The landscape of Retail, Marketing and Advertising in 2025

This year is the momentum of AI Agents, but there are also other significant advancements impacting the Retail and Marketing and Advertising work in general:

Virtual-try Ons becoming a popular tool to improve conversions or simply as a further?touchpoint to contribute to purchase decisions in omnichannel retail. Google VTO is a tool that uses artificial intelligence (AI) to show how clothing looks on real people based on their size and body shape. You can find the VTO tool on qualifying listings on Google Shopping. Do you want to incorporate the option on your website? There are several startups that allow you to make an outfit check with a single pic uploaded like Wear It, Dualme, Aicado.ai, SilkPLM, Openart https://openart.ai/generator/clothing and many more.. From eyewear brands, to more and more retailers are adding virtual try-ons on their e-store like Target, Kiko Milano or L'Oréal Paris for ie. And since there is a significant footfall instore demanding engaging experiences, the natural omnichannel extension is adding digital touchpoints in-store, such as the Dolce & Gabbana Beauty's smart XR mirrors in the physical space at Saks or the smart fitting rooms.

?Whatsapp + AI is emerging as an additional touchpoint for collecting consumers' big data in retail operations, sales and marketing for the team. Please note, as many think, to resonate deeply in the opportunity offered by this channel: before using Whatsapp to send commercial offers, you need a specific authorization by consumers and prospects. That's why Meta has recently introduced its new Customer Data Platform known as the Signals Gateway. Data is hosted in Amazon Web Services and, in theory, Meta can’t see it. Through its new Signals Gateway CDP Meta delivers enhanced customer behavior insights across Facebook, Instagram, and WhatsApp platforms for businesses. Signals Gateway enables advertisers to better handle customer data through improved collection and utilization methods. With this launch, Meta enters officially in the competition within the expanding CDP market. This launch consists on a unified business platform to monitor customer interactions while analyzing trends and refining marketing strategies. My concerns are related to the usage by brands of those data to generate market trends or simply contents, as we already?saw the?flat creative curve or - the as already seen effect?- of creative outputs in brands' new product?development and contents. Think about the overtourism generated by videos of the usual locations on TikTok and IG, or the proliferation of the similar contents and product shapes in FMCG or luxury.

AI, Social Media trends and the real collections in-store: the harsh reality of market fit

What am I referring to specifically? The sophisticated algorithms like Convolutional Neural Networks sift through mountains of visual data from platforms like Instagram and TikTok, identifying emerging trends in colour palettes, textures, and silhouettes with uncanny accuracy. Recurrent Neural Networks? analyse the ever-evolving language of social media, deciphering the nuanced meanings behind hashtags, captions, and user comments to understand the evolving consumer mindset. Unfortunately, those algorithms?can't perceive when the curve of desirability of certain goods reaches the peak or how many potential customers (TAM) a brand can have in its markets. This data-driven approach empowers brands to anticipate market demands, swiftly adapt to changing trends, and even identify niche markets with remarkable precision. But this data-driven revolution raises critical questions. Can algorithms truly capture the essence of fashion the emotional resonance, the cultural undercurrents, the sheer human creativity that drives iconic designs? Is there a risk of homogenization, where AI-driven insights lead to a predictable, data-optimized aesthetic that lacks the unique spark of individuality?

The Mass Market of Generated-AI Avatars

AI to support brand building and the creation of content and PR in marketing and e-commerce

Generative AI has been a game changer for many digital marketers, content creators, gaming publishers and small medium brands which didn't previously have much resources.

With the AI it is now possible to generate PDF, Decks, Infographics, but the real innovation comes from the mass availability of AI solutions to make virtual shootings like MakeUGC.ai, Social Clone, Poolday.ai, Creatify.ai, Arcads.ai, Google's Vlogger, or videos with virtual models or actors.?

The ease of producing localized and personalised contents is opening much more opportunities for brands selling in different?geographical?locations.?

Although, generated images sometimes still have limitations and shouldn't be published without supervision. There is also a main ethical impact that needs to be justified and probably addressed. For ie. Coca Cola's latest XMas Gen-ai commercial faced a lot of controversies by industry professionals for its human impact on labor market sustainability last year. And finally, the?public opinion also raises concerns about the potential for misuse in creating deep fakes and spreading disinformation.

What's hot in the generative-ai video production scene?

ByteDance, the company behind TikTok, has recently unveiled two significant advancements in generative AI: Goku AI and OmniHuman-1.

These models represent major leaps in AI-driven content creation, particularly in the realm of video generation. ? What we know about Goku AI: it is an internal open-source LLM by ByteDance. It's the AI latest model for image and video generation, designed to compete with Tongyi series?by Alibaba, or Baidu's Wenxin Yiyan or Tencent's?HunyuanVideo and the Westerners OpenAI's Sora or VEO2 by Google.

?What are the features offered by Goku?

• ?Marketing Avatar Creation from Text: Converts text into hyper-realistic human videos with natural expressive movements.
• ?Product Image to Video Conversion: Automatically generates promotional clips from product images.
• ?Realistic Human-Product Interaction: Helps create realistic videos showcasing product usage by humans. Its Rectified Flow Transformer (RFT) technique, which allows for smoother and more realistic video generation. The creation process is gradual, refining each frame step by step to ensure seamless motion transitions within the video. This technology also incorporates a dynamical rectified flow algorithm, ensuring fluid motion, reducing visual noise, and enhancing image sharpness. With this approach, GOKU AI can generate clearer and more realistic videos, making it a preferred choice for the creative and marketing industries.
• ?Optimized Video Ads: Enables high-quality ad video production at a cost claimed to be 100 times cheaper than traditional methods.

While nothing is confirmed, it’s highly likely that ByteDance will integrate Goku AI into TikTok to enhance content recommendations and video creation tools. From the innovation perspective, this is going to bring a wave of engagement by users and brands. However, I am also very curious to see how the company will handle the risks to avoid the proliferation of critical contents.

?What's the other tech?of Byte Dance??

OmniHuman-1 allows users to generate highly realistic human videos from a single image and motion inputs, such as audio or video. Using advanced motion conditioning techniques, the platform creates lifelike avatars that display accurate facial expressions, gestures, and lip-syncing, all synchronized with the provided audio or video. OmniHuman-1 can process various types of input, from portraits to full-body images, and can even produce high-quality video content from minimal data like just audio. In addition to human figures, it can animate cartoons, animals, and objects, making it an ideal tool for diverse applications in virtual reality, entertainment, and education. This AI model offers a groundbreaking approach to transforming static images into dynamic video content with impressive realism and versatility.

Main Features:?

• ?Single Image Input: Can create realistic videos from just one still image3
• ?Versatile Output: Supports various scenarios including talking, gesturing, singing, and playing instruments
• ?Extensive Training: Trained on over 18,700 hours of human-related data
• ?Multimodal Conditioning: Incorporates text, audio, and body movements in its training process
• ?High Realism: Generates extremely realistic human videos, potentially outperforming existing methods
• ?Flexible Input Handling: Supports images of any aspect ratio, including portraits, half-body, or full-body images

Google developed an AI project called VideoPoet, a simple modeling technique, can convert any large language model or autoregressive model into a high quality video generator. It is composed of a few components. The autoregressive model learns from video, image, text, and audio modalities in order to predict the next audio or video token in the sequence. The LLM training framework introduces a mixture of multimodal generative objectives, including text to video, text to image, image-to video, video frame continuation and inpainting/outpainting, styled video, and video-to audio. Moreover, these tasks can be combined to provide additional zero-shot capabilities. This simple recipe shows how language models can edit and synthesize videos with a high level of temporal consistency.

What is it possible to do with Google's Videopoet?

• Creating short video clips from text prompts or images?
• Generating realistic scenes and abstract art
• Weaving visual narratives by changing prompts over time
• Applying various camera motions and visual effects
• Potential use in filmmaking, advertising, and creative industries

Google's Veo2 is more recent and has probably a better usage in the cinematography?industry.

Meta has also launched its own AI, called Movie Gen.

Movie Gen is an advanced AI-powered media generation suite introduced in October 2024. It consists of several foundation models designed to create high-quality video and audio content from text prompts.?

Video Generation:?

• ?create high-definition videos up to 16 seconds long in 1080p resolution35.
• ?generate videos from text prompts, turn still images into videos, and edit existing footage7.
• ?create personalized videos using actual photos of individuals59.

Audio Generation

• ?produces synchronized audio for videos, including ambient sounds, sound effects, and instrumental background music34.
• ?generate audio up to 45 seconds long and uses an extension technique for longer videos4.

Editing Capabilities

• ?allows users to modify existing videos or images using text inputs9.
• ?users can alter styles, transitions, or add new elements to footage1.

Meta claims that Movie Gen outperforms competitive models in overall video quality3.

While Movie Gen shows significant potential, it is currently not available for public use. Meta has stated that high costs and lengthy generation times are barriers to its immediate release as a product.

Basically, while I hope that you will enjoy testing your personal abilities mastering?genAI tools, I am also thinking about human nature's risks and given the ease of content creation these days, brands should also invest in solutions to detect and prevent AI misuse and hence preserve their brand reputation.?

What do you think? Let's chat!!!

#TopRetailExpert