Créer des pubs avec son clone d'intelligence artificielle à 20€ : ce que vous cache les marques

The world of content creation is undergoing a massive shift, and as this transcript highlights, high-quality video production is now accessible to anyone with a few dozen dollars and a bit of imagination. The primary message is clear: the barrier to entry for creating professional-grade video content has vanished, thanks to the rapid evolution of artificial intelligence. One of the standout tools mentioned is HeyGen, a well-known platform in the AI video generation space. The speaker demonstrates how they created a digital twin, or avatar, of a person named Alec. Interestingly, they didn't even need Alec to film a specific video for the AI; they simply used a three-minute "rush" or raw footage from a previous advertisement. The AI analyzed Alec’s movements, his lip-syncing, and his voice. While users previously had to link external tools like 11 Labs for voice cloning, these features are now integrated directly into the platform. For less than €20 a month, the software can generate a video of this avatar saying absolutely anything. The level of control is remarkable. You can instruct the avatar to ask viewers to subscribe to a YouTube channel or leave a five-star review on a podcast platform. The AI doesn't just replicate the voice; it synchronizes hand gestures and body language with the script. The speaker notes that the user experience is so intuitive it feels "made for children," allowing users to adjust the speed, volume, intonation, and even the accent of the voice—for instance, making the avatar speak with a Quebecois accent. These generated videos are already being used in live advertising campaigns on platforms like Instagram, often without viewers even realizing they are watching an AI. To stay ahead of these technological leaps, the speaker recommends using X (formerly Twitter) as a primary source for "tech watch." By following accounts like OpenAI or ChatGPT, anyone can stay updated on the latest AI breakthroughs in just a few minutes of scrolling. The transcript mentions seeing 100% AI-generated advertisements for major brands like L'Oréal or clothing companies where the models in the videos don't actually exist. This trend was particularly visible during the Christmas season, where numerous brands produced their entire holiday spots—from beverages to gift advertisements—using AI, often without explicitly communicating it to the public. For those who find the technical side of AI—specifically "prompting"—intimidating, the speaker introduces a concept called "reverse engineering." Instead of struggling to write the perfect command, the speaker uses a conversational AI like Google’s Gemini to write the prompt for him. By giving a simple voice command, such as "Create a video of Alec walking in the streets of Paris," and asking Gemini to generate the most effective technical prompt for high-end models like Sora 2 Pro or Google VEO3, the user can achieve hyper-realistic results without being a technical expert. Cost management is another key topic. Rather than paying for multiple individual AI subscriptions, which can become expensive, the speaker suggests using aggregator platforms like Artlist.io or Higgsfield.ai. These services allow users to pay for credits and access various AI models in one place. Using these tools, the speaker has generated "man-on-the-street" style interviews set in Paris that look indistinguishable from real footage. These videos can be customized by selecting the aspect ratio (such as 16:9 or 9:16 for social media), resolution (up to 4K), and duration. Ultimately, these tools are being utilized by traditional businesses, such as companies selling pillows or makeup, to create targeted advertisements that speak directly to their audience's needs. The speaker concludes by suggesting a simple but effective strategy: look at what successful competitors are doing in the US market, replicate those ideas for a French-speaking audience, and use AI to generate the necessary video assets. With the technical hurdles removed, the only remaining limit is the creator's own creativity.