On returning from the family gathering in Buxton and during preparations for next year's trip, I wanted to make a clip about the process of choosing the destination. I tried to make a still image to use as a seed for a video clip.
I gave ChatGPT this prompt to get this image:
A woman astronaut is looking out of the panoramic window of the ISS as the planet slides past. Hyper-realistic.
It's a beautiful image but why is she wearing a helmet inside the craft, and why is the window so much smaller than the real panoramic window of the ISS, which looks like this real photo below? (See also this article) and the first minute of this video (it's slow so play it at double speed.
I asked ChatGPT to generate a new image with this prompt:
She is in the spacecraft so should not be wearing her helmet. Also, please include some electronic displays with colored lights in the background.
ChatGPT did not respond (offended, perhaps?) so I tried this:
Can you remove the helmet from the woman in the last image and include some electronic panels with colored lights in the background?
Still rather dark, a bit glum, and why isn't her hair floating around in the absence of gravity? And she's not even looking out of the window! Bored of the view?
I gave ChatGPT this prompt:
The panoramic window of the ISS is much bigger and has many facets, like a cut diamond. Can you adjust the window and also make the woman turn her head to look out of the window?
That's a step forward but still rather dark and the window is still too small. You might think the US flag is reversed but no - on a moving object the stars have to lead, otherwise it would symbolise a retreat. You don't believe me? Check out the flag on the space shuttle:
Next I gave ChatGPT this prompt:
That's better. Can you pull the camera back, make the window even bigger while keeping its shape, and show more of the spacecraft? Her hair should be shorter and floating since there is no gravity.
That's good - I like the floating hair. But the window is still too small and the image is too dark...
I make one final attempt:
Can you make the image in 16:9 format, have more light on the walls of the spacecraft and replace the US flag with a UK flag?
The result is below.
Oops - her hair has become shorter and is no longer floating, and the eye and mouth are a bit distorted. The image is in 3:2 format and not the 16:9 format that I requested. But at least there is a nicer flag. By this time I had used up my free ChatGPT image generation allowance for the day without getting a really good image...
... so I thought I'd try Grok instead. I used this prompt:
Can you make an image of a woman astronaut who is looking out of the large, multi-faceted panoramic window of the ISS as the planet slides past. She is not wearing a helmet. Her mid-length hair is floating in the absence of gravity. The walls are lined with display panels. Hyper-realistic.
I got these two images, (Grok generates two images at a time) both too dark and with unsatisfactory windows.
In the first image she's not even looking out of the window. I tried to fix those problems (images not shown), but wasn't very pleased with the result so gave up trying to make a still seed image and decided to just use a prompt to make the video clip with Veo 3. This is the script I used.
We see the earth slipping by through the large, multi-faceted (like a diamond) panoramic window of the International Space Station. The camera turns and we see a close up of a female astronaut who is looking through the window at the earth. She is not wearing a helmet and her mid-length hair is floating in the absence of gravity. There is the faint hum of electronic devices and there are panels with colored flashing lights in the background. There is a beep and a voice is heard over the PA system, with some static: Hello, Wardlings. How's the accommodation? She turns towards the camera and replies: Nice views but we wanted a proper kitchen and en suite bathrooms. Hyper realistic cinematography. No subtitles.
I had to shorten the dialog I had intended to use in order to fit the eight second limit. I didn't need to specify the 16:9 format because Veo 3 currently always uses that format. But actually the video has black bands top and bottom for reasons I don't know. Despite me specifically requesting no subtitles, Veo has put some in, and mangled them - a typical Veo problem currently. I didn't ask for an Asian Wardling, but I didn't specify Caucasian either. Maybe Veo 3 wasn't given much training data on the nature of Wardlings. There isn't the constant background hum I wanted, nor is there the beep I requested. Instead there is a 'whoosh' sound at the beginning that I didn't ask for. I was hoping for the first voice to sound like communications from Houston but instead it sounds like it's the cameraman speaking - I should have specified that more carefully. Her hair is long rather than mid-length and isn't really floating around.
As you will have guessed if you are still with me, it took a long time to achieve this 8 second video, but what took by far the longest time was finding an affordable way to use Veo 3. I want to use it to make lots of videos but at the usual price of about a dollar per second that's not affordable - I eventually found a good solution: sign up for a trial Google Cloud account and get $300 in free credit with a validity of 90 days. Here is the help I got from Gemini with that.
I was quite pleased with the clip, apart from the subtitles and the fact that I can't use real photos of my siblings or use their real voices, so I'm expecting the Wardlings to return to space very soon...
By the way, the ISS has been described as the most expensive single item ever constructed. As of 2010, the total cost was US$150 billion. For that much money you might expect it to represent perfection in design, with everything neatly organised. Here's the reality:
The Chinese Tiangong space station looks minimalist by comparison: