**UPDATE** I just wanted to point out that I have been getting some great productive comments on this video and would like to highlight a couple of them: *Image based alphas - adding a bitmap node between the alpha image and the mask input of the camera tracker is a more efficient way of approaching image based alphas. - Thanks J S ! *Removing noise in Fusion - "In Resolve 17 the noise reduction tool in Color page is added to the fusion page call 'NoiseReduction'" - Thanks CoreSign ! Timestamps 0:00 Intro 0:50 Analyzing the shot 1:24 Camera tracker inputs and outputs 2:01 Adding a mask to the camera tracker 3:04 Image based alphas for masking Removing noise or grain before tracking 6:23 Using channels 8:02 Remove Noise node in Fusion 9:07 Using blur 9:42 Noise removal in Resolve Studio Correcting the image contrast before tracking 10:47 Using channels 11:22 Color Correct Node Lens Distortion 11:46 Automatic Removal 12:25 Manually remove lens distortion The Track Tab 14:19 Auto track, reset and Preview 15:02 Detection Threshold 15:27 Minimum Feature Separation 15:54 Track Channel, Track Range 16:33 Bidirectional Tracking 17:00 Gutter Size 18:25 New Track Defaults 20:38 Auto track defaults 21:14 Auto tracking the scene Camera Tab 21:37 Focal Length, Film Gate 22:07 Resolution Gate, Center Point 22:40 Pixel aspect, Auto Camera Planes Solve Tab 24:07 Solve and Delete solve 24:26 Solve Options 25:45 Seed Frames 26:58 Refine Focal, Enable Lens, Refine center point 27:35 Refine lens distortion, Enabled Lens parameters 28:55 Solving our scene, solve summary 29:54 Visually checking our solve with a merge 3D 31:20 Track Filtering minimum track length 31:53 Maximum Track Error 32:30 Maximum Solve Error 33:02 Select and Auto select tracks 33:40 Deleting trackers auto and manual 34:45 Refining the solve, refine summary 36:00 Trim previous and trim next 36:51 Rename and set color 37:15 Zero weight trackers 40:30 Export flags 41:41 Solve weight 43:45 Show names, frame range and solve error 44:13 Selected trackers Options Tab 44:42 Trail Length, Locator size 45:45 Colors for tracks, locators and export 47:47 Visibility 48:58 Colors and reporting Export Tab 49:35 Export and Update previous export 49:53 3D Scene Transform , aligning the camera 51:34 Scaling the scene 52:41 Real world scale 55:48 Ground plane, export and Previous export options 58:06 Export scene 59:43 Lens Distort Node 1:02:25 Adding objects to the scene 1:04:49 Add objects using point cloud 1:08:10 Overscan for distortion
Finally someone creating tutorials from a VFX professional point of view, not an enthusiastic I do it like this because I touch all the buttons this is what I get. Great. Congratulations
not gonna lie this was the best hour spent on a tutorial so far! every tracking tutorial shows you the quick and dirty tracking but now that i know how this node ACTUALLY works i can finally fix some scenes that have been really hard to track! amazing, thank you very much
@@prophetless keep up the quality content! you're helping (literally) thousands of people get better every day by sharing what you already know in an understandable and well structured way :)
WAO!!!! As a 3D artist who uses Blender and Davinci Resolve this have to be the best tutorial ever when it comes to doing camera tracking on fusion, a perfect explanation of everything!!!, thanks!!!!🔥👏🏼
Looking forward a course packed with quality like those tutorials, covering how to finish shots in a professional way, and showing different workflows. Livr the way tou explain, so count me in!
That was an hour+ extremely well spent. I won't embarrass myself by describing how I've been handling some of this tracking stuff when I've wanted to composite multiple objects into a scene, but now I know how it should be done! Saved, subscribed and looking forward to learning more!
One of those videos I have been waiting for years learning slowly Davinci… very detailed and I love it … I think thanks to you I have somehow yesterday figured out that it is possible copy data from one locator and paste to for example 3D text in 3D space. Which make very easy to place text without fideling about in cloud and trying eyeball where it fits using cloud. I always was a bit annoyed about how aftereffects although not real 3D space was much more friendly when placing objects such like text using 3D camera and it always took me ages to do the same in fusion . But with your tutorial it is another step forward and I learned a lot … thank you so much . I love all yours videos about fusion Have watch them all . There is no content like yours on internet. 🙏🏻 thanks
This is great! Once I quickly got past the level of the countless “make this thing follow your hand” and match-motion tutorials, I was happy to finally find your excellent in-depth video here.
Thank you very much. You do everything you can to clarify the information. I have purchased a course and I have obtained additional information from you that was not in the course. Thank you very much.. you are a very generous person
This is such a great resource for learning about the camera tracker in Davinci Resolve. This has helped me alot! Thank you for taking the time to put this together.
This must have been my best spent hour in many years. Thanks for the amazing tutorial. and for explaining EVERYTHING!.... guys at The Pixel Farm should learn from you on how to do a Tutorial.
Yes absolutely! I've done some videos in the past using the technique. If yo look through my older videos. But that's a topic for another video. I'll put that on my to do list.
Thanks a lot for this great video ! A whole day to study in detail. I'm a bit lost with the control of the viewport: Shift, ctrl for zooming and rotation are so not user friendly! I am now trying to export my camera directly to unreal engine... When I export, the 3Dcamera didn't get the good focal length and my 3d object where not in the good perspective. Do I have to setup again ?
This is awesome! Should we converting our media sources to linear colorspace using colorspace transform node before any delta keyer or camera tracking operations? In another tutorial someone said fusion loves linear more. But then in other places I've seen that these kind of tools like to work with more contrasty images, what's your view on that?
Thanks! For compositing most certainly Linear is better, as adjustments will work as you expect. For the camera tracking a contrasty image is better to pick up details for tracking, if you are not getting the results you need. I typically do not alter the image unless it's not seeing points well enough to get a good solve.
@@prophetless awesome thanks so much for replying. So if tracking is struggling maybe even putting a rec709 lut on first? Since I'm working with RED raw files
If you are applying it to the image, and not the viewer, that should work. You can always check with the "preview track" and see if the 3d tracker is finding more points.
Great tutorial! Quick question though. How do you determine how much overscan is necessary and what the resolution is for rendering out the image plane for 3d users in Maya? I'm used to Syntheyes and the scenes exported there have the proper render size with the new Undistorted XY value. Thanks in advance.
Thanks! Good question, I may do a video on this later today. I usually use Syntheyes too, but when I use Fusion I typically use the crop node to get the right size of the plate for export.
I just did a video describing these issues if you haven't already figured it out: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-6AeNYVwfx5U.htmlfeature=shared
This is the most helpful video I have ever watched on youtube. I've learnt so much from this video and the way you go about teaching it is phenomenal. I work with FPV clips so motion tracking them has been so much trial and error and a pain in my arse. I now feel i understand what each setting does now so i can achieve what I've been wanting to achieve! Thank you Sir. In my merge 3d view the tracking points are so small compared to the Video frame. How do you change this?
@@prophetless thanks for the fast response. Turns out it was the focal length being really small. One more question sorry, how can you colour grade the base clip without affecting the tracked text? Fusion keeps crashing when I try to camera track a comp clip within a fusion clip. And you can’t turn off the media in and simply copy a layer above in the edit page because the text looses the tracking data with the media in off. Tricky to find a workflow.
Sounds like you are trying to do the comp in the timeline. I usually do everything inside of Fusion. You have full control of the look inside Fusion. I usually match effects to the plate/base clip and do color last.
@@prophetless Ive done a colour grade in the colour tab because it’s way more powerful. So would like to do that to match the previous clip, which is just the non tracked part of the same clip.
Agreed. In Fusion if you match the 3D text to the Original Plate, color, vfx, etc... (May not have been very clear the way I stated it in the previous post, now that I re-read it). Then you add the color grade after. You could use a LUT in Fusion to make it easier to visualize before the actual grade. Are the clips not matching after you do the vfx on the one part?
Thanks so much for going over this in-depth! In DaVinci 17, I can't find the "render" button to export my camera to take an FBX into Houdini. I have watched a bunch of videos, everyone seems to have a render button that allows the camera to be tracked...not sure why I don't?
If you are using Fusion inside Davinci Resolve you could go up to the top toolbar and under Fusion look for Render all savers. I can't recall if the render button under the timeline was ever in the version inside Davinci Resolve but it is in the standalone Fusion. Let me know if that works.
Absolutely! I have a tutorial here that shows you how. It's a ttuorial on masking/rotoscoping but the method is using a 3d stabilized camera. ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-Za4CtMBDvUc.html
What happens if your footage is vertical? Specially if u go for a GoPro footage on a FPV drone. Do u at least have to exchange the aperture height and width?
The camera tracker should automatically pick up on the ratio of the image being input to track. So you shouldn't need to switch them, it should do it automatically.
Great work, thanks! Is there a chance to figure out what might be happening wrong with my workflow? On the step which is reflected in this video approximately on 1 hour 02 minutes. When you add a merge node to connect media in and exported tracking data I can not get point cloud to become visible on top of the original footage. Tried to swap inputs but all I got was a tiny wireframed plane alligned to the ground of the original footage. That might mean that my entire tracking scene is way too small, but I tried to scale it up significantly, and nothing had changed. I know that it is hard to figure out, but may be someone had already been facing such an issue and knows the solution?
Do you have make renderable turned on in the point cloud? That would be the first thing . Also if your renderer is set to software render the point cloud will be small, set it to OpenGL Render and you should see it. If that doesn't work let me know and we can figure it out.
I know this is going back s couple of years, but i still refer to this video from time to time. A question, if I may? At 30:18, a large, kind of, "box" appears. What is this and how can one remove? A project I'm working on has such a large and thick lined, box, I'm struggling to see what I need to see. TIA! 👍
That has to do with the scene scale. that is the camera path as viewed in the 3d scene. If you increase the scale of your scene it should get to a reasonable size
@@prophetless I have to admit that "scale" setting is a mystery to me, and I usually keep changing it to different values, and re-solving, until the region occupied by the locators looks sensible (compared to the camera, in perspective view, when sometimes thousands of them seem to be located "within the lens"). But, by the time the markers/camera look sensible, and the locators aren't hiding "within the lens", that box has very thick edges meaning that it's obscuring so much.
Yeah I feel like there should be an option to turn that off, and honestly there probably is I just haven't looked into it. I've been able to scale the scene up and that usually takes care of it, but I have also been annoyed at certain points with the same issue. If I find out a way to shut that off, I will let you know. If you end up figuring out a way let me know. Thanks
@@prophetless Thanks, and will do. Though, if I find it, you'll probably hear my shrieks of delight first ... all the way from the UK. 🤣 Similar with the camera track ... that can be the thickness of an anaconda ... usually when the box edges are thick. But doesn't bother me quite so much.
Hey William, I don't have an iPad pro with Lidar so I have not worked with it yet. But if you have export options I would think there would be away to get it into Fusion. Might need to use a second app to translate it but I would think there would be away. Could be wrong. What kind of options does it have for exporting out the data?
CameraTracker SOLVE is using CPU rather than GPU. Is this correct? Win10, Nvidia RTX 3080TI, AMD 5950x. Performance settings on in OS & Nvidia control panel.
This is such a great video, thanks for making it. I still struggle to get anything below a 2.0 Average Solve Error. I was recently shooting with a GH5s and a Olympus 17mm. I spent ages working out the sensor area size based on the different ways the GH5s uses the sensor and then also did minor tweaks to this too helped a bit. But always the Average Solve Error is around 2 or above. By the way I have gone through all of your video and made all the tweaks I can and still nothing is helping it get below that. Any suggestions would be really helpful if you have a moment.
Hey Chris, A couple of things I can think of, off the top of my head, are I think the GH5 has rolling shutter which could certainly cause some problems and another common problem is lens distortion. A 17mm lens is probably going to have a decent amount of distortion. Did you already look into those areas?
@@prophetless Thanks for your reply :) I hadn’t thought of the rolling shutter as I have the GH5s which is better than the GH5 in this area, but I will definitely look into that to see if that is causing anything. The lens distortion is suppose to be corrected in camera but I do let fusion refine lens distortion.
Rolling shutter can be really problematic even if it is just a little. Also it depends on what type of effects you are trying to create. Sometimes even though the error is high it might be useful for the effect that you need to do. I have had solves that were higher than 2 but the camera solve was perfect for what I needed to do. Hard to say without seeing the footage. And as good as the camera tracker is in Fusion it certainly isn't as robust as a dedicated 3D Tracker. Let me know if you can't figure it out. Maybe you can post the clip or something so I can see what you are up against.
@@prophetless thanks so much for your help. When I am back home after Xmas I will shoot a test video clip that I can use to upload somewhere so you can see. Thanks again 👍😊 If it is a little rolling shutter, I wish fusion had something like adobe does for correcting it, but I’m pretty sure it doesn’t yet. What I might try is a very slow moving shot so as to not introduce any rolling shutter and then speed up the footage to a similar speed, just to try and rule out any rolling shutter problem. Because that would then leave lens distortion as the possible reason like you said.
@@prophetless o dear...! O meant 32gb... May I ask what card do you have ... I have card with only 6gb and it seams fusion eats all my ram . I too have 32 but it just use a lot during work with fusion like sometime even 20 GB... Thanks for your tutorials love how they are detailed
@@prophetless Hey ... I appreciate you are taking your time to answer, thanks a lot. If you have a second, one more question. I can't afford to get more expensive graphic card so I have to stay with 6gb rtx 1660 . However would it make a difference for fusion Ifat least upgrade ram from 32 to 64?
It will help depending on what you are doing but have you looked into your settings on how much ram is used by fusion? Also using proxy for while you are working can help a lot if you are having issues. Just some thoughts that might help instead of spending more money.
You may want to look at this video: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-Za4CtMBDvUc.html Discusses how to add masks to objects in 3D space.
The lens distortion at the beginning of the script is to un-distort the image so we can see things correctly in the 3D view and the one at the end, before we merge back over the original clip is to distort our 3d objects to match the lens distortion of the original clip. Here is another video that might explain it a little better starting around the 2:30 mark: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-6AeNYVwfx5U.html
Hey around 37:49 when I try to track my blue points within the rectangle mask; it ends up tracking everything within the rectangle but doesn't stick to the cage. The rectangle stays in the same position on the video and just creates blue points all over the right side. how do I fix my mask to track the cage ?
I had animated the roto shape before hand to match the area of the cage. So you would need to animate the shape if you are following along. Should only take a few keyframes.
Thanks! Add a card/plane in 3d space where you want the object and attach the png file as a material. You'll probably have to scale it unless you are using the projection method.
@@prophetless thank you my friend ! Is this the only way ? And if I want to add a non 3d shape ? The same method ? Need to add plane as with the picture ?
You can add the text to an image plane. Similar to how you would add photos in this video: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-P3pvqnMIgkM.html
@@prophetless thanks yes I discovered this yesterday, but I also noticed the image plane rasterizes the text so it's better to increase the size on the text node instead of the image plane node
Did you check out these videos: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-eDYCoe_Fa6U.html ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-qmPXPF9PzqQ.html I don't use blender that often but I can look into it if these methods don't work.
@Prophetess, Hi Brian, thanks for this awesome tutorial, I’m new to editing, new to Davinci , new to you channel and a New Subscriber 😊 Brian, I’ve attached a link, do you think that this could be done in Fusion ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-IwaLOJAHAGg.html I’ve attempted it many times please see link below ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-JzeLkBGgd_s.html but I found it very time consuming try to find the Polaroid after attaching it to the plane node? Any suggestions would be greatly appreciated and thanks again for this awesome tutorial 💯‼️
@@prophetless I forgot to add that my Polaroids were not stationery as they were in tutorial…. Thanks again looking forward to Fusion tutorial, have a great weekend 👍🏽
If you want to animate them, besides the camera move, you can animate them by key framing the polaroids to move around. Should be pretty easy once you have it set up.
@@prophetless I actually wanted them to remain stationary like they were in the After Effects tutorial. In the video I made I notice that they kept moving. Please send me a link so I can buy you a coffee, I appreciate your help
