AI video

Figure 1. Sign up to try VideoFX.

I was reading a Google webpage about Imagen 3 and I followed a link to another webpage about SynthID and that had a link to a Veo webpage which had multiple links to VideoFX (see Figure 1),

Figure 2. Test Kitchen.

On October 17th 2025, I filled out an online form to get on the wait list for VideoFX.

At Google's "AI test kitchen" they have several different types of software (see Figure 2). I had previously tried ImageFX once, about a month ago and found it to be poorly documented and not of much use. I gave ImageFX this text prompt: "a jungle of alien plant-like life forms, the plant-creatures have some anatomical features like humans, the plant life is from a distant exoplanet and mostly covered with green leaves, the alien plant life has appendages that look like the tentacles". The software generated the following:

Figure 3. ImageFX user interface, December 2024.

Figure 4. Prompt options.

Figure 5. "35mm film".

The yellow drop-down menus that ImageFX inserts into user-supplied text prompts have alternative choices as shown in Figure 4.

I clicked on the option for "35mm film" and ImageFX made 4 new images, one of which is shown in Figure 5.

I clicked on the "edit image" button and tried to get ImageFX to convert the "plant creature" into a plantimal with a humanoid face...  

Figure 6. In edit mode; selecting a leaf and asking ImageFX to insert a face.

Figure 7. Lame software.

The software processed my request (Figure 6) for several minutes and made 4 new images that looked identical to the original image.

Knowing that Gemini will refuse to make images of humans, I tried asking for a cat's head, and again there was almost no perceptible change in the image.

It really baffles me as to why companies like Google put software like this on the internet for people to use without documentation and with such a useless user interface.

I suppose there is some "trick" to getting this ImageFX software to work, but Google can't be bothered to tell users what the trick is. I used this text prompt: "Google's ImageFx is the lamest software in the world" and got the image shown in Figure 7.

Figure 8. Mostly generated by ImageFX in April 2025. Plant person by Mr. Wombo.

At some point towards the end of 2024, the ImageFX user interface was altered. The "edit" button was removed and drop-down menus were no longer generated in text prompts. About then, I began using Google's Whisk. At the start of April, I was given "early access" to a new feature in Whisk, the ability to generate short videos from Whisk-generated images.

Figure 9. Generated by Whisk

The image shown in Figure 9 (above) was generated by Whisk. Below is an example of the kind of short video that Whisk can make from images.

Here is the text prompt that was used to create this video (v1, above): "The woman, the man and the flaming face are all moving and talking at the same time. The yellow flames flicker continuously and violently while the stylized face in the flame opens and closes its mouth as if the face is speaking to the man and the woman. The man pushes a probe into the flame and speaks to the face that is in the flame. The woman quickly walks across the room and reaches towards the flames with both hands".

In my imagination, this video (above) could depict a rather frustrated Tyhry who is trying to communicate with an artificial life-form that is inside the orange plasma. Whisk seemed reluctant to actually animate the "plasma creature". The video shown below (v2) was the one video out of a dozen attempts that resulted in significant movement of the "plasma creature", making it look like the creature is speaking.

The video below (v3) was the most unusual of these "plasma creature" videos.

The next video, below (v4), was the best example of a Whisk-generated video in which the woman walks towards the plasma and then reaches out towards the plasma.

In the next video, below (v5), it seems like the man finally gives up trying to elicit a reaction from the plasma creature.

In the next video, below (v6), the man seems to run away and leave the room.

In the next video (v7), Tyhry seems to stumble then points at the plasma creature and then there seems to be a flash of something (nanites) emitted by her finger.

In the next video (v8), it looks like both the man and Tyhry are speaking as he tries to get a reaction from the plasma creature. He is using a "magic wand" that seems to emit blue flame.

In the next video, below (v9), the device held by the man is emitting sparks and flame. Tyhry is also able to launch something at the plasma creature (nanites?). Also, in this video, the camera seems to move or pan to the side.

The next video (v10) might be my favorite one. Tyhry actually runs a significant distance. Strangely, in some frames of the video she seems to be barefoot while in other frames she seems to be wearing heels. Also, another plasma reactor mysteriously appears at the left hand side of the scene. Unfortunately, there is only one small movement of the plasma creature's face.

In the next video (v11), Tyhry is definitely wearing high heels. She seems to grow impatient with how the man is prodding at the plasma. As she approaches, some flame seems to move from the man's device towards Tyhry's hand.

The next video (v12) is unique in that another woman seems to walk through the scene, close to the camera.

In the next video (v13), Tyhry backs entirely out of the scene.

The next video (v14) is the final one from this series that I am including in this blog post. This is an example of a video that Whisk generated without any additional instructions from me. Whisk created a zoomed in view of Tyhry standing close to the sparking plasma.

In this first session of generating video with Whisk, I felt like there must be some secret instruction system for explaining to the AI how to make a video. My results seemed largely random, except for an occasional snippet that seemed to be generated in response to my instructions.

Next: more videos generated by Whisk.

Visit the Gallery of Movies, Book and Magazine Covers