Google’s Veo 3: AI Video Realism Leaps Forward, Sparks Debate on Authenticity

Google is pushing the boundaries of AI video generation with Veo 3, its newest model capable of creating realistic 8-second video clips with synchronized sound effects and audio dialogue. This significant advancement raises questions about distinguishing between authentic and AI-generated media.

Following its debut at I/O 2025, Google is expanding access to Veo 3. Select Google AI Pro subscribers can now generate 10 Veo 3 videos through the Gemini app on Android and iOS. This offering allows users to bring ideas to life with sound, background noise, and dialogue, marking a significant step forward in AI-driven content creation.

According to Ars Technica, Veo 3 videos cost 150 credits per generation, allowing users on the Google AI Ultra plan roughly 83 videos before needing more credits, which are priced at $0.01 each. Is the price worth it? Testing the technology involved various prompts to see what this capable tech can do.

Veo 3 utilizes **diffusion technology**, similar to image generators like Stable Diffusion, refining random noise into coherent video based on text prompts. It comprises a series of AI models, including a large language model (LLM) to interpret prompts, a video diffusion model to create visuals, and an audio generation model to add sound, wrote reporter Benj Edwards.

To combat misuse, Google is implementing its proprietary SynthID watermarking technology, embedding invisible markers into frames. A visible watermark is also being added to all videos, with exceptions for Ultra members in Flow. All generated content from Google has an invisible SynthID watermark, said Google.

Reports from Ars Technica indicate that while Veo 3 can generate diverse content, occasional glitches occur. Spaghetti makes crunching sounds, dialogue might come from the wrong character's mouth, and garbled subtitles are common. Despite imperfections, Veo 3 showcases improvements in video synthesis quality and coherency over models from OpenAI, Runway, and others.

Edwards highlights that Veo 3 could contribute to casual media deception and that the ability to easily generate convincing video raises concerns. The rise of accessible AI video creation democratizes deception, requiring skepticism in media and an emphasis on the source's credibility.

While some find Veo 3 to show the birth of media deception, others see the launch as its mass democratization. What are your thoughts on Veo 3's potential impact? Share your perspective in the comments below.

• AI video
• Google AI
• Veo 3

X talks about this news