Update 25/06/2016
3DNes is released at :
https://geod.itch.io/3dnes---------------------------
Update 08/03/2016
Finally the beta is ready:
http://tructv.bitbucket.org/3dnes/It only runs on Firefox for now.
Have fun!
-------------------------------------------
Update 19/02/2016
The beta release candidature trailler is here
https://youtu.be/Ti2fwroyP2A with many new features : better shape segmentation, transparency, z layers etc ...
------------------------------------------------------
Update 12/31/2015
Hello Everyone!
Longtime no see
.
Lately i'm working to improve the vision and sound.
Lighting and shadows are added.
And now it's time for showcases.
Next step, i will work on z layers problem.
https://youtu.be/wzBjHEdh14khttps://youtu.be/_Sl_Ou3Wrbchttps://youtu.be/576mERxRnj4https://youtu.be/oPsWI5htHNwhttps://youtu.be/bKHYgiLuiNUwhttps://youtu.be/oPsWI5htHNw----------------------------------------------
I'm currently working on a 3D Nes Emulator. My work has achieved an initial result. Below are several clips to showcase my program.
https://www.youtube.com/watch?v=cUXYgxEfQrA&list=PLhplDJFRf1RINuz-T3s16_7OssOeiAx8y I would like to receive any remarks or suggestion.
Please share your thought. Thank you.
I am impressed.
At first I thought everything is beveled into a block with repeated texture wrapped around, until I realised that different tiles/sprites can have different breadths, and that pipes are properly cylindrical.
I think every single game has to be setup independently to achieve this result, right?
I never realized that piranha plants always rendered, even in the pipes.
Gilbert wrote:
I am impressed.
At first I thought everything is beveled into a block with repeated texture wrapped around, until I realised that different tiles/sprites can have different breadths, and that pipes are properly cylindrical.
I think every single game has to be setup independently to achieve this result, right?
No, every game is automatically rendered with the same engine, the same code.
I know that it's the same engine(same emulator), but different games has to use different data files, right?
I meant data files that define how to coop with different tiles/sprites, like how 'thick' a wall tile and a ground tile are (from the video it's obvious that a wall tile is not as thick as a ground tile, and the text fonts are even thinner, intentionally) and in case of pipes, etc., what models to use.
Of course when no such data file is provided an emulator may still consider each tile/sprite as a block of the same thickness, just won't look as good.
This is similar to the HD emulators, in which you need to provide new graphics for each game.
No, the 3d scene rendering is based on automatic "on the fly" analysis of 2d scene. I don't follow the approach of HD Nes. This approach requires huge amount of graphic work done to let just a game played. You can see that when an object is rendered for the first time, the game become a little lag due to the analysis computation.
How do you decide whether a given tile/sprite thing is rotationally symmetric or something else?
lidnariq wrote:
How do you decide whether a given tile/sprite thing is rotationally symmetric or something else?
General speaking , the emulator have a fixed set of predefined geometric pattern like cylinder, cube ... vertical or horizontal, when the game played, the 2d scene will be analysed to detect 2d geometry objects on it. Then a set of heuristics rules will be used to assign each object a pattern which will be used later to convert this objects into 3d. That's the mechanism behind it.
Very impressive for a heuristic method.
Super Mario Bros. works OK, except the treetops in 1-3 look like a green log on top of the tree trunk. But it seems to treat the outdoor background in
The Legend of Zelda as thick fog.
I wonder what it does with
Dr. Mario or any of
Shiru's games or any of
my games.
geod wrote:
Then a set of heuristics rules will be used to assign each object a pattern which will be used later to convert this objects into 3d. That's the mechanism behind it.
Now that's really really impressive!
So you're saying, for example, it can see that the pipe looks like a cylinder in that it's bright in the center and darker along the edges (and the pattern repeats vertically)? What I really want to know is how well this would work on an NES game with non perfectly front facing floors, (if that makes sense) like Battletoads. In order for this thing to work flawlessly (or even close to it) without hacks, it would practically have to be a human looking at the game. And just thinking, in the case of Battletoads, I guess it would also have to infer what is supposed to be up, and what is supposed to be further in the background for the characters.
It's pretty cool, definitely more interesting to look at than the trivial solutions.
That "first person Mario" video reminded me of many PSX/N64 games with short draw distances, some fog might be funny in that video in attempt to cover it up somewhat. (I'm sure this would lead to somebody making a comment in YouTube about NES's draw distance...)
In fact, that's one of the reasons games went 3D in the first place: in order to be able to draw things both close and far away. It's helpful in many outdoor scenarios and flat-out necessary in a racing game, with few exceptions.
Really impressive, especially given your explanation on how it works.
Espozo wrote:
So you're saying, for example, it can see that the pipe looks like a cylinder in that it's bright in the center and darker along the edges (and the pattern repeats vertically)? What I really want to know is how well this would work on an NES game with non perfectly front facing floors, (if that makes sense) like Battletoads. In order for this thing to work flawlessly (or even close to it) without hacks, it would practically have to be a human looking at the game. And just thinking, in the case of Battletoads, I guess it would also have to infer what is supposed to be up, and what is supposed to be further in the background for the characters.
Well we can say that the emulator is still in a very alpha stage. What the emu has succeeded to establish is the infrastructure, the framework. In the future, more semantic info will be collected from 2D scene,more predefined patterns will be added, more rules will be created to make the emu more robust. You say it right there will never be a perfect solution for this kind of problem but better solution is always possible. For the camera mode/direction detection: top down, left right or front back, perspective or orthographic or 2d projection, currently the emu has not detected it yet so the render will not be good for 3d camera or complex background game.
And when the emu becomes mature enough, the optimization for a specific game is as simple as values assigning for a set of coefficient. That's how i see things.
tepples wrote:
Very impressive for a heuristic method.
Super Mario Bros. works OK, except the treetops in 1-3 look like a green log on top of the tree trunk. But it seems to treat the outdoor background in
The Legend of Zelda as thick fog.
I wonder what it does with
Dr. Mario or any of
Shiru's games or any of
my games.
Fortunately my emu can run Thwaite game
https://youtu.be/TmxTL8Aw1AkFor Zelda game, it is the problem of background layers positioning in z axis. I have not tacked it yet.
Very impressive! I suggest removing the pixel from the background layer if a sprite pixel is on top of it. This reminds me of another project for game boy but this one use custom model as Gilbert asked:
http://www.emutalk.net/threads/55483-GameObj-Unity3D-GB-Emulator!Are you using Unity3D too?
mkwong98 wrote:
Very impressive! I suggest removing the pixel from the background layer if a sprite pixel is on top of it. This reminds me of another project for game boy but this one use custom model as Gilbert asked:
http://www.emutalk.net/threads/55483-GameObj-Unity3D-GB-Emulator!Are you using Unity3D too?
Yes, i'm using Unity too.
What you mean by "on top of it"? I don't fully understand.
I mean in front. When the sprite is in the front layer like Mario is in front of the plant in 1-1, remove any pixel from the background object that also has a front layer sprite pixel.
By the way, how will Kung Fu look with this emulator? The main character is on the background layer.
mkwong98 wrote:
I mean in front. When the sprite is in the front layer like Mario is in front of the plant in 1-1, remove any pixel from the background object that also has a front layer sprite pixel.
By the way, how will Kung Fu look with this emulator? The main character is on the background layer.
I'm afraid that thing doesn't work that way. Because 3d object models must be independent with each other. We don't have the right to remove any thing. What happen if the character move out of background object?
geod wrote:
I'm afraid that thing doesn't work that way. Because 3d object models must be independent with each other. We don't have the right to remove any thing. What happen if the character move out of background object?
Then is it possible to generate the model, but break the model into 64 pixel size mini models? And the emulator must implement the pixel priority to determine which pixel is visible. This way you can remove pixels that are obscuring others. This also applies to two sprites occupying the same location.
In Kung Fu, I mean the character is actually make up of background objects. Do you treat background objects and sprite objects differently?
mkwong98 wrote:
geod wrote:
I'm afraid that thing doesn't work that way. Because 3d object models must be independent with each other. We don't have the right to remove any thing. What happen if the character move out of background object?
Then is it possible to generate the model, but break the model into 64 pixel size mini models? And the emulator must implement the pixel priority to determine which pixel is visible. This way you can remove pixels that are obscuring others. This also applies to two sprites occupying the same location.
In Kung Fu, I mean the character is actually make up of background objects. Do you treat background objects and sprite objects differently?
I think that this approach come from the NES PUU itseft by this i mean 2D solution. Imagine that there are two 3D objects with different thickness, we filler out some pixel of back object which is thicker then when the camera is not perpendicular with the scene, the back object still hides the front object from camera view.
Btw, can you show me where to get the Kungfu game?
The legal way:
- Buy a Kazzo dumper from infiniteneslives.com
- Add Kung Fu for NES to your game collection
- Dump it
- Verify that your dump has the same CRC as the verified good dump
The illegal way:
Download a ROM
Ok that game i know it. I asked to make sure which game he is talking about?
I'm interested about what solution you will have for top down games (e.g. Zelda, 1943) or games (e.g. Ninja Gaiden) where the background is likely to be filled with solid tiles.
Hello Everyone!
Longtime no see
.
Lately i'm working to improve the vision and sound.
Lighting and shadows are added.
And now it's time for showcases.
Next step, i will work on depth layers problem.
https://youtu.be/wzBjHEdh14khttps://youtu.be/_Sl_Ou3Wrbchttps://youtu.be/576mERxRnj4https://youtu.be/oPsWI5htHNwhttps://youtu.be/bKHYgiLuiNUwhttps://youtu.be/oPL0TqP71eg
576mERxRnj4: Somehow the Castlevania logo becomes a cylinder that occludes the C!
oPsWI5htHNw: Yet something that's supposed to be a cylinder (the Dr. Mario logo) isn't. It tries to make separate cylinders out of the letters in the title. And is there a reason the top of the bottle and the "Game Over" mark pulsate when the bank changes? Is it that it did the depth recognition with one bank, and things differ enough between the banks that the recognition changes?
bKHYgiLuiNUw: Shadows in SMB1. Very nice.
Hi tepples,
I was working on Vision and Sound last month.
Concerning PPU 3D, i have changed the underlying code into more general one but haven't touched its functionality yet.
I'm tacking it now, and we will discuss this topic in the next version which is probably playable.
For now just enjoy the lighting and shadows please
.
By the way, the blinking effect in DrMario is done by bank switching technique? Awesome!
I was wondering, when the emulator looks at a 8x8 tile, does it look at any of the ones around it before it decides how it wants to render that tile? I don't know if that makes sense, but I was looking at how the magnifying glass on Dr. Mario has the middle bulging out in a staircase like pattern while the rest is rounded (which is incorrect, but it's a start.) I think some of the bushes and stuff in SMB do the same thing. Maybe you could have it check stuff around it to make it bulge out as much as the stuff around it instead of completely. Just trying to give advice. I really like what you've done so far though. I will say one thing that was strange is how the text in SMB was rendered perfectly, but in Kung Fu, different characters were at different depths for seemingly no reason.
Espozo wrote:
I was wondering, when the emulator looks at a 8x8 tile, does it look at any of the ones around it before it decides how it wants to render that tile? I don't know if that makes sense, but I was looking at how the magnifying glass on Dr. Mario has the middle bulging out in a staircase like pattern while the rest is rounded (which is incorrect, but it's a start.) I think some of the bushes and stuff in SMB do the same thing. Maybe you could have it check stuff around it to make it bulge out as much as the stuff around it instead of completely. Just trying to give advice. I really like what you've done so far though. I will say one thing that was strange is how the text in SMB was rendered perfectly, but in Kung Fu, different characters were at different depths for seemingly no reason.
The rendering mechanism is not just based on 8x8 tile, but based on "shape" - meaningful collection of consecutive tiles. Shape recognition - like any recognition problem - is hard and does not have an ultimate answer. On Nes System fortunately we can choose color pattern as the main factor for shape clustering, it works for 90% case. It's still an open problem for improvement through. Consequently an 2D tile is not always mapping to an "3D tile", for example an 2D solid tile can be used in the introduction cube at the begin of SMB, but also be used in grass or cloud. The corresponding 3d tile in each case is completely different. The original code only allowed one 2d tile to be mapped to one 3d tile. For the current code one 2d tile is technically allowed to be mapped to many 3d tile but still be mapped to only one 3d tile. That's why you see some small strange detail in grass, in pipe etc. I could tack it earlier of course, but it would affect greatly on emulation speed so instead i went for the missing aspect (light, shadow, adequate sound quality) first then performance re-optimization. Right now i'm ready to return to 3d convert problem.
I am highly, highly impressed with this. I can't wait to see a stable build come into existence.
But first, a few questions I'd like to ask:
1) How exactly does the 3D rendering work? After watching a few videos of your emulator at work, I am curious as to how the depth works.
2) Does extra code need to be written for every NES game? That seems improbable because so many of them exist, but it seems a little weird to me that the code for the games themselves would have the data for such a task.
3) Does the emulator alter the code of the games themselves? This seems likely to me, but I'm new to programming in general, so this could be due to ignorance on my part.
Other than that, this looks really cool. Great work so far, my man.
Hello, thank you for the encouragement.
1) Let say the depth of a shape is a function with inputs are shape type, shape size, shape layer and its relation with other shapes
2) No extra code for any particular game, this will very soon become a maintenance nightmare. In the future, in its complete form, the emu will have a set of coefficients, and we could do adjustment for a specific game by modifying this set, that's the limit.
3) No, the emu is a normal emulator plus a realtime 2d->3d converting engine. That's all.
Any updates on this? Very impressive work! Can't wait for a beta release...
How does something like Rad Racer look on this? It seems more suited for platformers, but would be interesting to see.
Hi the update news is almost there. I'm late compared to the schedule. Please just wait a few more days then the update news will come out. Thank you for the support.
By the way the Rad Racer game is already a 3D game by itself so i'm afraid that there are no many things we could gain here
This looks pretty amazing! And I really like that the emulator actually analyzes the images to give them proper 3D models, instead of just stretching them on the Z axis. I know it's a work in progress, but do you have a compiled version or something, that we can try out? I love experimenting with this kind of stuff and would like to see how a lot of games would look, running on this.
Sorry for the wrong timing prediction. I have just passed the New Lunar Year Holiday and couldn't do any work during this period.
I will try my best to make it available as soon as possible
Happy Monkey Year everybody
Hi all,
The beta release candidate ... trailer is here
https://youtu.be/Ti2fwroyP2A with many new features : better shape segmentation, transparency, z layers etc ...
Test Pool contains :
- DrMario
- Legend of Ninja
- Mega Man
- Super Mario Brothers 1
- Super Mario Brothers 3
- Zelda Link
- Contra
- Castle Vania
- Kungfu
- Battle City 1990
Please share your thought.
The beta release will come soon
My first reaction is probably "how the /!"(#/ did he pull that off?!"
Nice work!
You should sell this technique to Nintendo for them to implement in Virtual Console for 3DS.
Oh, I wonder ... it seems likely that Bubble Bobble would look good through this?
Good job!
It seems the emulator now detects when a front sprite tile is covered by a background tile and changes the background tile to semi-transparent. And after a short period of time, moves that background tile to the back. I suggest changing the length of the time period based on the size of the overlapping area. So if most of the pixels of the two tiles overlap each other, then reduce the time period. I also suggest lowering the opacity of the semi-transparent background tiles so that the sprite tile covered by it will be more visible.
I can also see that detection algorithm have problems with chr ram games because there are multiple copy of the same tile in the ram. Maybe you can check for repeated tiles by comparing the tile pixels.
Thanks @mkwong98
You got it right. This is the basic idea of z layer processing. Concerning your suggestion:
- Overlapping area detection in detail is a costly processing. For the time being i just check two shape overlap each other or not. And things sometime are quite complicated. For example in Castlevania, the bat moves in sin wave and goes completely through the ground for a sort period of time but we can not put the ground in the far layer.
- About the opacity and visibility, i will recheck it all.
- About chr ram game, do you see problem or guess it? Because i already define tile identity by it own pixel data.
CHR RAM games can construct tile data at runtime. I'd like to see what it does with
Qix (vectors),
Elite (vectors),
Videomation (vectors),
Hatris (vertically tile-unaligned bitmaps), and the text in
RHDE and
robotfindskitten (horizontally tile-unaligned bitmaps).
@tepples
Could you clarify the term "vertically/horizontal tile-unaligned bitmaps"?
Look at a screenshot or video of
Hatris. The hats are drawn to the background, but not at 8-pixel increments. They're drawn into tiles at arbitrary vertical positions, using AND and OR operations, and copied to CHR RAM. I'd embed a screenshot, but most pixel-perfect screenshots that Google Images found were 1. on ROM sites and 2. using HTTP's
Referer header to block embedding.
RHDE uses a
proportional font. A single glyph may be 2 to 7 pixels wide. The letters are drawn into tiles at arbitrary horizontal positions, using shift and OR operations, and copied to CHR RAM.
A screen full of text in a proportional font
geod wrote:
- About chr ram game, do you see problem or guess it? Because i already define tile identity by it own pixel data.
For example around 18:05 in the video, I think those wall tiles on the right part of the screen are the same as those on the left part of the screen? Or did you do that on purpose?
@mkwong98
- This is the side effect of shape segmentation algorithm and not related to chr ram
@tepples
Hatris Clip
https://youtu.be/8oP2PckYZIsRHDE
https://youtu.be/NVUxS_JjrYYWith RHDE the emu hangs after the introduction scene. It seems to me that it does not like working with homebrew games at all
Hatris
The BPS and Hatris logos were messed up, possibly because the background is not color 0. It tried to do a good job with the crown, but the background pattern overwhelmed the wizard hats. On the next screen, I was impressed that it almost got the jukebox, the "STAGE" cabinet, and the window right, but "SHOP" was recessed. "STAGE" also flipped to recessed once one was chosen. I'm not entirely sure anything will detect all possibilities correctly.
The gameplay screen looked surprisingly okay, though it got confused when rounded vs. not rounded. The conveyor belt flickered strangely, sort of like a bunch of things in Dr. Mario.
RHDE
The Blender-rendered, cel-shaded scene in the title screen is flat except for a couple cutouts in the guy towing the chair, caused by using sprites to cover attribute clash. (The title screens of Contra and Tetris 2 have the same thing.) I didn't really expect more, as recognizing depth from a scene like that would be very game-specific. "FURNITURE FIGHT" at bottom is recessed, making it look as if it were a wall that originally read "FURNITURE FIGHT" onto which "RHDE" had been spray-painted. I found this effect interesting. In the text at bottom, some words were extruded more than others, and the top ("5 Rounds" and "Endless") sort of got lost under the wall.
Select Race: The sides are more or less completely flat, except for under the poli character in the center, because it can't tell the sprites from the background. I was going to redesign this screen entirely for version 2 though. The "1" and "2" are rounded. The grass is rendered behind the house, possibly because it's behind sprites. And the sky is in front.
As for freezing, can you make your emulator output a trace log of executed instructions to see where it's getting hung up? It could in fact be my own bug in the game, where I'm depending on some poorly specified behavior of the system.
Very detailed feedback, thank you.
If you check the video of new version, DrMario does not flicker anymore
. In most case shape segmentation is the source of error. It's always hard and even harder when time constrains come into the show. I do not have the computation power to analyse the frame per pixel, only do it per tile so the situation is even worse. We still do not count the fact that shapes could share the same tile. Some pixels belong to one shape and some belong to another shape and the current mechanism couldn't do anything to deal with it
. Anyway the emu should have specific game coefficients to obtain near perfect rendering but it's the future work.
The new version become hangs while running game with save state and i still don't know why, maybe i will implement trace log when i deal with this bug.
It might end up the case that we'll need to go with game-specific depth hints per tile.
I think i will not try to achieve the perfect tile depth and z-position. Basically the creator designed the game with 2D space in mind so in some case it's almost impossible for perfect 3D space converting even by human mind. So my ultimate goal is much more modest - just a "decent" 3D scene with clear visibility of shape/tile.
Finally the beta is ready:
http://tructv.bitbucket.org/3dnes/It only runs only Firefox for now.
Have fun!
Kudos on the amazing work!
JOOC, are you using some machine learning algorithm to learn to classify the shapes? If yes, which one?
Aaand
AlphaOmegaSin just made a video about your emulator. Congratulations!
*Edit: Also, trying to open a game of my own, but the script crashes. (And, yes, I am running it on Firefox) Any help?
BioMechanical Dude wrote:
Aaand
AlphaOmegaSin just made a video about your emulator. Congratulations!
*Edit: Also, trying to open a game of my own, but the script crashes. (And, yes, I am running it on Firefox) Any help?
Please try to reread the just updated "How to play" section. If it still does not work for you then let's blame on the Unity and Firefox guys
.
Well... the online version is unplayable. You need high lucky to get it working.
I don't mind about videos, but if it's working or not here.
It is NOT.
Agree sir. It is not IN MANY CASES:D.
How does the emulator figure out what parts of the image to render in 3D? For instance, see this segment of the map from Castlevania:
http://www.nesmaps.com/maps/Castlevania ... 7bMap.htmlThe blue background tiles look almost identical to the orange floor tiles. What algorithm could automatically tell the difference?
Blue tiles are more numerous? Sprites are constantly placed in front of them?
tokumaru wrote:
Blue tiles are more numerous? Sprites are constantly placed in front of them?
The videos do not suggest that the emulator learns during game play. It's not like it starts out as 2D and it slowly figures out what to pop out into 3D.
As for where sprites can go, bats, ghosts, player weapons, enemy projectiles, items, etc. move over solid wall blocks. And, the player can jump upwards through blocks to climb; blocks only become solid again when you land.
One possible explanation is that the emulator contains a tile table for all the common games, providing hints on how to display the tiles.
zeroone wrote:
One possible explanation is that the emulator contains a tile table for all the common games, providing hints on how to display the tiles.
I remember the developer saying he didn't use databases, just heuristics. I don't know if that has changed between then and now.
Very impressive work!.
Quote:
The rendering mechanism is not just based on 8x8 tile, but based on "shape" - meaningful collection of consecutive tiles. Shape recognition - like any recognition problem - is hard and does not have an ultimate answer. On Nes System fortunately we can choose color pattern as the main factor for shape clustering, it works for 90% case. It's still an open problem for improvement through. Consequently an 2D tile is not always mapping to an "3D tile", for example an 2D solid tile can be used in the introduction cube at the begin of SMB, but also be used in grass or cloud. The corresponding 3d tile in each case is completely different. The original code only allowed one 2d tile to be mapped to one 3d tile. For the current code one 2d tile is technically allowed to be mapped to many 3d tile but still be mapped to only one 3d tile. That's why you see some small strange detail in grass, in pipe etc. I could tack it earlier of course, but it would affect greatly on emulation speed so instead i went for the missing aspect (light, shadow, adequate sound quality) first then performance re-optimization. Right now i'm ready to return to 3d convert problem.
Wow. That sounds pretty good. Only heuristics?. I've seen some videos and I think the result might improve if you apply some sort of machine learning algorithms although I understand that at this stage of development may be more convenient to improve the heuristics. In any case the result is impressive. I hope to see further progress soon.
geod wrote:
Agree sir. It is not IN MANY CASES:D.
Just to be crystal clear. The online version has problemas, it's not loading anything or working here. I really want to try out this emulator, and I can't wait for the "offline" version of it. ^_^;;
Ryoga wrote:
Wow. That sounds pretty good. Only heuristics?. I've seen some videos and I think the result might improve if you apply some sort of machine learning algorithms although I understand that at this stage of development may be more convenient to improve the heuristics. In any case the result is impressive. I hope to see further progress soon.
One important factor of machine learning algorithms is to define what is good and what is bad, and that is very difficult to do in this case.
Because many people said that this emulator may contain some kind of database for specific games, i would like to ask if there is a general method to know which game it is from a random input rom?
I don't know if this kind of method exist. Thank you.
geod wrote:
Because many people said that this emulator may contain some kind of database for specific games, i would like to ask if there is a general method to know which game it is from a random input rom?
I don't know if this kind of method exist. Thank you.
You would use something like NesCartDB and use the check-sum of the rom to determine what game was loaded.
Or use the database that comes with Nestopia
You've got competition. Someone else is working on a 3D NES emulator that runs
The Legend of Zelda in first-person view.
Source:
YouTube, via
Twitter
I almost believe that this is a serious work until i read some user comments - i suck hard at English, especially English listening
. By the way how does he do all the video drawing in this clip? That's just amazing!!!
tepples wrote:
You've got competition. Someone else is working on a 3D NES emulator that runs
The Legend of Zelda in first-person view.
Source:
YouTube, via
TwitterIn that link, the author provides some information on how his algorithm works out which blocks to make into 3D. It's still unclear what process 3DNes uses. Is it possible to do this without pre-constructing a tile table for each game ahead of time?
I do view the video the second time and understand almost all of what he said and wrote in screen. Stupid comments about April Foul and trolling .... This is the real work. So here is what i think about it. This project is based on the assumption that 3D blocks are always composed of 2x2 tiles. It allows for quick and easy 3d rendering. But that's it. No matter how many tricks and hacks are put in the pot, it wouldn't render correctly any game simply because in more than half of cases the above assumption is just plain wrong.
Hello everyone, long time no see
I'm here to report my progress on PC Build :
Tetris (by Tengen):
https://youtu.be/0ixlMNiXoX8Chip 'n Dale Rescue Rangers:
https://youtu.be/rSsVM4JuZHwFeel free to share your thought.
Tomorrow i will post new version of old game clip, so you can clearly see the difference
*drools over the Tetris gameplay section.*
The fact how the emulator guesses the geometry of different aspects is just amazing. How it determines the thicknesses of stuff is superb. Just look at the blocks compared to the surrounding "brick" frame in Tetris/Hatris, and the windows in Chip'n Dale's backgrounds. When I looked at the apples in Chip'n Dale I was like
and the player characters in both platformers looked very good too, just like
nanoblock models. The logo and opening screens of Hatris looked pretty bad, but the cute little heads and hats in-game pretty much made up for this.
One more thing. About the shadows, they're really nice. Sprites casting shadows on the ground looked really cool. But how customisable are the shadows(I haven't even tried the web version yet; sorry about that)? Can we tune them down or change the position of the light source ourselves, instead of disabling them completely? They sometimes obscured important stuff that made the games much harder to play. For example, in Tetris, the blocks in the left columns were being hidden by the shadows and the shadows sometimes made text(such as scores) hard to read too.
I think while turning a game 3-D wouldn't make much different to game mechanics or make games easier to play(actually this most probably makes games harder to play), it'd be quite awesome to use 3-D mode to playback game recordings (such as TAS recordings) originally done in the normal mode.
3DNes uses directional light, not spot light. Change the light direction in this case is the easiest thing i can imagine
. But how to make a good user interface on the upper layer is hard for me. So i just let it that way for now.
I have just tried 2D racing game and the result is quite good
https://youtu.be/MPmuWy9LgJ4
Geod wrote:
Super Mario Bros - PC Game Play
Cute. The score counter casts a shadow on the cloudbushes.
Hatris is improved.
Gilbert wrote:
the player characters in both platformers looked very good too, just like
nanoblock models.
In your link, can I get past this screen without creating an account?
Odd. I never saw that screen. Maybe it's because I for some reason had signed up to facebook (though I NEVER use facebook).
Anyway. You won't miss anything not being able to read that page anyway. As everyone does, I just googled for "mario nanoblocks" and posted a random hit as an example.
And since you saw a blocked Ironman you already got the idea.
Are these some sort of Lego rip-off?
https://youtu.be/VAlICbb6Ij8For this game, the player is blocked at the border of screen and can not move.
Could anybody give me some suggestions about where the problem is located?
3dnes has a scanline based ppu emulation.
Thank you!
Can you play other similar games with sprite wrapping, like Mario Bros., with no problems?
https://www.youtube.com/channel/UC_z3ctZrPwVfYjPIYSckRbAI have played and recorded clips of ~ 30 official games even BattleToads without any game blocking problem. Link above.
However 3dnes has problem with some home-brew games of shiru
Btw what does it exactly mean "sprite wrapping"?
Wrap-around means you walk off the right side and reappear on the left and vice versa. It's the rule in
Asteroids,
Pac-Man,
Joust,
Mario Bros., and Nintendo's
Joust clone titled
Balloon Fight.
Does
Driar also fail?
Driar is blocked at first screen "Press Start". I press enter but nothing happens.
3DNes runs Balloon Fight, Mario Bros, Pac Man just fine.
@Zepper: why did you remove your suggestion?
As a random guess, Driar uses thefox's MUSE sound engine which uses unofficial 6502 instructions.
It's not a random guess at all. In fact it's the cause of Driar faillure.
Battle Toads - Convert the build-in fake shadow to the real 3D one - demonstration
https://youtu.be/bP9GnCXZPPs
3DNes is coming out with ingame realtime editing feature.
Part 1:
https://youtu.be/tyVFxcvtPwE
Hello,
Whoever want to try the pc build please inbox me, i will send you the link.
The amount is limited so first come first served.
3dnes already supports controllers.
It looks like keyboard shortcuts for the depth editor would be helpful. There's a lot of awkwardness with click - move - click - type - move - drag - &c that would hopefully be more fluid with fewer mouse movements.
Thank you. Totally agreed, i will implement it in the next build.
Hello everyone,
3DNes is released at :
https://geod.itch.io/3dnes
While you have the choice of putting a price in your work, 3dNES, I really hope you reconsider such thing, at least for the NES community. Sorry.
Zepper wrote:
While you have the choice of putting a price in your work, 3dNES, I really hope you reconsider such thing, at least for the NES community. Sorry.
This is my fulltime job plus part time job since 1 year ago so now i don't have a choice.
I understand and respect your opinion though.
Author should look into what happened with iNES (the emulator). Hint: charging money for your NES emulator, despite being niche in what it does, will not turn out well. You'll start spending time trying to combat piracy (this never works -- ever), because "this is my livelihood/how I survive". It'll be a constant stressor + uphill battle that cannot be won, all while someone else will be attempting the same thing as you (and release it for free and/or as open-source). This has happened time and time again with emulators (iNES, VSMC, Bleem!, and several others).
There are very few commercial console emulators that survived or were worth it. The only one I can actually think of right now is Magic Engine (back in the DOS days), and that's because it was for a console that quite literally nobody had emulated (I actually bought Magic Engine back then and was thrilled with the quality). There are some NES emulators that are commercial and (somehow) surviving solely because they're on platforms that are "difficult" to develop for comparatively (specifically Windows mobile).
All that said: I do respect your choice to go this route as well, but I politely suggest you have a backup plan for your survival/livelihood/income.
As always i'm very glad and thankful when you guys share your thought about my project. And here is my thought:
- I will not spend time to combat piracy for now. I'm open up to it. My software has no anti-piracy mechanism whatsoever.
- I just hope having the financial support to continue the work from whom truly appreciate it.
- Donate just doesn't work. I receive totally 85$ from 4,5 people; one is my friend.
- I'm a single tasked kind of man. I can not do it in free time while having a normal job.
- I prefer seeing my contribution as the 3Dification mechanism which is applicable for all kind of 2d tiled-based console system rather than seeing it as the emulation one.
- If it fails then it fails. I prepare for it from the start, will find a job then continue the normal life. No regret though i have already had priceless joy with it.
Anyone has any thought just share it. Truly appreciate!
Personally it's totally fine by me if you want to charge money for it. (Though unfortunately my interest level doesn't even remotely match the price you have set.)
I wish you good luck.
rainwarrior wrote:
Personally it's totally fine by me if you want to charge money for it. (Though unfortunately my interest level doesn't even remotely match the price you have set.)
I wish you good luck.
That's why i thought set it at 10 or 15 doesn't change anything
. Thank you though!
My thoughts are the same as rainwarrior's. No problem with the idea whatsoever, but the price is much higher than I'd be willing to pay for something like this. I'd have seriously considered pulling out the credit card at $10-$15. Have you considered a time-limited demo version? Maybe something that automatically exits after 5 minutes or so. Enough to give you a taste of how it works, but doesn't let you really play through any games. That might get more people to bite at $30 if they get a hands-on demonstration of what they're paying for.
Thank you for the suggestion. I'm still thinking about a win-win solution and everything is still open to consider.
miker00lz wrote:
My thoughts are the same as rainwarrior's. No problem with the idea whatsoever, but the price is much higher than I'd be willing to pay for something like this. I'd have seriously considered pulling out the credit card at $10-$15. Have you considered a time-limited demo version? Maybe something that automatically exits after 5 minutes or so. Enough to give you a taste of how it works, but doesn't let you really play through any games. That might get more people to bite at $30 if they get a hands-on demonstration of what they're paying for.
It *will* be cracked in a few days. What's the point?
FaTony wrote:
miker00lz wrote:
My thoughts are the same as rainwarrior's. No problem with the idea whatsoever, but the price is much higher than I'd be willing to pay for something like this. I'd have seriously considered pulling out the credit card at $10-$15. Have you considered a time-limited demo version? Maybe something that automatically exits after 5 minutes or so. Enough to give you a taste of how it works, but doesn't let you really play through any games. That might get more people to bite at $30 if they get a hands-on demonstration of what they're paying for.
It *will* be cracked in a few days. What's the point?
Maybe, but anybody who buys it can make it public too, so what's the difference?
miker00lz wrote:
Maybe, but anybody who buys it can make it public too, so what's the difference?
Exactly.
Hello everyone.
After receiving feedbacks from the community, i have decided to change the price model of 3DNes as follow:
- 3DNes is now available for FREE access with NO COST AT ALL.
- Who buy it at the price 4.99$ will have the PRIVILEGE to access to the latest build as soon as it is released
- Already purchased users will be REFUNDED the price difference REGARDLESS THEIRS WILLING
geod wrote:
- Who buy it at the price 4.99$ will have the PRIVILEGE to access to the latest build as soon as it is released
That sounds like Martin "nocash" Korth's business model.
EDIT: clarified who nocash is
can you clarify your point?
The newest version of
NO$GBA, NO$SNS, etc. is made available only to donors. The rest get the second-newest version.
I'm feeling so stupid right now. Do you mean everyone will go for the free version or what?
geod wrote:
I'm feeling so stupid right now. Do you mean everyone will go for the free version or what?
I think he literally meant that your model is the same as the guy's whose nickname is "nocash", thats all.
If that so then i officially become stupid then
geod wrote:
I'm feeling so stupid right now. Do you mean everyone will go for the free version or what?
No. Your model is also the same as
Cemu, or
Unibios.
Some people (like me) like to use the latest version of the software, and give the developer some money.
And people who doesn't are still able to use your emulator.
So no, you're not stupid
Good, now you just need to switch to Ardour business model: everyone gets the source code, donors get binaries.
By "feeling stupid" i mean i don't understand what tepples means. Too tired to dig into the guy's site, only watch his front page. It's partially tepples's fault for referring to a guy named "
nocash", if he refers to cemu then i can understand immediately
. Everything is clear now though.
I saw the early works you have show (Castlevania, a few months ago). I never thought it could be so beautiful as it is now. It seems to be more than an emulator, this is an art tool. Is it possible to replace the nes tiles that are being put in the 3d shapes with high quality textures? If yes, then we would have real works of art being made. The PC Build link is unavailable to me now, but I will try it as soon as possible. I hope you are planning to make an Android port, I would definitely buy it.
I wasn't familiar with Cemu (appears to be an experimental Wii U emulator) or Unibios (appears to be a multi-region BIOS for Neo Geo), but I was familiar with Martin Korth's products, so that's what I used as an example.
joabfarias wrote:
I saw the early works you have show (Castlevania, a few months ago). I never thought it could be so beautiful as it is now. It seems to be more than an emulator, this is an art tool. Is it possible to replace the nes tiles that are being put in the 3d shapes with high quality textures? If yes, then we would have real works of art being made. The PC Build link is unavailable to me now, but I will try it as soon as possible. I hope you are planning to make an Android port, I would definitely buy it.
Theoretically everything is possible even custom 3D art replacement. And Android port is definitively in my vision.
Is it possible to make some object colors glow? It would be amazing to see the coins in Super Mario affecting the ambient lights.
Also, sometimes objects merge with the background, making it impossible to put them correctly in the Z axis. This is an example from Captain America:
Attachment:
CapAmerica.jpg [ 140.97 KiB | Viewed 9162 times ]
If we (the users/editors of 3dn files) could turn off the "shape detection" you made, using the actual nes tiles, perhaps we could solve it. Alternatively, we could use some "background removal" routine to make these objects look correct.
I apologize for making so many requests, you must be very busy with the program code.
Sorry for late response. What you pointed out is just one of some problems 3DNes will have to solve in the future release. I would like to gather as many as possible issues to create a as generic as possible solution
By the way could someone help me with the following question. Performance wise, which is the ranking order of popular nes emulator?
I would like to ask this question because in long term i will probably give up the emulation part of 3DNes and only focus on the 3Dification part. 3DNes will become 3DRetro something like RetroArch that we could load different cores of different console systems.
Thank you in advance!
By the way (again) i would like to introduce to you some of latest achievements of 3DNes modders - the Mega Man series in action:
MegaMan 1:
https://youtu.be/Qzwq1vzlD9kMegaMan 2:
https://youtu.be/KK4sr3wOyOYMegaMan 3:
https://youtu.be/9G9nEBzA4B8
It is very good to know that you are planning to make your technology similar to Retroarch. Genesis and SNES games will be beautiful to watch.
Also, would it be possible to substitute the "layers" to a pixel by pixel attributes of "where it starts" and "how many cubicles of depth"? It would make the preparation of 3dn files slower and harder, but would solve some problems. Your "automatic shapes" (H Cylinder, V Cylinder, Cube, etc) would still be useful as starting shapes, but we could refine the models in the editor (it does not need to be an advanced 3d editor, a 16x16 2d pattern would become a 16x16x16 matrix, for example).
A "forceful editing mode" would be good too, a mode that stops playing whenever a new pattern appears and asks the editor to enter the attributes of the pattern.
Near perfect in MM1, average in MM3 and bad in MM2.
Answering your question on the performance, it runs ok in a Intel Core i5 3.1 GHz. In one of my notebooks (dual core, 2.3 GHz), it has a huge lag between controls and response in the game (about half a second or more).
I am not sure, but NES emulation is quite easy on performance (in almost every emulator I tried). The hard part on performance seems to be the realtime 3Dification.
I planed to do it in 4 days. In the end it takes me 4 weeks. Here is the release of V1.1 with Mac and Linux support. More responsive, less laggy with multi-threading switch.
Full change log:
+ Add Linux OS support (32/64 bits)
+ Add Mac OS support (universal 32/64 bits)
+ Move to multi-threading framework
+ Remove Wellcome Dialog
+ Add command line parameters support
+ Increase camera drawing distance
+ Decrease input latency
+ Improve overall performance
+ Improve 3Dification algorithm
+ Add various small bug fixes
Special Notes: based on the result of the survey of Linux and Mac support that i'm running from the beginning up to now -
https://goo.gl/forms/Iun9FIvxpnmJgAlY2 -, more than 98% users are willing to pay to have their OSes supported so the minimal price for those two portings is 4.99$
Roadmap:
+ V1.1.x Bugfix, Customizable input mapping, Virtual Reality
+ V2.x New 3Dification framework to solve the limitation of current one, 3dn filesize will become much smaller, 3d shape editor will become more powerful and smarter.
Final Bonus: video of Zelda II - The Adventure of Link
https://www.youtube.com/watch?v=ir0AafhVxlw
Version 1.1.2 has been released at
https://geod.itch.io/3dnes.
Among new features, save-state is now supported. It's really a game changer for the manual modding process.
Full change-log can be found at
https://itch.io/t/40466/winlinuxversion-112-release
Hello again,
I would like to inform you the release of 3DNes V1.2 with Virtual Reality, First Person View and Control.
Here is the feature preview video:
https://www.youtube.com/watch?v=Yk7X05mw6oUFull change log:
https://itch.io/t/50041/version-12-release-virtual-reality-first-person-view-and-more
Complete new experience from the classic nes game
https://youtu.be/ZkRdiSeKOZI
Version 1.3.0 has been released at:
https://geod.itch.io/3dnesChangeLog:
- Switch to core architecture => possibility to use another emulation core
- Improve cpu, ppu, apu emulation accuracy as a whole
- Fix input latency issue
- Fix sound lag, cropping
- Reimplement of mmc3 mapper
- Porting Blipbuf to C# => code base is pure C# now
- Add Vrc2-4 mapper (21,23,25,27)
- Add Action53 mapper (28)
- Remove border masking => now you can observe the scene from trully 360 degree
- Add instancing shader, better rendering performance
- Onepass virtual reality rendering
- Smoother shadow rendering
- Add credits for BlipBuf and BizHawk project
- Save/Load backward compability is broken
3DNes V1.4 - VR Version has been released:
- Release Video:
https://www.youtube.com/watch?v=GQu_LsxP1Kg- [New Feature][Shape Editor] multi shape processing (adjust, delete), multi shape merging
- [New Feature][Shape Editor] manually select tiles to create new shape.
- [Facility] Add 79 pre build 3dn file of popular nes games
- [New Feature][Pro] Vive - Oculus - 3D Monitor support
- [Change][Window][Linux] 3dn files will be managed at $3dnes_path/3dn
- [Bug Fix] Fix texture rendering bug in some nvidia cards. This is a regression bug.
Hi all,
Deformation and animation is a new feature of 3DNes V2.
Here is a quick recap:
https://youtu.be/Sf5f07TnHAkI'm more than wellcome to any feedback and suggestion.
Thank you.
Retina Render Mode : Inside the pixels
https://youtu.be/y8hWvN9gBeo
News on development.
3DNes : Mario - Pipes - Vietnamese Flag - Water World
https://youtu.be/lo3_lr87Nqo
3DNes V2 Scripting System Preview
https://youtu.be/LWeAW7ehckI
First/Third Person View Rework:
https://youtu.be/OB5t1XTi7tg
Crazzzy awesome stuff, as always.
I think FPV would make most games rather difficult to play though, maybe it'd be easier if there is an option to display the original screen in a tiny window at a corner so the player can occasionally peek (it'd be great to play in a more normal perspective and then show the replays in FPV).
Well, the mini map feature had been implemented long time ago. In this clip i intentionally turn it off to prove that we can actually play without peeking it with this new first/third person view mode.
Super Mario Bros - Remastered 3D Edition
https://youtu.be/v2yJKlBSMwA
Auto Tracking And Adjustment with Legend of Zelda
https://youtu.be/3sTXMr2cZ7M
The Legend of Zelda Remastered
https://youtu.be/XwEAAJ_5iGI
That Zelda's lookin' pretty neat.
Glad that you like it
I really like the inventory screen transition, and the dungeon rooms look amazing.
russellsprouts wrote:
I really like the inventory screen transition, and the dungeon rooms look amazing.
Haha the inventory screen transition makes me proud the most
Why haven't singers used nes cartridges anymore???
https://youtu.be/1AGhv7HTtl8
Circus Charlie - perspective game in 3d
https://youtu.be/CUWKwygYr2g
A 3-D game in 3_D? Interesting. The line scrolling playfield is a bit screwed up though. Wonder how the road in racing games will look like.
Probably screwed up in the same manner
3DNesV2 Binary & SourceCode Release
https://youtu.be/weRajsyLaEI
Journey of Silius Remastered with the help of a newly added Fceumm core for TAS playback
https://www.youtube.com/watch?v=97_FGC851R0
Ａｗｅｓｏｍｅ ｓｔｕｆｆ ａｓ ａｌｗａｙｓ.
Ｆｏｒ ｓｏｍｅ ｒｅａｓｏｎｓ ｉｔ ｒｅｍｉｎｄｓ ｍｅ ａ ｂｉｔ ｏｆ ＭａｇＭａｘ：
https://www.youtube.com/watch?v=xY5RWq3Goj0（Ｗｏｎｄｅｒ ｗｈａｔ ｃａｎ ｂｅ ｄｏｎｅ ｔｏ ｔｈｅ Ｆａｍｉｃｏｍ ｐｏｒｔ ｏｆ ＭａｇＭａｘ ｗｉｔｈ ｔｈｉｓ ｅｍｕｌａｔｏｒ．．．）
Ａｎｄ ｔｈｅ ｂｒｉｃｋ ｗａｌｌ ｉｎ ｔｈｅ ｓｅｃｏｎｄ ｓｔａｇｅ! Ｉｔ ｎｏ ｌｏｎｇｅｒ ｌｏｏｋｓ ｌｉｋｅ ａ ｂｒｉｃｋ ｗａｌｌ ｂｕｔ．．． Jｕｓｔ ｌｏｏｋ ａｔ ｔｈａｔ!
Is it possible to set some background tiles to use the background color in place of transparency? For example the brick wall background in level 2 looks odd with the bricks floating in mid air. Otherwise the game looks fantastic.
mkwong98 wrote:
Is it possible to set some background tiles to use the background color in place of transparency? For example the brick wall background in level 2 looks odd with the bricks floating in mid air. Otherwise the game looks fantastic.
Yes it's possible. Thank you for the feedback.
Mega Man Remastered - Some shapes did use background color for rendering
https://youtu.be/xhHcVPg6lL8
I do not personally like to play nes game in 3d but from a technical point of view, I find the result quite impressive and enjoy looking at the result of each video. Good work! The gears was nice.
I need to check how it is made, out of curiosity since it must require to some degree some work done by hand for each game, right?
You are absolutely right. Work by hand includes pattern parameters adjustment, some texture adjustment if needed and even some little Lua scripting for certain profiles .
Woah that hanging skeleton!
[joke]Wonder whether anyone will make a profile for Chiller...[/joke]
geod wrote:
That's awesome. Can you show what a full walk-through of Castlevania looks like now?
New Contra Profile with new capabilities from V3 engine
https://youtu.be/csZCD0OiwEs
Megaman 2 Profile With the Engine Rewritten
https://youtu.be/TQhxYoWuGgQ
Demostration of new 3d engine power
https://youtu.be/inw8EtFt9ok
Megaman 2 is looking pretty good!
Much better. The Airman's Tiki Heads need a rework. ^_^;;
Care to create a new video with the new engine on the TAS of Journey to Sillus?
The current work on Silius is a little bit better than then one in this video.
https://www.youtube.com/watch?v=5iH5ApC9NY0
Hey i would like to here your opinions about the mega man version above and this version. Which one is better?
https://twitter.com/tructv/status/1110901412244803584?s=20
