Afrikaans
Akan
Albanian
Amharic
Arabic
Armenian
Azerbaijani
Basque
Belarusian
Bemba
Bengali
Bihari
Bosnian
Breton
Bulgarian
Cambodian
Catalan
Cebuano
Cherokee
Chichewa
Chinese (Simplified)
Chinese (Traditional)
Corsican
Croatian
Czech
Danish
Dutch
English
Esperanto
Estonian
Ewe
Faroese
Filipino
Finnish
French
Frisian
Ga
Galician
Georgian
German
Greek
Guarani
Gujarati
Haitian Creole
Hausa
Hawaiian
Hebrew
Hindi
Hmong
Hungarian
Icelandic
Igbo
Indonesian
Interlingua
Irish
Italian
Japanese
Javanese
Kannada
Kazakh
Kinyarwanda
Kirundi
Kongo
Korean
Krio (Sierra Leone)
Kurdish
Kurdish (Soranรฎ)
Kyrgyz
Laothian
Latin
Latvian
Lingala
Lithuanian
Lozi
Luganda
Luo
Luxembourgish
Macedonian
Malagasy
Malay
Malayalam
Maltese
Maori
Marathi
Mauritian Creole
Moldavian
Mongolian
Myanmar (Burmese)
Montenegrin
Nepali
Nigerian Pidgin
Northern Sotho
Norwegian
Norwegian (Nynorsk)
Occitan
Oriya
Oromo
Pashto
Persian
Polish
Portuguese (Brazil)
Portuguese (Portugal)
Punjabi
Quechua
Romanian
Romansh
Runyakitara
Russian
Samoan
Scots Gaelic
Serbian
Serbo-Croatian
Sesotho
Setswana
Seychellois Creole
Shona
Sindhi
Sinhalese
Slovak
Slovenian
Somali
Spanish
Spanish (Latin American)
Sundanese
Swahili
Swedish
Tajik
Tamil
Tatar
Telugu
Thai
Tigrinya
Tonga
Tshiluba
Tumbuka
Turkmen
Twi
Uighur
Ukrainian
Urdu
Uzbek
Vietnamese
Welsh
Wolof
Xhosa
Yiddish
Yoruba
Zulu
The video you're watching right now,
it's a gausian splatting environment was
captured from a real location right here
in Manchester. I scanned this location
using my 360 camera and after completing
a few important steps, I imported the
entire model directly in a real engine.
So, in this video, I'm going to explain
you full process step by step how I scan
the location, which camera I use and
what software I use to complete this
process. And most important, what
mistakes I've done along the way so you
can avoid them if you want to try this
for yourself. So, don't skip the video
because you're going to miss important
steps from the process and probably
you're going to get stuck along the way.
To capture the location, I use my Insta
360 X2. It's a solid camera, records in
manual mode, and gives me a four
resolution, which is great. Along with
the camera, I had the selfie stick,
which I used to mount the camera. I
picked one location in the city, the one
that I wanted to scan for a long time.
And the method I used was very simple.
First, I walk left to right in a zigzag
pattern across the entire bridge with a
selfie stick above my head, exactly how
you see in this footage. In this way,
you cover more details of the bridge. If
you want to shoot in a busy place, you
should wake up in the morning, otherwise
you'll have people walking all around
you, which is very annoying. I did a few
tests before and it was a lot of people
and was very frustrating. Once I reached
the end of the area I wanted to scan, I
turned around and walked back in a
straight line. No zigzag movement from
left to right anymore. Then I did one
more pass with the camera positioned
close to the ground. The first time I
scanned the bridge, I skipped the floor
and the result looked a bit funny. By
lowering the camera, I thought that
would fix the problem, but in the end it
didn't. All this was captured in a
single recording and it took around 8
minutes to complete.
[Music]
If you own an Insta 360, you
automatically have access to Insta 360
Studio. Just go on their website,
download the app, and install it into
your machine. Insert the SD card, and
the app should instantly ask if you like
to import your capture video, which is
very useful. Select your file on the
right hand side, then head to the export
tab, go on the 360 video, pick your file
name, and export the file into your
desired location. Set the bit rate to
maximum, resolution to the highest
available, and encoding format. In my
case, I'm going to use H.264. Depending
on your footage length, this step is
going to take a while. In my case, it
took around 30 minutes to process and
export the video. At this point, the
file is ready to go. If you watch my
other tutorials, you know, I always put
extra effort into this because even
small adjustments make huge difference
in the end. After some quick tweaks in
Premiere Pro, the footage looks a little
bit much nicer. You can see before and
after here. It's a very small change,
but trust me, in the end, all the
details matters. Hey YouTube, before we
move on, I just want to say thank you
for watching this video. I really want
to hit 5,000 subscribers by the end of
the year, and you are the only one who
can help me. So, please don't forget to
subscribe. Now, back to the video. Once
you color grade the footage, either you
export it from Premiere or Insta 360
Studio, I use the 360 portable kit by
Cinematic Tob. He created a script that
combines FFmpeg with Alice Vision. I'm
going to link his tool and his YouTube
channel down in the description below.
All you need to do is download the tool
and open the BAT file. Drag the exported
360 video file. Check how many frames
you want. I think this is going to split
each second from the video in one frame.
For the testing, I use one and I let it
run. This automatically will create a
folder with a bunch of image sequences.
Just open up the folder and you can see
them right there. Next, we are going to
drag the folder that just been created
in the second part of the script. This
process will split your 360 images into
multiple 2D images that represents that
photo. In my case, I set to have eight
splits with a resolution of 1,200 and a
field of view of 100. Depends how many
images you have in the folder. This step
can be quick or you might have to wait
some time. Now if we put everything side
by side, first you would have the 360
video that you capture it with a camera
and then you split that video into 360
images and then later the 360 image has
been splitted in eight image. Crazy how
powerful this tool is. At this stage you
can either use reality capture to speed
up the process or go directly in
postshot. For me reality capture
struggle to align all these images. So
I'm going to skip this step and jump
directly into postshot. Now, one thing
to note is Photoshot is no longer free
or sort of. You'll need a subscription,
but the advantage is that lets you save
the file that can be loaded into an
Unreal Engine with the plug-in. So, this
is very useful. Sadly, exporting as a
PLY is no longer available in the free
version, but don't worry, you can still
follow along with this tutorial.
Once you open postshot, I drag the
folder with the split images to drag in
the last folder you just created. Drag
it into the software. Keep all the files
max image count. Switch to splat MCMC.
This is going to give a better results,
but it's going to take a longer time.
So, be aware. Start training immediately
and stop it at 300,000 steps. The more
steps you have, the sharper and cleaner
the result will be. For me, it took
around 9 hours to train this file. After
that, the result looked amazing. better
than the first few tries. I could
finally explore the spot inside postshot
on the left. You can switch between
different viewings. Point cloud camera
alignment which did a very great job. If
you see fuzziness, use cleanup tool to
remove all the unwanted artifacts.
[Music]
And very important, very very very
important, don't forget to save the
file. Now it comes the fun part. Yes,
you can bring your spots directly into
Unreal Engine. Open Unreal Engine. Start
a new project in firsterson mode.
Once the level started, make sure to
enable the postshot plugin under the
edit plugins and type postshot
and give a quick restart. When the
project opens, I deleted the default
shapes, leaving just the floor and the
blue cubes.
Then I drag my save post shot file into
the new folder inside Unreal Engine.
If you want to move the splot in the
environment, the pivot point probably is
going to be off. So to fix this, I place
an actor in the center of the bridge.
Then made the splat a child of the actor
so I could position everything
correctly.
From there, I adjusted the scale of the
splot. The real bridge has 6 m wide, so
I align six andreal engine cubes side by
side to get the dimension right. Just
remember, always lock the scale so you
don't stretch the model weirdly. Once
the splot is aligned, you can hit play
and explore the splot. But there's a
catch. Gausian splatting files don't
usually have geometry, which means we
don't have physical collision. To fix
this, I built a basic shape in a real
engine. I created the floor, the side
fences of the bridge, stairs, light
pose. Essentially, I created the entire
environment using cubes. When I
finished, I made those cubes invisible
in the details panel. Select all the
cubes, type visible, and then check that
bit. So the game had proper collusion
while keeping the splat visual. If you
want to play in the game, just move the
cubes however you want. Ideally, it
should be position on the geometry that
you just created. Finally, I move the
spawn point to always start at this
little path before the bridge. Now,
every time I hit play, I'm going to
start from this location.
Heat.
Heat.
[Music]
And that's pretty much it. This is how
you bring any locations from reality
directly in Unreal Engine. All you have
to do is to have a 360 camera and follow
this workflow. You can bring anything
inside and you can play, you can create
your own video game. So, thank you for
watching. I really want to hit 5,000
subscribers by the end of the year. and
I know you are the only one who can help
me. Thank you so much for watching and
I'll see you in the next one.
[Music]
[Music]
Can't find what you're looking for?
Get subtitles in any language from opensubtitles.com, and translate them here.