@fdotinc
Instagram · Vlog
@fdotinc ai is scary lowkey #tech #engineering #ai #chatgpt #startup
@fdotinc
Instagram · Vlog
@fdotinc ai is scary lowkey #tech #engineering #ai #chatgpt #startup
Performance
Video #42345 · Vlog
[01]
The creator provides a step-by-step breakdown of how a colleague used AI to create a 3D model of 1920s San Francisco. He explains that the process involved scraping thousands of historical photos, using an AI to annotate and find the best references, cross-referencing images with historical floor plans like fire insurance maps to build 3D structures, and finally, mapping the original photo textures onto the models to create a realistic, historically accurate street view.
[02]
[03]
[04]
[05]
What stops the scroll in the first 3 seconds
Title hook · 15%
so look at this
Verbal hook · 15%
Look at this.
Visual hook · 70%
A first-person point-of-view shot moving towards a person working at a complex multi-monitor computer setup, creating immediate visual interest and curiosity about the work being done.
[06]
Steal this for your own video
A first-person POV shot slowly pushing in toward someone working at , where you can almost make out but not quite.
[07]
The promise the opener makes, and how the video pays it back
Hook
A direct, attention-grabbing claim that establishes the video's central thesis and creates immediate intrigue about the subject's capabilities.
Payoff
A concluding statement that summarizes the core value or achievement of the entire process, directly proving the initial hook's claim and providing a satisfying sense of completion.
[08]
The structural pattern this video runs on
Framework
The Layered Reveal
To demonstrate the power or complexity of a process by breaking it down into a series of increasingly sophisticated, layered steps, starting with a massive raw input and ending with a highly refined output.
The Foundation Layer
Establishes the raw, large-scale input or source material for the process, emphasizing the sheer volume of data being used as a starting point.
The First Intelligent Filter
Introduces the first layer of processing, where an intelligent system sifts through the raw input to identify and isolate key elements of value.
The Structural Cross-Validation
Introduces a secondary, different type of data source that is used to provide structure, context, or validation for the elements identified in the previous step.
The Synthesis & Application
Shows how the initial visual data is synthesized with the structural data, often by applying the former onto the latter to create a more complete representation.
The Granular Refinement
Reveals a final, highly detailed processing step that adds a layer of realism or sophistication, transforming the synthesized model into a polished final product.
ai is scary lowkey #tech #engineering #ai #chatgpt #startup
contextual · beginning
so look at this
contextual · beginning
powerful
contextual · beginning
scraped
contextual · beginning
of 1000s
contextual · beginning
of 1920s San Francisco
contextual · middle
model
contextual · middle
each image
contextual · middle
reference
contextual · middle
this one
contextual · middle
only see
contextual · middle
background
contextual · middle
it then
contextual · middle
the image
contextual · middle
historical
contextual · middle
like fire
contextual · middle
maps to
contextual · middle
a 3D
contextual · middle
another
contextual · middle
then grabs
contextual · middle
angles
contextual · middle
original
contextual · middle
as a
contextual · middle
onto the
contextual · end
signage
contextual · end
even get
contextual · end
to create
contextual · end
street
contextual · end
is based
contextual · end
history
Look at this. This is how powerful AI has become. This guy scraped hundreds of thousands of photos of 1920 San Francisco. An AI model annotates each image to find the best reference, like this one where you can only see this building in the background. It then cross references the image against historical floor plans like fire insurance maps to develop a three d structure. Another model then grabs any available angles no it can map original images as a texture onto the model. Signage and billboards even get extruded to create a realistic street view where everything is based on real history. Yeah.