Minghan's Blog
Created by Minghan Lyu

Month: May 2024

Categories
NUKE

NUKE London Fall Assignment 02

This week Gonzalo suggested a few changes to my video, which I optimised and tweaked, and I created a couple of new effects, and next week I’ll start working on the near window flames and the explosions!

Firstly, thanks to the development of human civilisation! the almighty AI helped me to draw a couple of broken houses as well as some flame effects, which I exported later on

Then I did a Roto of last week’s ground breaking effect

I then imported the picture of the house into NUKE, then did a 2D trace of the original screen to generate the Matchmove values

Then I did a Rotopaint of the original frame and made it work to all frames

Then I adjusted the size of the smoke in the distance as well as the colour

Then in the middle of the lesson, Gonzalo said that it wasn’t enough to have this house in the picture, because now it looked out of place, so I made the house next to it a broken effect as well
And because the house was obscured by the foreground, I created an additional Roto.

He also told me to modify the penetration level of the foreground smoke, so I continued the foreground smoke through this chimney and now it looks more layered overall

Adjusted last week’s Roto, where it is now obscured by the picture of the broken house, and I Rotoed it back in the back

Then I added the effect of bullets shooting towards the sky in the centre of the frame, and looked on Github for an effect that automatically adds a lens vignette

I added a smoke effect to the back of the furthest house and did colour grading as well as Roto

Then I added a big bang at the far end.

Colour grading and screen tracking…

And I’ve also tweaked the debris of the Shard Mansion, I’ve removed a lot of the extra breakage effects and modified the foreground fog so that now it doesn’t have a black edge to it

Categories
Collaborative

Collaborative Project Process for EP 02

This post documents the individual progress of the last group project up to 23.5

Our team uses GITHUB to record every time a team member makes a change to the project and uses it for important questions (within the project). Here is the link: Issues · Sturmt1ger/The-Bag-Choice (github.com)

My main task this time was to complete the waterfall plant and its compositing in NUKE, I firstly used Bridge to find a model of the car and imported it into the software, then I created a liquid emission source.


The technique I used was a volumetric particle transformation technique, I would use the Flip element and solved it, and then transformed each particle into a grass and a plant by replacing them with particles.

Then I realised that the car model was hollow in the middle, so I imported it into C4D and changed its internal model. Because in Houdini I need to convert the FBX file to a volume file, the original car has no thickness, so I can’t generate a volume

Then I changed the particle emission source to a single one (otherwise the plant would flicker) and checked the particle normals

The normal must be vertically up, I changed the direction in the node

Then I manually created the nodes of the Karma renderer and imported an HDRI map as the main light source.

Then I made sky-specific changes to the main NUKE file.

I plan to split the altered footage into two parts, with the sky being grey in the first half of the destruction scene and clear in the second half.

I plan to split the altered footage into two parts, with grey skies in the first half of the destruction scene and clear skies in the second half. First I did a 2D tracking of the sky at two points to get the XYZ as well as the rotation transformation data

I then used the Keyer node to get the Luminance data and added Roto to the parts other than the sky to get a mask of just the sky parts

Then I put in an image to check my Alpha channel

Then there was a separate Roto for the face part of the character (because the character was moving a lot and the face was so bright that the Keyer node treated my face as the sky too = =)

Imported the cloudy day video file

These are the node trees that replace the sky

I then optimised it for edges, as there was now a white edge to the characters as well as the trees, so I found the Colour–Smear node in the NUKE Github, which makes it easy to remove the white edge

Before
After

I then added scene fog as well as colour correction so that there would be a colour contrast between the front and back of the scene

Categories
Advanced & Experimental

Progress Report on Personal Project 04

This post chronicles the progress of my personal project this week, featuring the addition of new VFX lenses.

First of all, I am going to add some effect of blood being splattered on the wall in the shot of blood pouring into the tunnel, I first used a MOV format, but it didn’t work as well as it should have, because my shot was created by simulation of mono-water, so the direct compositing would have been very fake in the scene, and if I had this shot composited by pure CGI, or HOUDINI plus MAY or UE5, it would have been possible to achieve a very good effect

So I was going to try a new approach to this problem next week, and then I created a second FX shot, this one a horizontal one showing the blood rapidly catching up with the protagonist. So I was going to try a new approach to this problem next week, and then I created a second FX shot, this one a horizontal one showing the blood rapidly catching up with the main character. The difficult part in this shot was adjusting the size of the blood motion blur in proportion to the speed at which the blood was moving, and eventually I found a good state of affairs.

Then I started to FX the last shot, I was going to create a first person view of blood, firstly I mediated the lens in HOUDINI

Then to handle the high speed movement of the camera I did 3D tracking of the shot in NUKE, while tracking I used roto to remove the figure to keep the GPU from tracking the figure thus producing a completely wrong point cloud.

Then I speeded up that video in Ae, because the final section of the video where the water is supposed to go down due to gravity and thus become faster, but the original video was at a constant speed

Categories
Collaborative

Collaborative Project Process for EP 01

This post began by documenting the making of the team project in the Exploration Practice course, and this post documents the first phase and some of the second phase of the production process and insights.

Our team uses GITHUB to record every time a team member makes a change to the project and uses it for important questions (within the project). Here is the link: Issues · Sturmt1ger/The-Bag-Choice (github.com)

Below is a summary of the core elements of our team project

Project thrust:
A short visual effects film about people’s choice of bags to show the importance of environmental protection and to stimulate people to rethink environmental protection.


Project Tools:
NUKE, HOUDINI, MAYA, UE5, PR, BRIDGE


Synopsis:
Two people meet in a café, A asks B why they don’t use eco-bags, B thinks it’s okay. A then tells B that it’s wrong through a series of initiatives and visual effects if the environment is destroyed.

Then I drew this storyboard

Next, after getting the original video, the first thing I was responsible for was the ROTO work on the VFX part of the video, as our VFX needed to be masked against the VFX with a person

I first used Mocha Pro to auto-key the character, it’s as easy as Photoshop, then I found some problems with this tool, it’s hard to track when the character turns around (for example, when the character turns his head, the whole tracking plane is displaced) I think the reason is caused by the software’s problems.

So I did an accurate tracing of the character in NUKE, then converted the NUKE data to MOV format and imported it into Pr, which has a plugin that allows you to take full advantage of the performance of the computer’s GPU to auto-trace the character, but you still need to manually process some colour similarities.

After completing the tracking, I realised that the system wasn’t recognising the part where Mahadev walked in from outside the camera, I then manually tracked this part in NUKE, I divided him into 6 sections for accurate tracking

Then I tried a simple masking in NUKE to test ROTO usability

I then created a node tree for smoke synthesis in the NUKE file so that my squad members could easily make smoke through this series of nodes

Categories
NUKE

NUKE London Fall Assignment 01

After three weeks of preparation, this is where I officially started to create the big term assignment that Gonzalo set at the beginning of the term. This assignment will use all the NUKE knowledge we’ve learnt so far, so I’m going to be putting a lot of effort into it too!

In preparation, I would like to add smoke, flames, shattering and explosions to the picture, and I’m thinking of making a shockwave of a nuclear bomb explosion at the end

Here’s a sketch of my intended effect

My main reference was wanting to replicate the nuclear bomb explosion scene from the game Call of Duty 4 Modern Warfare, the video below is the reference

At first I added a fog effect, I used ROTO and traced the scene

In the fog composite, I colour-blended the fog with the original image

Next I used the same technique to add a second smoke effect to the image. The original video for this effect was in MP4 format, so I first inverted the image and sucked in the LUMINANCE channel.

Next I added a scene breaking effect behind a couple of houses in the mid-ground, after the ROTO I used SHUFFLE and added three GRADE nodes to colour grade the model.

Then I added a fire-breathing effect on the roof chimney of the foremost house

But in the process of doing the flame compositing, I want to make the flame erupt multiple times and add a MOTIONBLUE node, but I found out that after connecting the MOTIONBLUR, the flame clip can only be played once, so I’m going to ask Gonzalo next Tuesday

This image has an empty alt attribute; its file name is d67576f258ce85ac18aae3ed7f48dbd.png

Here’s how I staged this assignment this week

Categories
NUKE

NUKE Fire and Smoke Compositing

This fortnight we’ve been learning about flame and smoke synthesis, and our teacher has asked us to complete part of the London Trap project next week!

Fire Part

Notes (VFX effects material site in the file)
1 When keyer snapping the a-channel, you can do a log2lin before to make the luminance data more accurate.
After compositing, if there is anything out of the frame, add a crop to cut it out.
2 When doing 3d tracking, you need to customise the length of the two points, just one, and then customise it in the scene of the node.
3 Add a motionblur when compositing
4 Keyer can select a display model and export it to alpha.
5 OFlow is used to change the video settings of the input source, such as the frame rate range and acceleration/deceleration.
6 When compositing, merge a max to add luminance information to the b (to give more depth to the result).
7 When doing flame compositing, the reflections on the wall are done this way, the original flame material is added to a curvetool, then reset and go! Solve for different frame data (you can see it in the toolbar), then copy the concentration data, then paste it into the place you want to deal with the grade, then paste it into the gain, then right-click and select Edit Expression, add *30. and it’s done. 5.14 email from gonzalo to see)
8 Synthesise the fire first step PLUS a blur flame to form the final fire edge (this has the most detail)
Step 2multiply back to density (reversed)
Third step PLUS on the fire itself (add a plugin version of glow)
And finally PLUS the flame itself after a glow!

Smoke Part

Notes

When creating smoke, if there is no A-channel in the original material, use the Luminance node to draw smoke from the screen.

If the clip needs to be cropped, use the Retime or OFlow node to control the number of frames in the video.

Then use the Grade node to increase the Gamma value of the A-channel in Luminance, so that the result has more detail and does not suck in the background data from the original material.

Then use the Copy node to make the A-channel stick to the original video, Blur the original frame, and Merge an average value to make the smoke blend into the environment.

Then add some MotionBlur or Defocus to defocus it.

Next is a magic operation, you need to re-UNPREMULT the original screen, then convert it to YCbCr format, and then stack it with the YCBCR data of the original screen, so that you can import the colours of the original screen into the smoke, this step is particularly important, even if it’s difficult to understand

Add a Lightwrap at the end to further blend the edges of the smoke into the original image

Categories
Thesis

Thesis Proposal: Week 5-6’s Course Feedback and Initial Framework for the Dissertation (research directions, preliminary report)

Weeks 5 and 6 continued with the importance of multisampling the reference section of your dissertation and we also had an offline dissertation session to help us better cope with our dissertation proposal. In Nigel’s course we continued to learn how to go about analysing our dissertation topic and how to construct an argument.

In offline courses, we analysed the types of research framework that can be done in our thesis and the research methodology we are going to follow to develop our topic.

Types of research

There are two types of research that we could use for our thesis such as:

  • Quantitative research. This type of research provides numerical data (analysed using mathematical methods or statistics). It is an empirical research that explains trends so it is more objective.
  • Qualitative research. This type of research produces findings without using quantitative methods and it is more subjective. It explores the perception, feelings, or ideas of people.

Critical thinking process and thesis proposal structure

The process to follow to analyse a topic in a critical manner is the following:

  • Investigate a problem thoroughly
  • Prosecute and defend an idea
  • Cross examine witnesses (literature review)
  • Verdict (conclusion)

This process will help us to structure our thesis topic in a logical way until reaching a conclusion where we will explain our findings.

Our thesis proposal structure will be as follow:

  • Introduction
  • Background significance
  • Literature review
  • Research design, methods, and schedule
  • Suppositions and implications
  • Conclusion
  • Bibliography

Methodology

Our research targets will constitute the basis of the methodology used. In order to explain the methodology used, we will need to follow this structure:

  • Describe thesis or research problem
  • Explain chosen approach
  • Clarify any uncommon methodology used
  • Explain how the data used was collected
  • Detail methods used to analyse collected data

Topic research

Brain Storm

1 Will the field of film special effects continue to move towards computerised CG compositing rather than developing techniques for realistic filming?

i There are many times when real life filming will have a better display ii Pure CG filming will be false iii Which is better for film, focusing on CGI compositing or focusing on real life filming?

Focus on CGI compositing

Pros:

Unlimited Creative Space: CGI offers unlimited creative possibilities to create any imaginable scenes and characters without the constraints of real-world physical limitations (Pearson CMG).

Efficient use of resources: Scenes can be created and modified efficiently in a virtual environment, avoiding the costly and time-consuming process of actually building and dismantling sets (Filmmaking Lifestyle).

Safety: for dangerous scenes or complex stunts, safety risks for actors and crew can be avoided through CGI.

Disadvantages:

Costly: High-quality CGI requires significant computing resources and specialised staff, resulting in high production costs (StudioBinder).

Authenticity challenges: Despite advances in technology, CGI effects can sometimes lack the realism of live action, especially when combined with live action, resulting in an ‘off-kilter’ look.

Audience Acceptance: Some viewers may prefer the texture and realism of live action, and believe that over-reliance on CGI can cause a film to lose its rustic and authentic appeal.

Focus on live action

Benefits:

Strong sense of realism: Live-action filming can provide a detailed representation of natural light and shadow and real physical environments, increasing the sense of realism in the film and the audience’s sense of immersion (Filmmaking Lifestyle).

Interactive performances: Actors tend to behave more naturally and realistically in real scenes, which helps to enhance the authenticity and emotional resonance of the performances.

Audience Acceptance: Many audiences are more receptive to real-life filming, believing that it conveys real emotion and storytelling.

Disadvantages:

Limitations: Live-action filming is limited by the physical environment of the real world, making certain scenes difficult or costly to realise, such as alien worlds in sci-fi films (Pearson CMG).

Costs and Time: Live-action filming requires finding suitable locations, sets, transport and crew scheduling, which adds to the cost and time of production.

Weather and environmental constraints: Live action filming is highly dependent on weather and environmental conditions, which can lead to delays or changes in the filming schedule.

Comprehensive Analysis

Balance and Combination

In modern film production, the best results are usually achieved by combining CGI with live-action shooting to complement each other’s strengths. The Avatar and Avengers film series, for example, create stunning visual spectacles through a seamless blend of live action and CGI, preserving authentic performances and environments while expanding the boundaries of visual effects through CGI (StudioBinder). The

Project Requirements Determine the Choice

Ultimately, the choice to focus on CGI or live action should be based on the needs of a specific project. If the film requires a lot of fantastical elements that can’t be realised in reality, CGI may be a better choice, while if the film is more realistic and emotional, live action may be more appropriate.

Taken together, modern film production often takes advantage of CGI and live-action shooting through the combination of technology and art to present audiences with visual effects that are both realistic and creative.

(Derive the argument above to generate the question) 2 The increasing development of visual effects has required more money to be invested in the production of a film, which must have led to fewer and fewer new companies or new directors joining the traditional film industry. What will be the future trend of this problem? How can this problem be solved?

i Marvel’s films over the past few years have tended to be homogenised, with the hero genre no longer being accepted by the public ii One of the films that has invested the most in CGI compositing over the last few years has been Avatar: The Watercourse, which had an estimated budget of $400 million. It relies almost entirely on advanced CGI technology to create the visual spectacle of Pandora.iii Netflix has a growing number of short films based on CGI compositingiv

Jedi: The Fast and the Furious (Bad Boys for Life, 2020): two new directors, Bilal Farah and Adil El-Abi, collaborated with several VFX companies, including Image Engine and DNEG, to complete high-quality visual effects

Summary

Thesis Research Topic

The impact of the development of visual effects on the entry of new companies and directors into the traditional film industry and future trends

Main Argument

With the development of visual effects (VFX) technology, the cost of film production has risen significantly, a trend that may limit the entry of new companies and directors into the film industry. However, through the democratisation of technology, the support of streaming platforms, collaborative production models and education and training, new directors and small companies can still find opportunities and success in this high-cost environment.

Abstracts.

VFX, CGI, cost of film production, new ventures, democratisation of technology, collaboration, independent film, streaming platforms, etc

INTRODUCTION

BACKGROUND: The importance of visual effects in modern cinema.

PROBLEM STATEMENT: High VFX costs are a barrier to entry for new directors and small companies.

Categories
Thesis

Thesis Proposal & Summary of Previous 4 Weeks’ Lessons

In the first few weeks, our focus was on researching our possible dissertation topics and analysing how to structure our dissertation proposal. As well as Nigel also told us how to construct a properly framed thesis, as well as giving an account of example arguments as well as the academic nature of the thesis.

The structure of a thesis would be the following:

  • Title page
  • Abstract
  • Acknowledgements
  • Contents page(s)
  • Introduction
  • Literature review
  • Methodology
  • Themed topic chapters (main discussion)
  • Results
  • Discussion or findings
  • Conclusions
  • References
  • Appendices

Before starting with the actual thesis, we will be doing a thesis or research proposal this term. This research proposal’s goal is the plan a structure for research like books, questionnaires, possible interviews, etc. It needs to explain why that research is necessary for the topic like for filling a gap in the subject, proving already existing researches, adding new knowledge to the topic, etc. Also, a literature review is important in our proposal to see if the material gathered is enough and useful for our thesis. It will let us demonstrate that we understand the key information of our goals. The proposal structure would be the following:

  • Introduction
  • Background significance (why research is needed and how it relates to field of study)
  • Literature review (introduction to sources we plan to use in our research)
  • Research design, methods, and schedule (planning)
  • Suppositions and implications (what results are we expecting and how are they going to contribute to the field of study)
  • Conclusion
  • Bibliography

In order to develop a research topic we should be asking questions like the following:

  • What motivates me to research and discover?
  • How can I connect it with my previous research?
  • Can this research impact the current knowledge of that topic and add new one?
  • Who is going to be the audience?
  • What is the reader going to learn from this research?
  • Will my topic meet the assignment’s requirements?
  • Will it be connected to my FMP?

Thesis objectives and structure

Before starting to write a thesis, it is important to set the targets, audience, and structure of it:

  • Determine audience
  • Determine what hit is the audience is going to learn
  • Break down thesis into main questions
  • Put questions in an order that makes sense
  • Develop sub-questions
  • Plan investigation two find answers

In order to build my thesis argument, I will need to use my own voice, meaning to express in a formal way but following my own style of writing. I could build my own views on its but not using my own opinions into it without any evidence to prove it. I should not use ‘I’, ‘my’, ‘mine’, only to be used when citing my own research. Before starting to write the thesis, it is good practice to try and sum the argument in a few words, so I can have it as reference to go back to check if I am still following this initial argument while writing the thesis.

In the chapters, it would be interesting to add counterarguments to our arguments that could at the same time argued again. This would prove that the analysis is not biased and it is taking in consideration all perspectives.

The conclusion should summarise the main points discussed on each chapter and should answer the question asked originally.

The methodology followed to research and gather information could be by researching peer reviewed sources, surveys, interviews, experiments, etc.

Literature review

This will help to orient ourselves and the reader towards what we want to cover in our thesis. A literature review is developed by:

  • Communicating relevant research methodology.
  • Showing theoretical framework we could establish.
  • Detailing where our work fits into the bigger picture.

When structuring a literature review we need to state our research question and explain how we tackled it. Then a body of paragraphs will explain the research in detail. Lastly, a conclusion section will reiterate the research question while summarising the insides. Within our body sources, we will need to identify themes, patterns, and gaps.

Once we have read and understood our sources, we will need to organise a strategy for writing about how we have used them in our research. We can organise them chronologically, by themes, or according to methods used by their author. Once this outline is done, we can start writing our literature review.

Categories
Advanced & Experimental

Progress Report on Personal Project 03

This chapter is the third report on the individual project Blood Flood. It consists of three main sections: 1 Further simulation of blood. 2 Ambient light sources. 3 NUKE synthesis

I’ve continued to import more accurate camera data into Houdini.

Checking the rendered result

Here’s the original video, you can see that there are five main light sources in the frame, which I then added to Houdini’s tunnel scene

Here the project files are all ready, I have increased the accuracy of the blood simulation by a factor of 10 here, and you can see that the result is now very much like the real scene (the simulation cache is also very large)

This is the particle effect of the model just now, you can find that the particles are super, just look at the effect of the particles is already perfect!

This is what the rendering looks like at this point!

Then I started the rendering and imported the result into NUKE, where I tracked the light source of the original video in 2D and transferred the data to the blood video

Then I added a Lightwrap attribute, a node that allows the high brightness of the edges of the blood to blend more with the original tunnel video. I then used a trick taught by Gonzalo to check exposure as well as saturated frames to check if the blood video was too bright? Or is it too saturated? If so, go ahead and add a Grade or ColourCorrect node to adjust it.

I ended up rendering the NUKE and editing the video in Pr

Here are the results of my video this time

Categories
Advanced & Experimental

Progress Report on Personal Project 02

This post is about 1: more fluid details and tweaking of renderer settings. 2: confirmation of final storyboards. 3: filming of the artwork and preliminary editing (including sound)

In terms of fluid detail, I continued to tweak the emissive source density, fluid viscosity, and surface tension of the water body; I still don’t actually have specific tunneling parameters at this point, but the tweaks to these nodes are permanent, and I can import them directly for use in a new model.

And I added two data saving nodes after solving, so I don’t need to ask the computer to recalculate every time, which saves a lot of time!

After I learnt about the KARMA renderer, I did an initial render of the scene, I added a darker EXR sky file to the background to for the dimness of the tunnel. Then I thought the result was still good at this point! But I realised later that I was too happy too soon, and every time I solved one problem, two new ones popped up!

Doesn’t the right side look good! Actually, the big trouble is yet to come!

What comes next is the most important thing, imagine if you had to blend a Houdini render with a formal scene?


1: Use the Karma renderer to render the parts of the body of water and mask out the tunnels, as it’s very cumbersome and ineffective to deal with in NUKE.
2: Export the water as a separate ABC file then import it into Maya for rendering
3: Direct rendering, not shooting the real scene later, using green screen to shoot the characters and composite them into the CG scene.


After thinking about it and trying it out, the only possible ways to handle it were 1 and 3, but at this point I had already finished drawing the storyboards as well as preparing for the shoot, so I ruled 3 out.
But there is a big problem in option 1, if I exclude the wall of the tunnel from the rendering, with the same time I have to not show the water behind the wall?

The most likely solution I tried was to split the water (blood), the wall and the emission source into three different Geo’s with different render numbers for them, but it still failed!

Then I asked Alwin (so reliable!) , he didn’t know either at first, but he started out by saying that Houdini’s default Mantra renderer could separate the two layers and render them very quickly, which I found out it could be done, but by this time I had already prepared the render in Karma and Alwin was happy to help, so he called to a friend of his from a VFX company and he finally found a solution with it! (It’s really hard to solve, you can’t even find examples in the official answer site)

Here’s what it looks like when it’s all settled, it’s so convenient! And you only need to add one node and write a command!

YEAHHHHHHHHHHH!!!

Here are the solutions

I also completed all the shoots this time and here are some images from the shoot. And since I needed to make a model of the tunnel exactly like the scene, I shot an extra video for 3D tracking.

I have to say that the Canon 6D MARK II performs really well in dark environments, and there is no excessive noise at all.

This video below is my initial edited work, and it’s already just under a minute long