Je crée une vidéo Deepfake avec l'IA en 5 minutes (FLIPPANT)

Je crée une vidéo Deepfake avec l'IA en 5 minutes (FLIPPANT)

by Ludo Salenne

Applied Sciences & Engineering AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence
🎙 Ludo Salenne 👥 267K 📅 April 28, 2024 ⏱ 20 min 👁 105K 📄 tutorial 🧭 2026-08-21
Available in: English (current) Français

Keywords

deepfakeAI clonevoice cloningavatarC2PA

Summary

The video is a tutorial by Ludo Salenne demonstrating how to create a deepfake video using AI tools, specifically ElevenLabs for voice cloning and HeyGen for video avatars. The creator begins by highlighting the dangers of deepfakes, citing a real scam in Hong Kong where a finance employee was tricked into transferring 24 million euros via a deepfake video call. He also mentions the potential for misuse in elections and geopolitical conflicts. After the warning, he provides a step-by-step guide: first, using ElevenLabs to clone a voice from a short audio sample; then, using HeyGen to create an AI avatar from a webcam video. He shows how to combine the cloned voice with the avatar for a realistic result. The video also covers additional features like importing a custom voice via API and using ChatGPT for script generation. The creator discusses future developments like Microsoft’s VASA-1, which can animate still images, and emphasizes the importance of detecting deepfakes. He offers tips such as reverse image search and looking for visual inconsistencies. He also mentions the C2PA standard for content provenance and labeling by social media platforms. The video concludes with a demonstration of the final deepfake and a reminder to use such technology ethically.

204 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video provides valuable, practical information on creating deepfakes, with clear demonstrations and a balanced discussion of risks and opportunities. The argumentation is solid, as the creator supports his points with real-world examples (e.g., the Hong Kong scam, political deepfakes) and references to ongoing developments like VASA-1 and C2PA. The tutorial is well-structured, and the creator’s enthusiasm is evident, but the argumentation could be strengthened by citing more specific sources for the claims made.

Scientific Rigor, Source Quality, Title Accuracy

The video demonstrates scientific rigor by acknowledging the dangers of deepfakes and providing detection methods. The sources cited in the description include official pages for VASA-1 and Sora, which are relevant. However, the creator does not always provide detailed citations for statistics or incidents, relying instead on anecdotal evidence. The title accurately reflects the content, and the video’s structure is coherent, with clear chapters and a logical flow.

157 words

Title / Content Match

The title accurately reflects the content: the creator demonstrates creating a deepfake video in about 5 minutes, and the result is indeed impressive.

Quality & Reliability

7/10

The video provides a practical, step-by-step tutorial on creating deepfakes using ElevenLabs and HeyGen, with clear demonstrations and a discussion of risks and detection methods. The information is accurate and up-to-date, but the presenter's expertise is not formally verified, and some claims (e.g., the Hong Kong scam) are presented without detailed sourcing.

Chapters

Cited Sources

  • HeyGen — Tool used to create AI video avatars.
  • OpenAI Sora — AI video generation model mentioned as an example of advanced deepfake technology.
  • Microsoft VASA-1 — Research project that animates still images, presented as a future deepfake tool.
  • Formation ChatGPT — Promotional link for the creator's own training, not directly related to the video's content.

Concurring Sources

  • Microsoft VASA-1 — Supports the claim that AI can animate still images, as discussed in the video.
  • OpenAI Sora — Supports the claim that AI-generated videos are becoming more realistic.

External References

Contribution & Novelties

The video offers a practical, up-to-date tutorial on creating deepfakes with accessible tools, highlighting both the ease and the risks. It contributes to public awareness of deepfake technology and its potential for misuse, while also providing a hands-on guide for legitimate uses like content creation. The inclusion of detection tips and the mention of C2PA add value for viewers concerned about misinformation.

Pour aller plus loin :

108 words

Radar Profile

The radar profile shows high scores in information quantity and quality, with moderate technical depth and reliability. This indicates a well-rounded tutorial that is informative and practical, though not deeply technical or heavily sourced.

Reliability 7/10