. This option allows you to extend your own image beyond the original boundaries and has been used on the covers of Yorokorbu, Cosmopolitan and in the advertisement for La Laitière (Image 03). The campaign created by the advertising agency Ogilvy Paris took the oil painting by Johannes Vermeer entitled The Milkmaid (1657-1658), which serves as the brand identity for the multinational, and used GAI to create a new, enlarged work (Image 04).

Image 03. Use of the outpainting tool in the advertisement of La Laitière

Source: Ogilvy Paris YouTube channel.

Image 04. Comparison of the original work (left)
and the resulting GAI work (right)

Source: Google (left) and Ogilvy Paris YouTube Channel (right).

Role and Narrative Function of GAI

Another issue to be considered is how GAI was included and what it contributed to the audio-visual piece. There were only three cases where the piece was purely the result of GAI as far as the image was concerned. There were three video clips: Paintress (The Streechers); Age of Illusion (Die Atwood); and Entergalactic (Kid Cudi). The rest of the items (n26) that made up the sample analysed chose to include GAI as one element that was combined with others, such as software and editing, post-production, layout, retouching or programmes to generate texts to insert in the image, a task in which image GAI is currently not very efficient. 

Perhaps the most complex case in this respect was Coca-Cola’s ‘Masterpiece’, which combined several techniques. Reuters Fact Check (2023) reported statements from the campaign’s post-production studio which reported that the production of the campaign involved live action footage with actors, animation, computer-generated imagery (CGI) and some GAI techniques.

It is possible that one of the essential questions in the emerging debate on the suitability of GAI for the creation of pieces in the audio-visual industry is whether its use is an opportunistic decision, solely leveraging the contribution from a new technology, or, on the contrary, whether its contribution goes beyond and breaks into the narrative or discourse itself. 

Of the units analysed, 62% (n18) made a purely technological use of GAI, compared to 14% (n4) in which a narrative purpose was present. There is also room for a combination of both, which held an intermediate position with 24% (n7).

Moreover, there were cases in the sample where GAI was explicitly alluded to, such as in publications with issues dedicated to it, including Cosmopolitan and The Economist. There were also campaigns that took advantage of GAI itself to build their discourse and did not hide it. This was the case at Heinz Ketchup, KitKat, Burger King and Publicis Groupe, among others. 

Image 05. Frames of the KitKat campaign (top)
and Heinz Ketchup campaign (bottom)

Source: KitKat Australia YouTube Channel and Heinz Ketchup YouTube Channel.

The first two refer to the use of GAI to reinforce the brand or even the claim associated with it (Image 05). The ketchup brand showed that even GAI recognised it as the only point of reference when it comes to creating images of this ketchup in different styles and contexts. KitKat, meanwhile, thanked technology for giving it a chance to take a break. Burger King has created flavours through GAI that have then crystallised into real products and Publicis Groupe is taking the opportunity to embarrass all those who six years earlier, in 2017, laughed at its investment in an AI-powered platform called Marcel.

Image 06. Publicis Groupe (left)
and Burger King product (right) advertising posters

Source: Marketing Directo (2023) and Instagram Burger King.

Visual Design Generated with GAI

In terms of product aesthetics, the problem faced by brands is how to differentiate themselves and their products from what could be classified as the usual styles that GAI generates if the originator does not go much further in achieving something unique and distinct through the prompt. This leads to one of the now common criticisms, namely that GAI will not result in new and original products, as it is created through an algorithm based on references to existing elements. Although the following catalogues were not included in the sample (because they were published after the sample’s time frame), they are a clear example of what has just been mentioned.

Image 07. Juguettos and Toy Planet Catalogues for Christmas 2023

Source: juguettos.com and toyplanet.com.

Different styles can be found in the sample; those that imitated realistic photography (n10), those based on illustrations (n14) of different types (psychedelic, Y2K, science fiction, futuristic, anime, etc.) with a greater or lesser degree of realism, and those that mixed different styles in the same product (n5), such as the Generation Bershka perfume campaign, which used anime-style illustrations and realistic photography, among others. The truth is that most of the products studied have an aesthetic related to the brand or the discourse, and on a few occasions the style is indifferent in relation to the product, giving rise to examples such as the one above. The adaptation of the visual style to the narrative was evident in advertising campaigns such as ‘The Next Rembrandt’ by ING, ‘Which football club would the most famous Spanish authors throughout history support?’ by Panenka magazine, and ‘The French Re-Connection’ by Citroën. These three works recreated ancient paintings to show what something ‘in the style of’ would look like but created by GAI.

Image 08. Products with a pictorial style

Source: YouTube channel @thenextrembrandt2907 (top left),
portalpublicitario.com (bottom left) and panenka.org (right).

However, adapting it to the brand that created the product is something that can also be seen in the campaigns by Burger King, Stradivarius or Glamour, among others. Their results are reminiscent of their other advertising campaigns, something that can be seen at a glance on their social media accounts such as Instagram.

Image 09. Instagram feeds from Burger King, Stradivarius and Glamour
featuring AI-generated products

Source: Instagram accounts of Burger King Germany, Stradivarius and Glamour Singapore.

Errors in AI-created images have been the object of major criticism. Some examples can be seen in the construction of the image for several of the creations in the sample. These include Cervantes’ six fingers in Which football club would the most famous Spanish authors throughout history support? (Image 08), the KitKat and Publicis Groupe (Image 06) advert with disfigured faces and the video clip Age of Illusion by Die Antwoord, which merges images together to generate deformations creating extra heads. However, the narrative for these products either contains the information that they are images created by GAI, or speak of illusions and simulation, so these errors often tend to reinforce the intended message. This was the case with the KitKat Burger King and Heinz Ketchup advertisements, which alluded to GAI as a creator of products and then mocked the results; it also applies to some video clips such as the one mentioned above and to Atropos by Periphery, which deals with the subject of human artificiality and its consequences.

Rationale for Using GAI

Finally, our study sought to explore the reasons for using GAI in the different products. While many of the creators have given interviews or explained the reasons behind their productions, in other cases this information could not be found, although it was evident in the final product for some of them.

Thus, there are cases where artificial intelligence played a secondary role, acting merely a tool that facilitated the achievement of a final result, which would be more time-consuming and costly to develop without it. This was the case of the opening of the Marvel series and of the ‘Masterpiece’ campaign by Coca-Cola. In these products, artificial intelligence is not essential to achieve the final piece, as they are the work of different professionals and software backed by brands with a large budget.

In contrast to these, there was a second set of productions that incorporated artificial intelligence motivated by the desire to experiment, be popular and/or capitalise on the virality associated with the use of these technologies today. In turn, these could be classified into two categories: those where the narrative depends entirely on the existence of GAI (without which their creation would be meaningless), and those which could have been created in the traditional way if this type of technology had not been available.

The former include ING’s ‘The next Rembrandt’, which, besides being the oldest product in the sample, was a campaign aimed at creating a painting in the style of the painter through GAI after analysing his works (also carried out by artificial intelligence) and was taken to the physical plane with the help of a 3D printer; other examples are also the KitKat, Burger King and Heinz Ketchup campaigns, which had a more informal tone and expressed amusement at the results. These brands that have created pieces that take advantage of the popularity of GAI (without having any link to it), and have made it part of the discourse, justifying the use of these type of images.

Also, in this group were covers taken from issues of The Economist, Somos and New Accountant, all of them partly or fully devoted to the topic of GAI within their sector, which therefore have a more direct justification for its use.

Image 10. Covers for issues of The Economist, Somos and New Accountant

Source: economist.com, elcomercio.pe, newaccountantusa.com.

On the other hand, there were products where there was no particular reason for the use of GAI, nor was it narratively justified. These were not sectors or products related to artificial intelligence and the campaign was not linked to its use in any way; however, this ‘partnership’ was built by connecting the concept of Artificial intelligence to their brand with the sole intention of jumping on the bandwagon for these types of creations to encourage discussion about the product simply because it was created with a GAI. This seems to be the case of advertising campaigns such as Martini or Bershka’s fragrance. Although this motive is not exclusive to those which linked AI to their narratives, such as some of those mentioned above.

The reasons behind the creation of some of these products may range from the mere fact of having the knowledge available to use this technology, as was the case with the video clip Paintress by The Streechers, or La red, by the group Kinesis4; to the simplicity and speed of creating an audio-visual product with GAI, as indicated by Cosmopolitan magazine, which acknowledges this ‘ease’ on its cover by mentioning that it took less than 20 seconds to create it.

Image 11. Cover of Cosmopolitan the A.I. issue

Source: cosmopolitan.com.

DISCUSSIONS AND CONCLUSIONS

The previous sections have addressed and answered the research questions posed in the introduction, which form the basis and starting point for this paper.

Revisiting the research questions, it’s possible to propose a plurality that transcends all the aspects analysed about the inclusion of GAI in audiovisual production.

Q1. After the analysis, no preference for a specific technology was detected, although the Midjourney tool is the most prominent. This could be justified because, at the time the pieces were created, this tool was free or low cost, but also stood out as offering the best results. This advantageous condition has been reducing with the arrival of new GAIs and improvements in the existing ones. In addition, the use of technology was mostly oriented towards creating new images from scratch, however, some cases like La Laitière are observed where the outpainting technique is used.

Q2. It has been found that GAI is primarily used as just another technological tool, as opposed to its minority use leveraging the popularity that the democratization of AI has brought to include it as part of the discourse.

Q3. Beyond errors or recurring elements that sometimes intentionally evidence that it is a creation with GAI, the analysed products achieve, with varying degrees of success, quite a fit with the visual style of the brand they represent.

Q4. Regarding the motivations for using GAI, two distinct profiles have been identified that are linked with all the previously mentioned aspects. On one hand, the choice of this technology due to the reduction of production costs compared to other forms of creation. And on the other hand, those who incorporate GAI for the possibility of experimentation for the creator and/or for the popularity and virality that its use offers.

By way of conclusion, we would like to make some considerations, or rather, to suggest a few avenues for research in the future that lead to new questions and possible applications of generative artificial intelligence.

Products from different sectors and brands were included in the sample. If the period had been extended, there would have been a greater number of creations, new sectors and brands would have been added, such as the aforementioned catalogues, new Amazon Ads (Puro Marketing, 2023), book covers, and new fiction series such as La Mesías (Javier Ambrosi and Javier Calvo, Movistar+, 2023) with a sequence created by GAI.

However, before embarking on this practice, content creators will have to choose between embracing the trends and opportunities offered by GAI, or facing criticism from a certain part of the public for the use of these technologies. An example of the latter can be illustrated by the reactions to the covers of toy shop catalogues, and the Bloomsbury Books of The New York Times.

The latter used an Adobe Stock Image that had been generated by GAI for one of its covers5, something similar happened with the graphic tablet brand Wacom when they launched their New Year campaign. Both cases were strongly criticized on social media for not having used the services of an illustrator (El Comercio, 2023; Wacom, 2024). These examples hint at the complex situation from an ethical standpoint in the use of GAI and set up a context in which it will be necessary to observe and discuss the repercussions of its inclusion in the professional field.

Beyond that, the future of GAI in product creation within the creative industries looks promising, especially in advertising, as shown by the 12 advertising campaigns in the sample of this study. In addition, as some forecasts in the MarketsandMarkets have reported (cited in IP MARK, 2023), there will be an annual growth of 29.6% in the global AI-generated advertising market until 2026. This will represent a market value of $13 billion.

In short, it will be necessary to wait and see how GAI is finally integrated into audio-visual creation and whether it becomes an ally (as anticipated) or the latent or manifest threat envisaged by some.

Silvia Magro-Vela. PhD in Audiovisual Communication from the Rey Juan Carlos University. Assistant professor in Media Audiovisual Graphic Design and TV Production at the Department of Audiovisual Communication and Advertising at the Rey Juan Carlos University. Her teaching and research interests are within the framework of Graphic and Social Design, TV series, and image composition. She is currently interested in artificial intelligence imagery generation and design. She also holds Films, Television and Interactive Media Master and is a research member of the project Figure and background. Iconic action to rediscover rural women. The photographic archive through art, funded by Spanish Ministry of Science and Innovation (FCT-22-18138). She also has specialised academic training and professional experience in Television Production and Film Production.

https://orcid.org/0000-0002-0695-8813

Pablo Sánchez-López. Holds a PhD in Audiovisual Communication from the Rey Juan Carlos University. His doctoral thesis examined the geopolitics of popular culture. He is a visiting professor in Media and Communication Studies at the Department of Audiovisual Communication and Advertising at Rey Juan Carlos University, and a lecturer in the master’s program in Entertainment Management at the Core School. His line of study is politics and social representation in communication. He has participated in research projects such as Study of the conditioning factors of disinformation and proposal of solutions against its impact according to the degrees of vulnerability of the groups analysed, financed internationally by the Luca de Tena Foundation and Facebook or Teenage self-harm representation in Social Media financed by Spanish State Research Agency (PID2021-124550OB-I00).

https://orcid.org/0000-0003-0113-509X

Nuria Navarro-Sierra. Lecturer at Rey Juan Carlos University in the Faculty of Communication Sciences, Department of Audiovisual Communication and Advertising. In addition, she has a six-year research period (CNEAI). She holds a PhD in Audiovisual Communication from the Complutense University of Madrid with an international mention and a PhD Extraordinary Award. She is a research member of the projects Iberian digital media research and factchecking hub (IBERIFIER) of the European Digital Media Observatory of the European Commission (2021-2024); and Politainment in the face of media fragmentation: Disintermediation, engagement and polarisation —ref. PID2020-114193RB-100— (2021-2023), funded by the Ministry of Economy and Competitiveness. Her main lines of research include work on the study of disinformation in the media, social and sustainable design in relation to the SDGs, as well as the use of serious games in journalism and education. Recently, she has started a new line linked to design through artificial intelligence as a tool for image generation.

https://orcid.org/0000-0002-1431-1534

References

Accenture (2023). Accenture Technology Vision 2023. Cuando los átomos se cruzan con los bits: Los cimientos de nuestra nueva realidad. https://newsroom.accenture.es/es/news/2023/technology-vision-2023. Accessed 28 October 2023.

Boté-Vericad, Juanjo José and Vállez, Mari (2022). Aplicaciones de deepfakes: Manipulación de contenido audiovisual y riesgos para los usuarios basados en las políticas de privacidad. Documentación de las Ciencias de la Información, 45(1), 25-32. http://doi.org/10.5209/dcin.77256.

Caballero, Jorge (2023). Hacia una nueva dimensión del montaje cinematográfico: Explorando las posibilidades de la inteligencia artificial. Hipertext.net, 26, 53-58. https://doi.org/10.31009/hipertext.net.2023.i26.08.

Cerdán Martínez, Víctor and Padilla Castillo, Graciela (2019). Historia del ‘fake’ audiovisual: ‘Deepfake’ y la mujer en un imaginario falsificado y perverso. Historia y Comunicación Social, 24(2), 505-520. https://doi.org/10.5209/hics.66293.

Chatman, Seymour B. (1990). Historia y discurso. La estructura narrativa en la novela y en el cine. Taurus Humanidades.

Chaykin, Adlan and Barty, Michael (13 June 2023). ‘No soy yo’: El creciente uso de deepfakes plantea nuevas amenazas para las organizaciones. Control Risks. https://www.controlrisks.com/es/nuestro-analisis/spanish/no-soy-yo-el-creciente-uso-de-deepfakes-plantea-nuevas-amenazas-para-las-organizaciones? Accessed 8 October 2023.

Cheng X., Karen (21 June 2022). Creating the first ever artificially intelligent magazine cover for @cosmopolitan! [@karenxcheng]. https://www.instagram.com/reel/CfEwohiJdXW/?utm_source=ig_embed&ig_rid=cc776791-1317-48eb-8565-d17162bd4302. Accessed 17 November 2023. 

Creative Brief Martini (2023). AMV BBDO. https://www.creativebrief.com/agency/amv-bbdo/work/unbottling-martini. Acces­sed 2 November 2023.

Crevier, Daniel (1993). AI: The tumultuous history of the search for artificial intelligence. Basic Books.

Dinur, Eran (2017). The filmmaker’s guide to visual effects. The art and techniques of VFX for directors, producers, editors, and cinematographers. Routledge. 

Disturbed (2022). Disturbed - Bad Man [Official Music Video]. https://www.youtube.com/watch?v=fpUpVznI4Yc. Accessed 11 October 2023.

El Comercio (15 May 2023). Un famoso libro usó inteligencia artificial para su portada y las críticas le llovieron. El Comercio. https://elcomercio.pe/tecnologia/inteligencia-artificial/un-famoso-libro-uso-inteligencia-artificial-para-su-portada-y-las-criticas-le-llovieron-bloomsbury-new-york-times-bestseller-ia-adobe-stock-noticia/. Accessed 18 October 2023.

Franganillo, Jorge (2022). Contenido generado por inteligencia artificial: Oportunidades y amenazas. Anuario ThinkEPI, 16. https://doi.org/10.3145/thinkepi.2022.e16a24

Gómez Gómez, Hernando and Rubio Tamayo, José Luis (2023). Algorithmgraphy as a milestone and phenomenon in the production of still images in the digital era: Resignification of the notion of the photographic image and projection of the medium in a context of image production with artificial intelligence and machine learning. VISUAL REVIEW. International Visual Culture Review / Revista Internacional de Cultura Visual, 14(2), 1-13. https://doi.org/10.37467/revvisual.v10.4607

Gross Salvat, Begoña (1992). La inteligencia artificial y su aplicación en la enseñanza. CL& E: Comunicación, Lenguaje y Educación, 13, 73-80.

IP MARK (2023). El mercado global de publicidad con IA crecerá un 29,6% anual hasta 2026. https://ipmark.com/el-mercado-global-de-publicidad-con-ia-crecera-un-296-anual-hasta-2026/. Accessed 14 October 2023.

Ivakhiv, Adrian (2016). The art of morphogenesis: Cinema in and beyond the Capitalocene. In Shane Denson and Julia Leyda (Eds.), Post-cinema. Theorizing 21st-Century Film. REFRAME Books, 724-749. https://doi.org/10.25969/mediarep/13475

Jiménez, Berta (10 January 2023). Martini, la primera marca de bebidas espirituosas en crear una campaña con inteligencia artificial. Marketing Directo. https://www.marketingdirecto.com/anunciantes-general/anunciantes/martini-lanza-campana-inteligencia-artificial. Accessed 15 October 2023

Kasparov, Gary (2017). Deep thinking: Where machine intelligence ends and human creativity begins. Hachette.

Lago, Carmen (6 September 2022). Así es la primera portada de Yorokobu creada por una inteligencia artificial. Yorokobu. https://www.yorokobu.es/portada-de-yorokobu-creada-por-una-inteligencia-artificial/. Accessed 10 October 2023.

Liz Gutiérrez, Antonio Manuel (2020). ¿Un mundo nuevo? Realidad virtual, realidad aumentada, inteligencia artificial, humanidad mejorada, Internet de las cosas. Arbor, 196(797), a572. https://doi.org/10.3989/arbor.2020.797n3009.

Magro-Vela, Silvia and Navarro-Sierra, Nuria (2024). Inteligencia artificial generativa ¿Una nueva era de la imagen o un ciclo más en su evolución? In Francisco Segado Boj and Jesús Díaz de Campo Lozano (Coords.), La investigación sobre comunicación en España. Tendencias actuales y perspectivas de futuro. Fragua. https://hdl.handle.net/10115/31465.

Marketing Directo (19 June 2023). Publicis Group se ríe ahora de todos los que se burlaron de su IA Marcel en Cannes Lion. https://www.marketingdirecto. com/especiales/cannes-lions/publicis-groupe-burlas-ia-marcel-cannes. Accessed 14 October 2023.

Ministerio de Transformación Digital (7 November 2023). España, pionera en el impulso de la regulación de inteligencia artificial, pone en marcha el entorno controlado de pruebas del Reglamento Europeo de Inteligencia Artificial. [Press Release].

Palomo-Domínguez, Isabel (2021). Del mito a la viralidad: El caso de la campaña de Cruzcampo que resucitó a Lola Flores. aDResearch ESIC International Journal of Communication Research, 26, e262. https://doi.org/10.7263/adresic-026-02

Puro Marketing (25 October 2023). Amazon transforma la publicidad con su IA generativa para crear imágenes impactantes en los anuncios y campañas publicitarias de las marcas y anunciantes. https://www.puromarketing.com/76/212788/amazon-transforma-publicidad-generativa-para-crear-imagenes-impactantes-anuncios-campanas-publicitarias-marcas-anunciantes. Accessed 22 October 2023.

Ransbotham, Sam; Kiron, David; Gerbert, Phillip, and Reeves, Martin (September 2017). Reshaping Business with Artificial Intelligence. MIT Sloan Management Review and The Boston Consulting Group. https://web-assets.bcg.com/img-src/Reshaping%20Business%20with%20Artificial%20Intelligence_tcm9-177882.pdf. Accessed 14 October 2023.

Reuters Fact Check (16 June 2023). Fact Check- El anuncio de Coca-Cola ‘Masterpiece’ no fue ‘hecho 100%’ por inteligencia artificial. https://www.reuters.com/article/factcheck-cocacola-ai-idUSL1N3883SL. Accessed 20 October 2023.

Ruiz Cano, Diego (30 October 2023). Amazon presenta una herramienta de IA generativa para anunciantes. Programmatic Spain. https://www.programaticaly.com/portada/amazon-presenta-herramienta-ia-generativa-imagenes-anunciantes. Accessed 2 November 2023.

Russell, Stuart and Norvig, Peter (2020). Artificial intelligence: A modern approach. Pearson. 

The Economist (1 July 2022). How a computer designed this week’s cover. The Economist. https://www.economist.com/news/2022/ 06/11/how-a-computer-designed-this-weeks-cover. Accessed 11 October 2023.

Tucher, Andie (2017). I believe in faking: The dilemma of photographic realism at the dawn of photojournalism. Photography and Culture, 10(3), 195-214. https://doi.org/10.1080/17514517.2017.1322397

Villafañe, Justo and Mínguez, Norberto (2006). Principios de la teoría general de la imagen. Ediciones Pirámide.

Wacom [@wacom] (9 January 2024) A message from the Wacom Team [Tweet]. X. https://twitter.com/wacom/status/17448
14055137919388
.

Wimmer, Roger and Dominick, Joseph (1996). La investigación científica de los medios de comunicación. Una introducción a sus métodos. Bosch. 

Xian, Xuelin (2021). Psychological factors in consumer acceptance of artificial intelligence in leisure economy: A structural equation model. Journal of Internet Technology, 22(3), 697-705.

Appendix 1. Analyzed Products

Table 1

No.

ID

BRAND

DATE

PRODUCT

SECTOR

The next Rembrandt

ING

April 2016

Advertising campaign

Banking

Is it ok to talk about AI in Cannes now?

Publicis Groupe

April 2016

Advertising campaign

Advertising

Paintress

The Streechers

March 2021

Video clip

Entertainment (music)

Meet the world’s first artificial intelligence magazine cover (No. the A.I. issue)

Cosmopolitan

July 2022

Magazine

Publishing (trends)

Heinz A.I. Ketchup

Heinz Ketchup

July 2022

Advertising campaign

Food

Age of illusion

Die Antwood

September 2022

Video clip

Entertainment (music)

Entergalactic theme

Kid Cudi

September 2022

Video clip

Entertainment (music)

Romance between human & an artificial intelligence (No. 130)

Yorokobu

September 2022

Magazine

Publishing (design)

Artificial intelligence’s new frontier (2 September issue)

The Economist

September 2022

Magazine

Publishing (trends)

La Laitière

Nestlé

September 2022

Advertising campaign

Food

Bad Man

Disturbed

November 2022

Video clip

Entertainment (music)

Untitled (no. 28, 2 versions)

Gràffica

December 2022

Magazine

Publishing (design)

Unbottling Martini

Martini

January 2023

Advertising campaign

Food & drink

Lost

Linkin Park

February 2023

Video clip

Entertainment (music)

Artificial community

Burger King (Colombia)

February 2023

Advertising campaign

Food

Atropos

Periphery

February 2023

Video clip

Entertainment (music)

Welche kreation würdet ihr euch gönnen?

Burger King (Germany)

March 2023

Advertising campaign

Food

AI made this ad so we could have a break

KitKat

March 2023

Advertising campaign

Food

April

Stradivarius

March 2023

Advertising campaign

Fashion (clothing)

Untitled (no. 1893, article ‘The rebellion of the machines’)

Somos (El Comercio magazine)

March 2023

Magazine

Publishing (trends)

Roots. Fashion Meets the AI Revolution

Vogue Singapore.

March 2023

Magazine

Publishing (trends)

Masterpiece

Coca-Cola

March 2023

Advertising campaign

Food

The benefits of using AI in accounting (No. 791)

New Accountant USA

April 2023

Magazine cover

Publishing (economics)

La red

Kinesis

April 2023

Video clip

Entertainment (music)

¿De qué equipo serían las leyendas de nuestra literatura? (Which football club would the most famous Spanish authors in history support? (no. 127)

Panenka Magazine

April 2023

Magazine

Publishing (football)

The French re-connection

Citroën

April 2023

Advertising campaign

Automotive

Generation Bershka

Bershka

May 2023

Advertising campaign

Fashion (fragrance)

Secret invasion

Marvel Studios

June 2023

Series

Entertainment (series)

Chic summer with Lisa Opie (magazine 13 July 2023)

Glamour Bulgaria

July 2023

Magazine

Publishing (trends)

Source: Own elaboration.


  1. 1 The table with the basic data of the pieces that make up the sample can be found in Table 1 (see appendix).

  2. 2 This refers to the main format, as some of the still image products have video versions for some social media.

  3. 3 This format was used to indicate the total number of products that were classified in the different categories indicated.

  4. 4 The Streechers explained that one of the band members, Ihor Melikhov, was fond of programming and used a form of AI he created himself in 2016. Kinesis is about a band from the 1990s for which its creator generated the video 25 years after the song’s release. Both bands explained this on the YouTube platform where the songs are shared.

  5. 5 This product was not included in the sample as the image was not created by the company for their book, but a previously created stock image was used which was taken from an image bank. Not only was it unwittingly created by artificial intelligence, but there was no real intention of using AI by the brand.