AI Based Video Advertisement Generation Using AI

Abstract

The fast growth of Artificial Intelligence (AI) has opened new possibilities in digital marketing, media production, and content authenticity checking. This paper presents a single, end-to-end system that tackles three connected challenges: automated video advertisement creation, AI-assisted logo building, and multi-modal fake video detection. The proposed architecture combines a Transformer-based Natural Language Processing (NLP) engine for marketing script writing, Stable Diffusion models for scene-level visual creation, and an FFmpeg-powered pipeline for video rendering. For verification, the system uses a hybrid deep learning and heuristic approach—combining EfficientNet-B4 with Bidirectional LSTM for deepfake detection, CLIP zero-shot classification for brand logo checking, and YOLOv8 for real-time logo detection. Heuristic methods including Error Level Analysis (ELA), Fast Fourier Transform (FFT) frequency analysis, noise profiling, and Structural Similarity Index (SSIM) support the neural pipeline to flag AI-made artifacts. The system runs through a FastAPI backend and a React/Next.js frontend, supporting real-time advertisement creation and authenticity reporting. Test results show that this combined approach produces high-quality advertisement output while reliably detecting manipulated or AI-generated content. This work highlights the potential and ethical need to build generation and verification tools together.

Introduction

The text discusses the growing impact of Artificial Intelligence (AI) in digital advertising, particularly in automated video advertisement creation and deepfake detection. Traditionally, producing video advertisements required significant budgets, creative teams, and long production times. Modern AI technologies now enable scalable, automated, and personalized advertisement generation using simple product information.

The paper highlights the role of advanced AI models such as:

• Generative Adversarial Networks (GANs) for synthetic media generation,
• Diffusion-based systems like OpenAI Sora for high-quality video synthesis,
• Large Language Models (LLMs) for generating marketing scripts, slogans, and scene descriptions.

While these technologies improve content creation, they also increase risks associated with fake media, including deepfake advertisements, fake endorsements, and manipulated brand logos. Such content threatens brand reputation, consumer trust, and information reliability. This has led to growing research in deepfake detection using deep learning and signal-processing techniques.

To address both advertisement generation and verification, the paper proposes a fully integrated AI system with two major components:

1. AI-Based Advertisement Generation Pipeline
   • Automatically creates video advertisements from product data.
   • Includes script generation, image synthesis, voice narration, and video assembly.
   • Uses technologies such as LLMs, Stable Diffusion, gTTS, and FFmpeg.
2. Fake Detection and Brand Authenticity Suite
   • Detects whether advertisements are real, AI-generated, or deepfake.
   • Combines deep learning models like EfficientNet-B4 and Bi-LSTM with signal-processing methods such as:
     • Error Level Analysis (ELA),
     • Fast Fourier Transform (FFT),
     • Structural Similarity Index (SSIM),
     • Noise analysis.
   • Includes zero-shot logo verification using OpenAI’s CLIP and YOLOv8 without requiring brand-specific training data.

The system is deployed using modern web technologies including FastAPI, React, and Next.js for real-time advertisement generation and verification.

The literature survey reviews prior work in three main areas:

• AI-driven video generation,
• Deepfake detection,
• Brand/logo verification.

Previous studies show that GANs, diffusion models, and transformer-based architectures can generate highly realistic and personalized content. Research also indicates that deepfake generation capabilities are advancing faster than detection methods, making verification systems increasingly important. EfficientNet-based architectures and Bi-LSTM temporal analysis are identified as strong approaches for detecting synthetic media.

Conclusion

This paper has presented a combined, end-to-end system for AI-based video advertisement generation, logo synthesis, and fake video detection. By bringing together the complementary strengths of Transformer-based NLP, Stable Diffusion visual synthesis, algorithmic audio generation, FFmpeg video rendering, EfficientNet-B4 + Bi-LSTM deepfake detection, CLIP zero-shot logo verification, and signal-processing heuristics, the system shows that generation and authenticity verification can and should be built together as two parts of the same infrastructure. The work addresses a real and growing challenge in digital marketing: as AI makes it steadily easier to produce convincing synthetic advertisements, the tools needed to verify their authenticity become equally important. This system is a practical step toward making both capabilities available in a single deployable platform. The results confirm the approach works well, with strong performance across both advertisement quality metrics and deepfake detection benchmarks, while acknowledging that the ongoing contest between generation and detection remains an open research problem. The modular architecture ensures that individual components can be upgraded as better models become available, making the system adaptable to the rapid pace of progress in generative AI.

References

[1] Kim, J., & Hwang, S.-H. (2025). Personalized Video Ad Creation via Generative Adversarial Networks. International Journal of Creative Research Thoughts (IJCRT), 13(6), a625–a628. ISSN: 2320-2882. [2] Xie, W., Hu, A., Xie, Q., Chen, J., Wan, R., & Liu, Y. (2025). Bibliometric analysis and review of AI-based video generation: research dynamics and application trends (2020–2025). Discover Computing, 28, 130. https://doi.org/10.1007/s10791-025-09628-9 [3] Goodfellow, I., Pouget-Abadie, J., Mirza, M., et al. (2014). Generative Adversarial Nets. Advances in Neural Information Processing Systems (NeurIPS), 27. [4] OpenAI. (2024). Sora: Creating video from text. OpenAI Technical Report. [Online]. Available: https://openai.com/sora [5] Bengesi, S., El-Sayed, H., Houkpati, Y., & Irungu, J. (2023). Advancements in Generative AI: A Comprehensive Review of GANs, GPT, Autoencoders, Diffusion Model, and Transformers. arXiv preprint arXiv:2311.10242. [6] Kavinkumar, M., & Kanishka, R. G. (2025). How Generative AI is shaping the Future of Content. CORE Open Access Publication. Available: https://core.ac.uk/download/648319648.pdf [7] Veerasamy, N., & Pieterse, H. (2023). Rising above misinformation and deepfakes. Council for Scientific & Industrial Research (CSIR). Academic Conference Limited. [8] Sharma, I., Jain, K., & Behl, A. (2024). Examining the motivations of sharing political deepfake videos: the role of political brand hate and moral disengagement. Deepfake Detection Model Overview. Journal Reference [See Discover Computing, 2025, Table 8]. [9] Shen, X., Xiao, X., Zhang, R., & Zhao, S. (2024). PMG: Personalized Multimodal Generation with Large Language Models. arXiv preprint arXiv:2404.08677. [10] Hao, Y., Hui, P., Kangasharju, J., & Liu, Y. (2024). A Survey on Generative AI and LLM for Video Generation, Understanding, and Streaming. arXiv preprint arXiv:2404.16038. [11] Ercan, H. D., Tanr?verdi, N. S., & Ta?k?n, N. (2024). A Systematic literature review for Artificial Intelligence in Advertising. CORE Open Access. Available: https://core.ac.uk/download/616566925.pdf [12] OZCAN, A. K. (2024). Exploring the Perceptual Boundaries of AI-Generated Content in Modern Content Marketing. CORE Open Access Publication. Available: https://core.ac.uk/download/620850908.pdf [13] Kowalczyk, P., Röder, M., & Thiese, F. (2023). Nudging Creativity in Digital Marketing with Generative Artificial Intelligence: Opportunities and Limitations. CORE Open Access. Available: https://core.ac.uk/download/567667169.pdf [14] Du, H., Han, Z., Jamalipour, A., & Kang, J. (2023). Unleashing the Power of Edge-Cloud Generative AI in Mobile Networks: A Survey of AIGC Services. arXiv preprint arXiv:2303.16129. [15] Kweon, I. S., Zhang, C., Zhang, C., et al. (2023). Text-to-image Diffusion Model in Generative AI: A Survey. arXiv preprint arXiv:2303.07909. [16] Almeida, S., & Ivanov, S. (2024). Generative AI in Hotel Marketing – A Reality Check. CORE Open Access. Available: https://core.ac.uk/download/613207532.pdf

Copyright

Copyright © 2026 Dr. Prerna N. Khairnar, Ms. Pranjal S. Elamame, Nikita K. Toche, Ms. Anuradha B. Pandav, Ms. Namarata N. Magar, Ms. Rushali A. Dhamale. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.