Google’s best Gemini showcase video turns out to be fake

When Google first announced Gemini, we were all excited for things to come but when the company released a video showcasing Gemini in action, we were all in shock and awe because of the fact that the video showed the AI model was able to not only interact with the video’s drawings in real-time but also answer questions that were posed by the narrator at that time in the video. This is also the reason why we went gaga over Gemini’s capabilities but it turns out that the video released by Google was faked. This is because all of those things did not happen in real-time and rather the inputs were given to Gemini as stills from the video and the audio was added later on.

This does not mean that Gemini is bad but it also tells you that the model is not as good as we expected after seeing the video. It is reported that “We created the demo by capturing footage in order to test Gemini’s capabilities on a wide range of challenges. Then we prompted Gemini using still image frames from the footage, and prompting via text”. As we mentioned earlier, this is still impressive that a LLM can identify the images and its features and answer questions based on the same. But the fact that they showed us the video gave us the impression that Gemini can identify drawings in real-time which is not the case.

One thing that got most of us is that we didn’t watch the video on YouTube where Google put out the description but rather saw it on social media platforms like X or Facebook where the caption did not have any description. It is reported that “In its YouTube description Google also admits the video is edited for latency – which makes it look like the model is responding more quickly than it is”. The speed at which Gemini gave the responses also impressed everyone but now that is revealed that these were faked, we feel betrayed at the marketing done by Google and the hype around Gemini seems to have faded away, at least for me.