AI Streetscape Animation from One Street View 🤯 | Full Workflow (Sketch to Render)

In this video, I'll show you how to create this streetscape animation starting from just a single Google Street View screenshot, all using one AI tool. So let's get started. First, I open Google Earth, find the street I want to redesign, and save a Street View image. You can use Google Maps Street View or just a photo you took on site. Once we have the street image, we can start generating with AI. I'm using Lovart.ai. It brings together a lot of the latest image and video generation models, including GPT image too. For this workflow, I'm using Nano Banana 2 and Pro. Then I simply drag the street photo into the chat. Next, I enter a prompt. Basically, I'm asking the AI to transform the Street View image into a hand-drawn sketch overlay. And I describe the style I want in detail. Within seconds, I get the first image I need for the animation. As you can see, it keeps the original Street View but overlays the design proposal as a hand sketch, which is perfect for the next step. Next, click on the image and continue prompting in the chat. This time, I ask the AI to add color while keeping the hand-drawn sketch style and generate a section perspective. This type of section diagram is really common in streetscape presentations. I've listed all the prompts I used in this video in the comments. You can copy them directly into your Lovart workflow to save time. Then we get the second image. It's really well aligned with the original photo and the sketch proposal. There are a few text labels I want to adjust. So I just click on the image. In the top toolbar, there's an Edit Text option. It automatically detects and extracts all the text from the image. From there, I can edit the text directly in the panel. Once I'm done, I just regenerate the image and it updates the text without changing anything else. This is honestly one of my favorite features in Lovart. The sidewalk width from 3 meters to 1 meter and the rest of the labels are corrected as well. Now using this revised image, I'm going to generate a realistic rendering based on the design sketch. In the prompt, I remove all annotations and ask the AI to translate the sketch into a realistic scene, keeping the trees, pavement, bike lanes, and planting strips consistent with the design. Then I hit Generate and we get the third image. This is the one I'll use for the animation frame. It translates the original sketch really well into a realistic rendering. Pretty accurate overall. There are just a few small things I want to tweak. I use the Mark tool from the bottom toolbar, click on the area I want to edit, and it automatically adds that selection into the chat. Then I just type a prompt, for example, add a black car here and generate again. Now we get an updated image with the change applied. Here's another thing I want to adjust. I want to change the paving on the right lane. It's pretty common for clients to request specific materials or for us to propose materials from a manufacturer. In this case, I want to replace it with this permeable paving. So I use the Pen tool to mark the area I want to change. Then I select both the image and marked lines, upload the reference paving image and enter a prompt like replace the marked area with the reference paving, and the AI updates just that section while keeping everything else consistent. Now the paving texture is applied to the marked area with the correct perspective. If we compare it to the previous image, you can see the paving has been replaced really cleanly. One last small adjustment. I want to move the black car a bit further back. So I add two marks, one on the car and another at the end of the street. Make sure both marks show up in the chat. Then I enter a prompt like move the car from point 1 to point 2 super quick and simple. And now we're done editing image 3 and ready to generate the animation. Now let's move to step 2, generating the animation. After generating and editing, we have four images. We'll use them to create three short videos with each video using a start frame and an end frame. Then at the end, we'll connect all three videos into one complete animation sequence. Let's start with the first video. First, switch the model to video generation. Then select the images from the canvas and set the google street view image as the start frame and the sketch proposal image as the end frame. After that, enter the prompt to guide the transition. In the prompt, I describe how I want the white annotations to appear gradually. And how the sketch overlays should come in naturally over time. This helps guide the transition so it feels smooth and intentional instead of just popping in. After a few seconds, the first clip is ready. How does it look? Drop a comment and let me know what you think. Using the same method, we set up the second video. This time, I use the second image as the start frame and the section perspective as the end frame. In the prompt, I describe how I want the color to gradually fade in. And how the section drawing appears step by step while keeping the background street scene consistent. After that, we generate the second animation. And for the third video, we just repeat the same process. This time, I use image 3 as the start frame and image 4 as the end frame. Follow the same setup, enter the prompt, and generate the final clip. You can find all the prompts I use in the comments. No worries if you can't see everything clearly here. Next, we'll move into some more advanced edits to create multiple views of the street design and turn them into animations. This helps make the final video feel more dynamic and engaging. If you click on the image, you'll see a toolbar at the top with a bunch of editing tools. These tools let you make changes without affecting the rest of the image, which helps avoid common AI editing mistakes. For example, I can do a quick edit here. Using the weather switch tool, I can turn this into a rainy scene. Or I can use the erase tool to remove elements like this bike right here. After a few seconds, we get an updated image with a rainy atmosphere, and the bike is cleanly removed. And here's a really cool feature, the multi-angle tool. Click on it, and you'll see a simple 3D model that lets you choose any camera angle you want. I rotate it to a bird's eye view so we can see the overall street design from above. Then just click generate. The first result gives me a slightly rotated bird's eye view of the street. But for the animation, I want a straight top-down view. So I adjust the angle and generate it again. Now I have a clean straight top-down view of the street. We'll use image 4 as the start frame and image 5 as the end frame to generate the next video. Make sure to switch the model to video generation. And you can also choose between different models. Cedence 2 or Clean 3 both work really well for architecture animations. Once the model is selected, follow the same workflow. Set your frames and enter the prompts. This time I describe a transition from a first-person street view to a bird's eye view to create a smooth camera movement. For the last step, I take a screenshot of the existing aerial view from Google Earth and bring it into the canvas. Then I add both the aerial photo and street design rendering into the chat. In the prompt, I ask the AI to apply the streetscape improvements onto the aerial image. And that gives us the final result, a top-down master plan rendering of the streetscape. With this top-down plan, we can use the bird's eye rendering and this image as the start and end frames for the final video clip. This creates a smooth camera movement from a bird's eye view down to a clean top-down plan. Then we bring all the clips together into one full sequence from existing street to design proposal to final rendering. Hope you enjoy this workflow.