120K Views in 2 Days: How I Made My AI Time Travel Video I Higgsfield + Claude Code

I made this video fully with AI and it got over a 100,000 views in just 2 days. It's me traveling for nearly 2,000 years of London history. So, in this video, I want to show you exactly how to do it step-by-step. And every single prompt that I used to make it is going to be in a free prompt pack which you can grab from the link in the description. You will need four things for this video.

The first one is a video generation platform. I personally use Higgs Build and you'll need credits on the account. And I will tell you exactly what everything costs as we go. The second is two or three clear photos of your face. Ideally, one should be straight on and one at a 3/4 angle.

It should be in natural light and don't use any filters. Every single shot in this video is going to be built from those photos. So, if they're heavily edited and don't actually look like you, the video will inherit all of those problems. The third one is uh just having an idea of what historical moment you would like to make. And the fourth one is Claude Code which is what is going to tie all of it together.

And the reason why I use Claude Code that's connected to Higgs Build is that once they're connected, I can describe what I want to make in plain English and Claude then writes a long and detailed prompt which is sent to Higgs Build and then Claude downloads the results into a folder on my computer and it keeps track of each take and every job that I give it. So, in step one, we need to connect Higgs Build to Claude Code and the easiest way to do that is open your Claude Code account and click settings and then connectors. Think of connectors like a bridge that lets Claude use another app on your behalf. So, instead of copying a prompt out of Claude and pasting it into Higgs Build, Claude can just talk to Higgs Build directly. So, once you've opened settings, open connectors, find Hicksville, and click connect.

And then a Hicksville sign-in window opens, you log in with your normal account, approve it, and then you're connected. And you can check if it's working simply by asking Claude Code, "What is my Hicksville credit balance?" And if it comes back with a number, you're connected. From here on, Claude can see all the Hicksville models, generate images and videos, upload your reference photos, and pull back your passion generations. One habit I would recommend straight away is making a folder for the project and asking Claude to save everything there.

It just makes things much more organized. Next up, we need to create a character sheet from the photos that you have taken of yourself. So, the front one and the one from the sides and the back. Upload several of these photos into Claude, and then just ask it to create a character sheet. And this will be a photo of yourself from front, back, and the angles.

And this is probably one of the most important steps, because if that sheet doesn't look like you, all of the generations going forward are not going to be great. Next [snorts] up, we're going to create a character sheet, but now it's going to be for every era. This is going to be the thing that will make it look like one person is traveling through time rather than eight random AI women. And it's similar to the character sheet that you have done of yourself, but this one is going to be from each era. And those will be single images of you in that era's costume.

A a big close-up of your face, then a full-length front view and side view, and a back view, all on a plain gray studio background. Because video models are much better at keeping someone consistent when they can see the whole outfit from every angle. So, this sheet becomes a reference for everything else in that scene. I make one sheet per era using GPT image 2.5 high quality with my free face photos uploaded as references.

And this step also is going to save you a lot of money because if you run the video before it generating the costume, it is much more likely that that video is not going to look like you. And generating an image costs like two free credits, while generating the video is 150. So, you can save a lot of money by just getting the photos right and then moving on to generating the videos. So, once that sheet looks right, it's time to animate. And I use C-Dance 2.

5 video model. You can use others, but I've tested a lot of other model, and often times they don't get your face right or the movement looks very odd. As far as I know, C-Dance is the model that takes up the most amount of credits, but also the output that you get is the most high-quality, so it's a balance. Now, I make the videos 8 seconds long with audio switched on. And the references go in a specific order.

The character sheet first, and then two clear photos of my face. And that order tells the model that sheet is the costume, and the photos are the face. The video prompt has a structure, and I reuse it for every era. You'll find it word for word in the prompt pack, but it should have several elements. The first one is a character section, so who I am and what I'm wearing.

Then a historic moment line, which is the date and the event in one sentence. Then the setting, like the place, the weather, the light, the crowd. And then an action section, like the moment happening in front of me and how I react to it. And then I put in the information about how the camera should move, as well as the audio that should be produced. I don't generate those prompts myself.

I describe to Claude what is it that I want, and then it gives me the output being the prompt. Now, it got me quite a few tries in order to get it right. So, I would start first by looking at the prompts that I've created and what elements are in there and make sure that you explain to Claude that these are the things that you want included in the video. And there's a couple of things that I've learned as I was generating these videos. The first one is instructing HeyGen that only I should have the my face and every other person should clearly be a different person.

Because there were a couple of videos that HeyGen generated where the background people look like me and obviously it looked very creepy and unrealistic. Next up, the audio, which is where I list three to five sounds from that moment. Like the sound of the bells for the coronation or the sound of the fire. And I always finish saying that I don't want actual music in the background. That way each clip arrives with its own soundscape and you keep the space free for the music that you can add in later.

And I'll come back to that because sound was one of the elements why the videos actually went viral. So, each of the clips took about five or six minutes to render. So, the process was that I would start one and while I was cooking, I would ask Claude to write the next eras prompt. And by the time the first one landed, the second one was ready to go. And I've set up things this way where I was first asked for the character sheet.

Once I approved the character sheet, only then would I ask for the video to be done. If the character sheet looked odd, then I would ask it to change that first and then do the video. And I would do the videos one by one instead of all of them being done at the same time because it's much easier to spot a problem in one of the videos and fix it instead of then reworking six or seven videos and also using up a lot of the credits. And not everything came out right the first time. So, I want to show you the redos because knowing how to fix a clip cheaply is honestly half of the skill.

One of the biggest mistakes that I've done was making clips where nothing happened. Where the era in the video was right, but the actual historical event was not happening in that video. So it would have been actually hard for people to imagine the event that I'm talking about. And to fix that, I explained to Claude that the historical event should be happening literally in front of my eyes and I should be uh partaking in it. The second mistake was eye line and that's a lesson I properly learned while doing the Paris video.

Because if you're facing the camera with the big event happening behind you, it looks like you're looking the wrong way and immediately the video feels off. So now I asked those to be made in two angles, so that one is over my shoulder looking at the event and the other one is from the front with the event behind the camera, which means that I can express the emotions that I'm feeling. And the way that I make changes to these videos is I just tell Claude in plain English what is wrong. For example, in the 1894 clip, the bridge doesn't lift until the last second. Make it start lifting by second three.

And Claude reuses the same sheet and face references from the notes file and only regenerates that one clip. And once I had all of the videos and I was happy with them, that's when I put the video together and added elements like the calendar flip as well as the name of the event. And I just asked Claude to stitch those clips together and keep each clips own sound and put the date on the screen. And [snorts] Claude wrote a small script that does the whole thing in one go. And it then joins the clips at full quality and draws a hanging calendar card, which is in the corner, and the year in a big, old-style serif font.

And it adds the day and month underneath it, and the name of the event in a handwritten script. And at every cut, the calendar flips, and it has its own sound of the calendar flipping. And on the longer jumps, it riffles through the years in between. I also asked Claude to move the cards to whichever side of the frame I'm not on, so that it never covers my actual face. All right.

And it added the intro title at the start as well. And you don't need to know any code for this. I just typed in roughly this, and I watched it render and gave notes like, for example, make the event label smaller, or the flip of the calendar is too slow. And Claude was able to adjust the script and render it again in a couple of minutes. And because it's a script, I've reused it for other videos as well.

You could do this by hand in CapCut or some other editing software, but making the script is just much easier and quicker. And I want to spend a minute on the sound and the role that I played, because I generally think that is a big part of why this one got past 100,000 views. So, the music that I added was Viva la Vida by Coldplay, and I feel like it fits almost too well this video, because the whole song is sung by king looking back on the empire that he's lost, which is exactly the feeling of watching 2,000 years of history go past you. It's also a song that a lot of people recognize within the first couple of beats. So, before anyone has worked out what they're watching, they already feel something.

And that is why a lot of people stop the scrolling. And the other half of the sound is the clips themselves, because every CeeDance clip was generated with its own audio that you hear the moment underneath the song, like the bells, the crowds, the fire. And that layer is quiet, but that's what makes each moment feel like you're standing in it rather than just watching a slideshow. And I promised also to mention how much it actually costed and how many credits I used. So, for the London video, it took me about 1,500 credits in order to make the video, which adds up to being about $50.

So, it's not very cheap, but bear in mind that it was a learning process for me, and I didn't really have any prompts that I previously used. So, it took me a while to get to things being right. I feel like you could do the video in about 1,000 credits, uh which turns out to be about $30, especially if you use the workflow that I described. All of the prompts that I used, the costume line for each era, the sound list, are going to be in the prompt pack linked in the description. And if you could subscribe to this channel, it would mean an absolute world to me.

Right now, only 6% of people watching these videos are subscribed, and it makes such a huge difference, and it allows me to make more videos like this. So, I would be really grateful, and I would love to see what videos you make. So, link those in the comments below. And see you in the next one.