Osipa, J. (2007) Stop Staring: Facial Modelling and Animation Done Right, 2nd Ed.
Sybex.
FLORES, V.C. (2005) ARTNATOMY. [Internet] Available from:
< http://www.artnatomia.net/uk/artnatomiaIng.html > [27 February 2011].
Subdivision Modelling. [Internet] Available from:
<http://www.subdivisionmodeling.com/forums/showthread.php?t=907> [Accessed 27
February 2011].
Sunday, 27 February 2011
Final Sculpt
After two weeks working on the digital sculpture of Dominic, it is finally complete. From a personal perspective, the growth has been huge this unit. On the digital sculpting I have progressed from no knowledge of Mud box -to a relatively high level, while Topology (seemingly terrifying at the beginning) is now a term and theory I not only recognise, but can also map out my own models with.
Of course with any steep learning curve there were issues (one such being in the sculpting of human mouths) but each has been overcome either taking advice from others on board or working through it. The final sculpture lines up almost perfectly with the reference photo – as shown below, which I was incredibly impressed with, considering using a method of sculpture which required estimations of perspectives– rather than layering a stencil directly over the model.
Unfortunately, since most of the time was spent modelling and researching topography, two things I did not get to explore was the resource “Stop Staring: Facial Modelling and Animating done correctly” and also the area of texturing. The final image below does have some level of texture – but it was done via projecting the photo directly onto the model. This is not really a texturing technique and was only done to showcase the model in an optional fashion. However in this case it is not a huge loss as the sculpting technique and understanding of topography are the key aspects being marked. Still with facial animation/ modelling and texturing being important areas that are sure to be revisited at a later date, I will likely look at the two more – even if out of term time.
Friday, 25 February 2011
Topological Preparation
Thus far my understanding of topology as a concept is somewhat solid – especially since the Monday lecture. But the final step in Dominic’s digital sculpt is to apply this concept onto the face directly. This is something which will require much more than an understanding of the concept so before I begin work on a topological mesh, I will look again at two resources which have been somewhat underused and understated in my blog thus far but will be crucial assets for the task. These are:
Artnatomia is a brilliant website which allows the user to look at a generic human face in relation to the placement of the skull and also the underlying muscles – being able to freely move between the two using sliders:
The reason that this is an important tool for the construction of a topological study layer is that the shape and prominence of the facial muscles and skull shape essentially craft the topology of the face – and in turn the mesh. By identifying those muscles in my own sculpt, and my model Dominic, it will mean that I can identify how my mesh should flow.
Still beyond knowing where a mesh will flow, it is also an important to know how these meshes should look. This is one of the reasons subdivisionmodeling’s forum discussion is important. The forum discussion is full of topological examples – which highlights the flow of particular areas in a mesh, in addition to how the use of ‘Poles’ can manipulate that flow.
(Two types of Poles - E: (When a square is protruded from a flat object you get four E poles (one at each corner)) and N: (“The nose pole”))
Taking this information, I looked at my model and identified the key muscles before placing the mesh over the top – making sure that it flows well.
This was my final topological mesh:
Thursday, 24 February 2011
Thursday's Modelling
Wednesday and Thursdays modelling sessions were varying in success. Wednesday I largely worked at home, but through frequent software closing issues did not get much that I had done saved.
Because of this while doing the necessary tweaks on Thursday, I saved frequently – meaning that most of the edits done can be clearly seen:
 |
| Using a large stamp brush on low strength, I attempted to add texture to the model - but this ended up adding unneeded detail and age to the face |
 |
| Began bulking out the cheeks and area around the mouth |
 |
| Bulked out areas smoothed - face starting to take shape |
 |
| Evened out the hair line - an issue which had been bugging me for a while. Lowered corner of mouth for a serious expression, more smoothing |
 |
| Final detail and material in place |
This final sculpt is almost perfect and aside from the topology layer – which will be addressed tomorrow, the digital sculpture is done. Looking through the steps taken, sculpting can clearly be attributed to not taking any kind of compromise over particular features. Even now looking over it, I think that tomorrow I will also edit the eyes as they are not quite as defined as I would like.
Tuesday, 22 February 2011
Tuesday Modelling
Today’s session was an independent session in the media lab, meaning I could solidly focus on getting the face sculpted – beginning with yesterday’s issue of lips. After seeking advice from others modelling, I made several realisations about modelling lips correctly.
The main element I had not taken into account was the rise from the bottom of the nose to the tip of the upper lip. This rise for inexperienced artists is often ignored/misconstrued as the upper lip rising – but in fact the upper lip begins at the dip. By knowing this it meant my mind-set for crafting that lip was different and the result was an accurate recreation of the lip.
The bottom lip was next and again the crucial area to think about was the maximum point of ‘rise’ in that lip. Once I had worked that out the lips had been almost perfected.
It was at this point that Dom, my facial model once again appeared – meaning I could photograph him again – this time with eyes shut. These were the two shots which would be the reference material for the digital sculpt:
With these photographs, I could safely begin modelling eyes knowing that the position of the eyebrows would also be accurate. These came out really well – however from the profile view there was a definite issue.
With half a week still to go, I feel that there is only a small amount which needs to be done to complete Dominic’s facial sculpt to a high standard.
Monday, 21 February 2011
Mondays Late Night Modelling
After the last post, I decided to begin work on Dominic’s sculpt early so I had a solid foundation to continue work with on Tuesday. Using the reference photos taken on Wednesday proved somewhat difficult as I had not told Dom to close his eyes – an action which would undoubtedly cause a number of movements in the eyebrows. Because of this I couldn’t start work on the eyes so I focused on working with the nose and mouth. The nose was relatively easy to sculpt correctly but the major issue was the mouth – which despite numerous attempts I could not get accurate – or even on some level successful. Still, the nose was a solid basis for the work tomorrow so I was happy with this late night session.
(Update: would upload a photo of this stage but deleted the file by accident [24/02/2011])
Monday's Topology Lecture
The morning lecture today was largely based around explaining the concept of Topography. Having been asking around over the weekend about the topic, most of the responses related Topography to the geographical term for contours on a map. This meant that I roughly knew that it was going to be a similar vein to that response – but in relation to the contours on the human face.
Marks explanation was almost this. Topography is about the understanding of contours on the face, and the overall flow of the face around key areas such as the eyes and mouth. By knowing this, the mesh can then be converted into a form which reflects this. Why is this necessary? Well topological meshing means that low res facial models can retain a huge amount of detail as each face is optimised in its use – a crucial property in the games industry for placing characters into a game engine. The more polygons a character has, the more the computer has to work – meaning that keeping a low count can be crucial on a game project. Furthermore, simply understanding the presence of topography on the face can improve an artist’s work at a fundamental level. From my new found understanding of Topography means now I shall be able to revisit the link from before and use it to aid me.
From today’s lecture it is clear three tasks still await me:
-Start and complete the high-res facial sculpt
-After Completion, analyse the topography through a paint layer
-Continue to study topography.
Tomorrow’s Media lab time means that I will be able to get started on the high res facial sculpt, while Stop Staring is a resource I shall read on and off this week, dependant on if any issues arise during the modelling stage.
Thursday, 17 February 2011
Experimenting with mudbox/ installing mudbox at home
Today I did not manage to get much more work done, as I had to organise installing Mudbox onto a new computer – a task which while not offering much in the way of short term benefits, did mean that next week I will be able to work outside of studio time to optimise the time spent on my final sculpt.
Still despite the set up taking up much time, I created another 2 heads on Mudbox. The first one was not my greatest work but much like my other models, it had a particular focus. This was wrinkles - Small or otherwise.
Key criticisms: Triangular nose, eye definition not perfect, bumpy texture, muscles not finished.
The second head was focusing on asymmetrical features and pushing the face into obscure expression – especially when it came to the mouth and eyebrows. It was not particularly created for a realistic purpose so I can’t really reflect on the look – but it showed me how not mirroring entire faces can add an entirely new dimension to the feel of a face.
Wednesday, 16 February 2011
Additional Mudbox Work/ notes and reference photographs for the final sculpt
Today I got into the studio at around 10:30, which was fortunately in the middle of Mark’s lecture to another group – meaning I had ample time to note down the procedure for superimposing a reference image into Mudbox. These were the basic notes I took:
Go to Image browser
Set image to camera
(Options) Depth (in front or behind mesh) and Visibility sliders (opacity)
Transform using the select objects tool (making sure to keep the exact same scale!)
Lock camera/ camera bookmark to get the profile and frontal comparisons on easy access)
Grab areas of the face – line it up (use mesh as guide) – don’t be picky (rough guide to shape)
(A neat trick on Mudbox)
Base head – mirror (select face tool) to select the side of the face to mirror
In addition to these notes, I once again created two heads. ‘Alice’ the first head I created was a recreation of a character idea I had. Considering the sculpts that I have made thus far being all male, it was a good opportunity to attempt a female face and get in much needed practice with the software.
Overall, despite somewhat disappointing lips, I was impressed with the end sculpt – especially with the detail to the back of the head and hair. Still the overall focus seemed to drift away from the facial detail and linger more with the hairstyle. I was especially guilty of this with my rough sculpt of Final Fantasy character ‘Snow Villiers’ (the second sculpt of the day).
Given the over the top hair and bandana, the hair should not had really been attempted. Still after mapping out facial features and positions in accordance with the images available, I tried to get the hair and bandana perfect – which was not particularly successful. Below are the images of Snow and My Snow sculpt (with or without the bandana/ hair).
Fortunately, towards the end of the day, Dominic – my facial model from the course, had come in to work on Mudbox. This gave me the perfect opportunity to take the reference shots for the sculpture next week. These are the photos taken:
Tuesday, 15 February 2011
First Mudbox Session
Today Mark introduced the fundamentals of Mudbox modelling to the group and within minutes, we were set on the way to working with the software. From that session, this was my method of working on a face without any reference materials:
The first step is to turn on mirroring when modelling the generalities of the face. Select the Mirror option and set it to ‘Tangent’. Because of this each edit will now follow the exact symmetry of the face and affects both sides in identical ways – even if the face is asymmetrical (to a degree) Furthermore Tangent will still follow that symmetry should the head be moved from the origin. X, Y and local Mirroring can be chosen but each has its own issues. Keeping the habit of Tangent Mirroring will help in the long run.
Beginning with a lowest polygon count, select the Grab tool. Setting the brush size to be relatively large, begin to morph and change the base head shape to what is necessary or wanted. Aspects such as the base nose shape can also be edited – but the freeze tool should be used to make sure that other details are not affected.
Once the head shape is correct, using Shift + D, the polygon count should be raised to around level 2 (32000) so that the figure can begin to be shaped and detailed to a high standard.
Using the Sculpt tool (setting it to be medium or large sized and of low strength) begin mapping out details such as bone placement, muscles and eyebrow shape. By using CTRL + Left mouse click, layers can be dug away using sculpt. This is crucial for creating Nostrils, Ears and the Mouth. For the most part Sculpt can be used when modelling in facial detail – although at points where layers need to be built up, the wax tool can be used.
(Note: Most Functions on Mudbox have 3 ‘modes’
Their intended action – Left Mouse Click
The Opposite of the action (removing, etc.) – Left Mouse Click + CTRL
Smoothing – Left Mouse Click + Shift )
Once all of the detail, including the muscles around the throat and torso, has been built up, the Pinch tool can be used to tighten up edges of detail – giving them a much firmer look. For faces which have many wrinkles or hard lines, Pinch can raise the quality of the face massively.
By now the face should have all of the details wanted. Once other areas have been addressed (shoulder blades, Spinal Column, veins if needed, etc.) the hair can be built up. Unlike in detailed drawing suites where each hair can be mapped into the work, in Mudbox this is not the case. Should the hair be a pressing concern, then the mind set when constructing hair should be to map out the hair shape, and then detail.
Before this, a new layer should be added, so that the hair can be removed and added at will. The Add New Layer option is in the top right hand corner of the default layout and works in a similar vein to Photoshop layers. New layers can be added, removed and the visibility can be turned on or off.
Once the layer is added, the next step is to build up the hair. For this, switch off Mirroring – else the number of hairstyles is vastly restricted as very few are exactly symmetrical. Using the Wax tool, strokes should be layered on where necessary – whilst taking into account crown placement, fringes and facial hair. After this base coat is added, use the Spray tool – on small brush size and reasonable strength (making sure that the stamp option and stamp placement are off) to add all of the ‘lines’ and markings of the hair. In doing this a sense detail begins to creep into the work. Be sure to ‘take away’ lines in addition to adding them to give the hair three layers of detail. Use the ‘smooth’ mode to clean up lines and give a smooth finish.
In addition to these steps there is much more that can be done in Mudbox. Stamping texture, changing lighting and materials are all steps that can be done post production and I am sure that there are many other methods of working or ways to use each tool. Still with another week or more, I hope to be able to add or change this working method to progress my work even further.
During the session I created two heads. The first I felt went well. The detail on the face was reasonable and overall I thought the outcome was excellent for a first try. The hair was slightly flawed in its manipulation and the lip profile highlighted some issues that I had with lip creation.
The second head was much better. I attempted to gain a much better grasp on the lips and also I chose for the character to appear menacing. To do this I focused on the eyebrows (which are both prominent and lowered), while creating a solid overall face shape (most noticeable with the solid jaw and chin). I also added two layered ‘flaps ‘around the eyes which heightened the menacing feelings given off by the character (this technique is often used for film villains (most notably the Emperor in the blockbuster, Star Wars).
Tomorrows targets: Practice more with Mudbox, note down procedure for superimposing an image onto the screen for reference, take reference photographs of Dom for high quality sculpt
Monday, 14 February 2011
New Project: 3D Digital Sculpture
The latest unit was shown to be Digital Sculpture. Much like the animation unit, for the next two weeks studying anatomy is crucial – especially as the task and outcome of the unit are as follows:
- (Task) complete high quality 3D model of another student, complete with a topology study layer which makes reference to the first outcome
(Include a ‘turntable spin’ animation with the 3D head)
- (Outcomes) A thorough understanding of Topology
Understanding and familiarity with the 3D modelling software Mudbox
I am expecting that once again there will be a learning curve (whether steep or not remains to be seen) - which is further amplified by the short time frame to complete the project in relation to large amount of work needed. Nevertheless, given tomorrow’s scheduled modelling session and numerous available points where the studio can be made full use of, I am not too unnerved by the challenge of completing the 3D model in time.
However Topology at this point in time sparks absolutely no understanding so this is something I will have to look at between modelling this week. I will also ask tutors and technicians about the procedure of creating a Topological study layer for my work. At the moment these are my only concerns – although time remaining is always a factor. Still, with the animation project complete and Contextual analysis moving along at a steady pace, the workload is still manageable.
Saturday, 12 February 2011
Bibliography
Video links used/ referenced in the blog:
(2008) Principles of animation from Ice Age 2. Available from:
[Last accessed: 4th January 2011]
(2009)Ice Age Trailer 1. Available from:
(2008) The Cat Came Back. Available from:
(2009) Pixar’s Short Film For the Birds. Available from:
(2009) Ichigo Vs Byakuya English Part 1. Available from:
(2007) Dynasty Warriors 2: Xiahou Dun part 1. Available from:
(2007) ‘Pixar tennis commercial’. Available from:
(2009) Titanic: The Animated Movie (English) Part 1. Available from:
(2007) Ratatouille – Trailer. Available from:
(2008) Ratatoing (Part 1). Available from:
Last Window: Secret of Cape West Gameplay. Available From:
(2010) The Worst Movies Ever – The Frog Prince. Available from:
(2008) Animation Basics - Bouncing Ball. Available from:
(2010) Pain Vs Naruto part. Available from:
(2008) Burnt Face Man 7. Available from:
Walt Disney Animation (2009) Steamboat Willie. Available from:
(2007) Felix running home to dinner. Available from:
(2009) Burnt Face Man. Available from:
[All of the above last accessed: 9th January 2011 ]
Books read:
Books read:
Wiedemann, J. (2007) Animation Now!. Tashen.
Williams, R. (2001) The Animators Survival Kit. Faber & Faber.
Friday, 11 February 2011
Reading Week Animating
This week was largely studio based, with most of the cat movement and around 4 seconds of ‘Doll’ movement to cover. Mondays session was incredible, in terms of work covered – with around 3 seconds of cat movement animated (Cat runs, hits box and box tips over). The work method was very much pose to pose for the cat run up to the box, although at this point I deviated largely from the storyboard made in pre-production and chose to work Straight ahead. The previous plan was that the cat would hit the box, paw at and drag the doll into it and proceed to drag it out, I decided that actually the Cat would knock the box onto its side, which would propel the doll from the box and the Cat would climb out of the box and look at the camera. This decision was largely chosen as it might not be possible to do all of the actions of my original idea in the time frame available.
Of course, having the cat climb out of the box was an important step, so I set aside 3 hours on Tuesday to make sure that I got that action perfect. Unfortunately, this proved to be difficult and issues with not getting the correct scale or motion meant that while I had a better understanding of how to move forward, my animation did not physically progress at all. Still with the ruins of the animation Tuesday, it gave me a solid basis for Wednesday’s animating.
Photoshop had unforeseen issues on Wednesday, which caused the work which was done Tuesday to corrupt. This wasn’t a huge setback as I had been saving work on a regular basis in case of such issue. Still after discussions with the media lab technicians about the problem seeming to have a no solution to the problems that I have been having, I decided to be much more vigilant when saving – making sure I could open and close an extra save file before closing the stage I was at.
Despite not having the framework build on Tuesday to work with, I managed to get the Cat out of the box, sitting down and looking at the camera. Again the movements of the cat as it squeezed out of the box was very difficult to animate – at points the cat seeming to just move on the spot with far too many steps in a short period of time. To rectify this, I moved the box back slightly to reveal the legs. This worked well as when asked, everyone felt that the animation was smooth.
Thursday was the final day I spent working on the Cat animation. I focused on animating the doll, which was done on a separate video layer. The first time I attempted to animate the doll flying from the box, the speed was far too high – meaning that the end result was a bad example of secondary action. Before I could alter it after lunch, the file had corrupted. Again, like with the Cat moving out from the box, the second attempt was much better as I captured the arc and timing well. The doll then landed in a bowl and follows the various panning of the ‘camera’.
This was the end result.
All that needs to be done to finish this project, is placement of audio in the animation – which I will try to get done some time over the weekend or early next week.
Friday, 28 January 2011
Friday Lecture - Tim Searle
Today was probably the most interesting lecture thus far as an experienced animator, Tim Searle, had come to discuss the UK animation industry. Firstly, having not known Searle by name, it was amazing to discover that he was behind the well-known opening title animation of the panel quiz show: Have I got News for you.
Currently working on the Horrible History animated series, Searle has been behind (or part of the production team) on a number of British animated projects – including a George Michael music video (Shoot the Dog), I am not an Animal and 2D TV. It wasn’t specifically discussed, but Searle’s lecture did highlight how commissions are both the most important prospect for a commercial animator and can also be the proverbial rug that can be pulled from under their feet. For example, despite being hugely popular with an audience of about 7 million, 2D TV’s commission was dropped when due to a switch in scheduling (to Monday night) as those numbers dropped dramatically. I am not an animal suffered a similar fate.
Despite this, Searle remains positive about the state of the UK animation industry, especially since he has worked with, and has aided the development of several ‘up and coming animators’. Furthermore at points in the lecture, he showed the works of several young talented animators of which he has much respect for. Searle also highlighted how important Comic timing is to a good animation – pointing out a student’s ‘astronauts’ animation as a good example of this.
One inspirational figure to Tim Searle has been John Halas. Tim showed the following quote, just to emphasize this:
“Animation is the most contemporary form of human expression, combining the elements of motion, Storytelling, Sound and Space”
Having been working with animation for several weeks now, even at the most basic of stages, it’s clear how much an animator does have to understand all of those elements mentioned in the above quote. Expression, motion, storytelling, sound and space – all of these have to be purposely constructed by an animator. Tim Searle’ visit has been a great insight into how fickle the world of animating is (but having been studying the Games Industry, they both share that feat) but at the same time, just how powerful the medium is.
Thursday, 27 January 2011
Thursdays Animating Session
I began my project in earnest today. With the rough storyboard made yesterday, I had a rough idea of what I wanted to produce today – a simple walk cycle at the start of the animation. This would be before the audio clip begins. Of course I did not know how much work I could possibly do in a day so there were a number of crucial lessons that were to be taught today;
-Cat movement
-Work Rate
-Design
Working out the movement of the Cat in the animation was a somewhat tough task. Before I began to start, I recorded several movements of my own kitten as a reference material. This was a helpful ‘crutch’ in that should working out tough movements prove to be troublesome, that footage could be used as a cement hint as to where to go from there.
The general work rate was around 6 hours solid work equalling 20 frames. This was somewhat terrifying in a way, but hopefully I will be able to pick up the pace over the next two weeks. With a total of 6 seconds for the audio, plus the 2 seconds before the animation, this should mean less than 6 days’ work to finish the animation.
The design and work rate really went hand in hand. I chose to keep the design as simple as possible. Given the complicated idea that I hope to execute, combining that with a complex full colour design would be impossible to execute in the time frame that the project runs for. So I chose to very much model the cat design on my sketched out storyboard – a simple black and white silhouette. This means that I can focus any spare time on keeping the animation as smooth as possible, rather than colouring objects.
I was impressed with my final result. The only notable issue was the shadowed back leg’s movement as it ‘jumps’ through three steps and does not look fluid.
Wednesday, 26 January 2011
Animation Storyboard
My idea for the animation is pretty simple. First a cat is walking, and then notices something out of screen. The animation pans to a box with a small doll on the top and back to the cat. When the camera has panned back to the cat it is in preparation stage for a pounce. The sound bite begins:
“Globo …he’s uh, he’s tough, he’s a brute...”
Cat pounces into box and hits it – knocking a pull cord from the doll into the box
“But he’s got a good heart…”
Drags the doll into the box
“And at the end of the picture… he saves the girl”
Cat drags the doll out of the box (looks at camera?)
Here is the storyboard which runs coherently with the above description.
[Possible points of research: My own cat (record movements and study) and Felix the cat advertisements]
Tuesday, 25 January 2011
3D Character Animating
Moving from animating a 3D bouncing ball to a 3D character rig, from the first time animators perspective seemed a unique combination of slightly intimidating and intriguing at the same time. Nevertheless, after the introduction on Monday and Today’s animating tutorial, the task of making a character move across the screen is much more appealing.
Much like the movement of a bouncing ball in Maya, moving a character rig follows the same strategy – albeit with more controls. The character rig had several controls to work with;
Between 1 and 3 on the Arms
3 on the torso
Between 1 and three on the Legs
And head controls.
The reason for the alternate controls on the arms and legs was due to two different ‘modes’ of movement available; Forward Kinematics (FK) and Inverse Kinematics (IK). Forward Kinematics is very much a hierarchy of movement. An example:
Top of arm > Moves: Top of Arm, Bottom of Arm and Hand
Bottom of arm > Moves: Bottom of Arm and Hand
Hand > Moves: Hand
Both Kinematics designate a particular control as a ‘Parent’ and the rest as the ‘Child’. The Parent’s movements influence that of the child, while the Child’s movement has no influence on the Parents. In the case of Forward Kinematics, the top of the Arm is the Parent in that it has the control over all other areas of the arm, while the bottom of Arm and Hand are the children. Of course this structure can become complicated in more complicated rigs and feature multi-layered levels of Hierarchy. In the case of the ‘FK’ Arm while the top of the Arm controls the bottom, the bottom of the Arm still retains control over the Hand – so even in this basic rig, the rigger has used a multi-layered structure.
Of course for Inverse Kinematics, this logic is flipped on its head. In the IK Arm hierarchy the Parent is in fact the Hand. By moving the Hand, the rest of the arm compensates for the movement – however extreme it is. By using Inverse Kinematics, it can make quickly moving an arm or leg into position very easy, but it has a few issues. The first is the fluidity of the end animation. By moving all areas of an arm at the same time, the animation will likely not be as fluid as the computer will choose the most direct route to the end ‘goal’. This could mean that an arm which is meant to swing naturally into position might simply move diagonally into place.
Another issue with the IK is that the original length of an arm might become distorted to compensate for a particular hand movement. With FK this isn’t a problem as form is maintained at all times.
These pros and cons are not to say that IK is useless, but more that it has to be applied to the right aspects of a rig. For example often FK will be used for arms but IK will be used for legs as the IK animating is much more suited to quick movements of legs.
In addition to these basic areas which can be manipulated, there are levels of detailed movement which can be captured using the Maya software. For example: fingers, foot movement, and facial expressions can all be edited by imputing particular numerical values or by physical moulding the control. This just adds an entirely new level to what can be done with a basic character rig at the lowest level of Maya Control.
Having had around 3 hours on the software to just experiment, I constructed a simple animation, which I will upload later this week. Still, despite this new interest in the medium of 3D animation, it is somewhat restricting to just work with the ‘Moom’ so I will likely still work in 2D for my final piece.
Subscribe to:
Posts (Atom)