One technique I mentioned was using Kdenlive to simulate gate weave – that strangely pleasing effect whereby picture moves almost imperceptibly around the screen as you watch. If you’re not familiar with what gate weave looks like, here’s an example:
I mentioned in my previous article that I discovered I could simulate gate weave manually using the Kdenlive “Pan and Zoom” filter.
I did this by zooming in on my video slightly…
Video resized by 108% to zoom in on it slightly
…and then moving the picture around randomly at 5 frame intervals.
Keyframes added a five frame intervals; X and Y changed randomly
Once this is done, I could save this portion gate weave as a custom effect so I could re-use it:
Save effect button in Kdenlive
When you do this, your zoom, keyframes and random movements are stored in the ~/.kde/share/apps/kdenlive/effects folder as an XML file. The XML file created by Kdenlive for some manually created gate weave looks like this:
<effect tag="affine" type="custom" id="test"> <name>test</name> <description>Adjust size and position of clip</description> <author>Charles Yates</author> <parameter opacity="false" default="0%,0%:100%x100%" type="geometry" value="-29,-23:778x622:100; 25=-30,-25:778x622:100; 50=-29,-22:778x622:100; 75=-31,-24:778x622:100; 100=-28,-24:778x622:100; 123=-31,-21:778x622:100" name="transition.geometry"> <name>Rectangle</name> </parameter> </effect>
Obviously, creating gate weave by hand for a long piece of video using the Kdenlive interface would take hours. Luckily, because the resulting gate weave custom effect is stored as a simple XML file, you can write a quick script to create the gate weave instead.
So I thought I’d use this post show you the script I use to create “automatic” gate weave.
When I do scripting jobs I prefer to use Python if possible. For this task I needed to get it to write out an XML file. Python comes with a selection of complex but hugely flexible ways to do this. However, to make this as quick and easy as possible, I used a lovely Python module called pyfo which was developed by Luke Arno and did everything I needed.
Although you can install pyfo if you want to, there’s no need; you can just extract the pyfo.py file and put it in the same folder as your Python scripts that use it.
As you can see below, my Python script is pretty self-explanatory. The script below is suitable for adding gate weave to PAL 4:3 video.
The variable value_string determines how much the video is zoomed in by initially. For wide-screen PAL 16:9 video I adjust this value to “-29,-23:1034×622:100;”. The value of the step_value variable determines how often the video frame moves. I think every 5 frames often works best.
You can run the script multiple times to create different xml files, but if you do remember that you will need to change the value of the custom_effect_name variable each time to something different so you’ll be able to tell your gate weave custom effects apart.
#!/usr/bin/env python2.4
""" This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/."""
import random from pyfo import pyfo
custom_effect_name = "weave"
#Number of seconds of gate weave required frames_required = 750
#Initial value of frame value_string = "-29,-23:778x622:100;"
for i in range(0, frames_required, step_value): x = random.randrange(origin['x'] - max_distance['x'], origin['x'] + max_distance['x'] + 1) y = random.randrange(origin['y'] - max_distance['y'], origin['y'] + max_distance['y'] + 1) value_string += str(i) + "=" + str(x) + "," + str(y) + ":778x622:100;"
xml_output = ('effect', [ ('name', custom_effect_name), ('description', 'Adjust size and position of clip'), ('author', 'Charles Yates'), ('parameter', ('name', 'Rectangle'), {'opacity':'false', 'default':'0%,0%:100%x100%', 'type':'geometry', 'value':value_string, 'name':'transition.geometry'}), ], {'id':custom_effect_name, 'type':'custom', 'tag':'affine'})
result = pyfo(xml_output, pretty=True, prolog=False, encoding='ascii') print result.encode('ascii', 'xmlcharrefreplace')
As it is, my Python script writes the XML file it produces to the console window. You can copy and paste the resulting XML into a blank text file and save it to ~/.kde/share/apps/kdenlive/effects folder,
However, you can pipe the output into straight into an XML file instead. For instance:
$ python gateweave.py > gateweave.xml
Obviously, my Gate Weave solution isn’t very elegant, but who cares – it works, and it’s all free software!
If you’ve read my blog before, you may have come across some posts about my friend Roddy Buxton. Roddy is an incredibly inventive chap – he’s rather like Wallace and Grommit rolled into one! He has his own blog these days and I find everything on it fascinating.
One of Roddy’s cracking contraptions
One of the subjects recently covered on Roddy’s blog is the home-made telecine machine he built. The telecine was a device invented by John Logie-Baird at the very dawn of broadcasting (he began work on telecine back in the 1920s) for transferring pictures from film to television.
Roddy also shares my love of everything ATV, so naturally one of the first films Roddy used to demonstrate his telecine was a 16mm film copy of the ATV Today title sequence from 1976.
This title sequence was used from 1976-1979 and proved so iconic (no doubt helped immeasurably by the rather forgetful young lady who forgot to put her dress on) it is often used to herald items about ATV on ITV Central News. Sadly, as you can see below, the sequence was not created in widescreen so it usually looks pretty odd when it’s shown these days.
How the sequence looks when broadcast these days.
The quality of Roddy’s transfer was so good I thought it really lent itself to creating a genuine widescreen version. In addition, this would provide me with a perfect opportunity to learn some more about animating using the free software animation tool Synfig Studio.
The first thing to do when attempting an animation like this is to watch the source video frame by frame and jot down a list of key-frames – the frames where something starts or stops happening. I use a piece of free software called Avidemux to play video frame by frame. Avidemux is like a Swiss Army knife for video and I find it handy for all sorts of things.
Video in Avidemux
I write key-frame lists in text file that I keep with all the other files for a project. I used to jot the key frames down on a pad, but I’ve found using a text file has two important advantages: it’s neater and I can always find it! Here is my key frame list in Gedit, which is my favourite text editor:
Key-frame list in Gedit
After I have my key-frame list I then do any experimenting I need to do if there are any parts of the sequence I’m not sure how to achieve. It’s always good to do this before you start a lot of work on graphics or animation so that you don’t waste a lot of time creating things you can’t eventually use.
The ATV Today title sequence is mostly straightforward, as it uses techniques I’ve already used in the Spotlight South-West titles I created last year. However one thing I was not too sure about was how to key video onto the finished sequence.
Usually, when I have to create video keyed onto animation I cheat. Instead of keying, I make “cut-outs” (transparent areas) in my animation. I then export my animation as a PNG32 image sequence and play any video I need underneath it. This gives a perfect, fringeless key and was the technique I used for my News At One title sequence.
However, with this title sequence things were a bit trickier – I needed two key colours, as the titles often contained two completely different video sequences keyed onto it at the same time.
Two sequences keyed at once
Therefore I had to use chromakeying in Kdenlive using the “Blue Screen” filter, something I had never had a lot of success with before.
The first part was simple – I couldn’t key two different video sequences onto two different coloured keys at once in Kdenlive. Therefore I had to key the first colour, export the video losslessly (so I would get no compression artefacts), then key the second colour.
The harder part was making the key look smooth. Digital keying is an all or nothing affair, so what you key tends to have horrible pixellated edges.
Very nasty pixel stepping on the keyed video
The solution to this problem was obvious, so it took me quite a while to hit upon it! The ATV Today title sequence is standard definition PAL Widescreen. However, if I export my animation at 1080p HD and do my keys at HD they will have much nicer rounded edges as the pixels are “smaller”. I can then downscale my video to standard definition when I’ve done my keying and get the rounded effect I was after.
Smooth keying, without pixel stepping
The other thing I found is that keying in Kdenlive is very, very sensitive. I had to do lots of test renders on short sections as there was only one “Variance” setting (on a scale between 1 and 100) that was exactly right for each colour.
So now I was convinced I could actually produce the sequence, it was time to start drawing. I created all of my images for the sequence in Inkscape, which is a free software vector graphic tool based around the SVG standard.
However, in order to produce images in Inkscape I needed to take source images from the original video to trace over. I used Avidemux to do this. The slit masks that the film sequences are keyed on to are about four screens wide, so once I had exported all the images I was interested in I needed to stitch them together in the free software image editor The GIMP. Here is an example, picked totally at random:
She’ll catch her death of cold…
Back in Inkscape I realised that the sequence was based around twenty stripes, so the first thing I did before I created all the slit mask images was created guides for each stripe:
These guides saved me a lot of time
The stripes were simply rounded rectangles that I drew in Inkscape. It didn’t take long to trace all of the slit masks for the title sequence. Two of the masks were repeated, which meant that I didn’t have as many graphics to create as I was fearing.
Once the slit masks were out of the way I could create the smaller items such as the logo:
ATV Today logo created in Inkscape
And, with that, all the Inkscape drawing was done. It was time to animate my drawings now, so I needed to export my Inkscape drawings into Synfig Studio. To do this I was able to use nikitakit’s fantastic new Synfig Studio SIF file Exporter plug-in for Inkscape. This does a fabulous job of enabling Inkscape artwork to be used in Synfig Studio, and it will soon be included as standard in Inkscape releases.
When I did my Spotlight title sequence I exported (saved) all of my encapsulated canvases (akin to Symbols in Flash) that I needed to reuse within my main Synfig file. This was probably because I came to Synfig from Macromedia Flash and was used to the idea of having a large file containing all the library symbols it used internally.
I have been playing with Synfig Studio a lot more since then, and I realised a far more sensible way to work was to have each of what would have been my library symbols in Flash saved as separate Synfig files. Therefore I created eight separate Synfig Studio files for each part of the sequence and created a master file that imports them all and is used to render out the finished sequence.
The project structure
This meant that my finished sequence was made up of nine very simple Synfig animation files instead of one large and complicated one.
The animation itself mainly consisted of simply animating my Inkscape slit masks across the stage using linear interpolation (i.e. a regular speed of movement).
I could type my key-frames from my key-frame text file directly into the Synfig Studio key-frame list:
Key-frames for one part of the animation
The glow was added to the ATV Today logo using a “Fast Gaussian Blur”, and the colour was changed using the “Colour Correct” layer effect – exactly the same techniques I used in the Spotlight South-West titles.
ATV Today logo in Synfig
In order to improve the rendering speed I made sure I changed the “Amount” (visibility) of anything that was not on the stage at the present time to 0 so the renderer wouldn’t bother trying to render. You do this using Constant interpolation so that the value is either 0 or 1.
I had a couple of very minor problems with Synfig when I was working on this animation. One thing that confused me sometimes was the misalignment of key-frame symbol between the Properties panel and the Timeline.
This misalignment can be very confusing
As you can see above, the misalignment gets greater the further down the “Properties Panel” something appears. This makes it quite hard at times to work out what is being animated.
Some very odd Length values indeed!
Another problem I had was that the key-frame panel shows strange values in the time of length columns – particularly if you forget to set your project to 25 frames per second at the outset.
However, overall I think Synfig Studio did brilliantly, and I would chose it over Flash if I had to create this sequence again and could choose any program to create it in.
The most important technical benefit of Synfig Studio for this job was the fact that it uses floating point precision for colour, so the glows on the ATV Today logo look far better than they would have done in Flash as the colour values would not be prematurely rounded before the final render.
I rendered out my Synfig Studio animation as video via ffmpeg using the HuffyUV lossless codec, and then I was ready to move onto Kdenlive and do the keying.
Obviously I needed some “film sequences” to key into the titles, but I only have a small selection of videos as I don’t have a video camera. To capture video I use my Canon Ixus 65, which records MJPEG video at 640 x 480 resolution at 30fps.
My 16mm film camera
Bizarrely, when the progressive nature of its output is coupled with the fact it produces quite noisy pictures, I’ve found this makes it a perfect digital substitute for 16mm film camera!
I “filmised” all the keyed inserts, so that when they appear in the sequence they will have been filmised twice. Hopefully, this means I’ll get something like the degradation in quality you get when a film is then transferred to another film using an optical printer.
Once the keying was done the finished sequence was filmised entirely using Kdenlive using techniques I’ve already discussed here.
And so, here’s the finished sequence:
Although I’m not happy about the selection of clips I’ve used, I’m delighted with the actual animation itself. I’m also very pleased that I’ve completed another project entirely using free software. However, I think the final word should go to Roddy:
Thanks for the link. I had a bit of a lump in my throat, seeing those titles scrolling across, hearing the music, while munching on my Chicken and Chips Tea… blimey, I was expecting Crossroads to come on just after!
If you are interested in ATV, then why not buy yourself a copy of the documentary From ATV Land in Colour? Three years in the making, over four hours in duration, its contains extensive footage (some not seen for nearly fifty years) and over eleven hours of specially shot interviews edited into two DVDs.
Going through my old Flash files, I stumbled across an early version of this image, which I first produced in 2005. I didn’t know how to make it look realistic then, but I’ve since been given lots of good advice from Rory Clark. This new version was produced in Inkscape and aged in The GIMP.
In case you’re wondering, these were all real IBA Transmitters.
I often spend idle half hours looking around flickr for anything of interest. The other day I found a very nice Anglia logo from 1959. Obviously, I couldn’t resist recreating it in Inkscape while I was listening to a pod-cast:
Click to enlarge
This stylised Anglia pennant logo formed the basis of Anglia Television’s original end-caps, including the one seen on their opening program.
Three years ago I talked about the programs I used to simulate old 16mm film. Back in 2008 I was using Windows XP, Adobe Premiere Elements 4.0 and a VirtualDub filter called MSU Old Cinema. I found I could use them to create some half-decent 16mm film:
These days I’m using Fedora 15 as my operating system and Kdenlive as my off-line video editor. That meant I’ve had to change the way I simulate old film quite a bit. I have been continuing to use VirtualDub and the MSU Old Cinema plug-in via WINE. Although VirtualDub is free software, the MSU Old Cinema plug-in is not, and this bothered me. I wondered what I could achieve in Kdenlive alone and I started experimenting.
In the course of this blog post I’m going to use the same image – an ITV Schools light-spots caption from the 70s that I recreated in Inkscape. Here’s the original image exported directly from Inkscape as PNG:
Created in Inkscape
The most obvious sign that you are watching something on a bit of old film are the little flecks of dirt that momentarily appear. If the dirt is on the film itself it will appear black. If it was on the negative when the film is printed it will appear white.
Kdenlive comes with a Dust filter that tries to simulate this effect. However, it has a very small database of relatively large pieces of dirt. In total there were just six pieces of dirt, drawn as SVG files, and that limited number led to an unconvincing effect. If I used the filter on a long piece of video I found I began to recognise each piece! There were also no small bits of dirt.
I drew 44 extra pieces of dirt in Inkscape and added them to the Dust filter. I also redrew dust2.svg from the default set. I call this particular piece of dirt “the space invader” as I found it was too large and too distracting!
The video below compares the Dust filter (with identical settings) before and after I added my extra files:
You may find you prefer the Kdenlive dust filter with just the default six SVG files. However, if you prefer what I have done you can download my extra SVG files from here.
With the modifications I’ve made, I actually prefer the dirt created by the Dust filter in Kdenlive to the dirt you get in the MSU Old Cinema plug in. The dirt from Kdenlive’s filter is less regular in shape and simply by changing the SVG files in the /usr/share/mlt/oldfilm folder I can tailor the dust to any specific application I have in mind.
After flecks of dirt, the second most obvious effect that you are watching old film is a non-uniform shutter causing the picture to appear to flicker very slightly. The MSU Old Cinema plug-in can simulate this effect, but wildly over does it. It is not suitable for anything other than simulating silent movies, so I never used it.
Luckily the Kdenlive Old Film plug-in does a much more convincing job. The settings that I found worked for me are shown below:
KdenLive Old Film settings for uneven shutter
And they create the results shown below:
It looks a bit odd on it’s own, but when added to all the other effects I’m describing here it will look fine.
I’ve noticed that when I am creating these effects it’s best if I move away from the monitor to a normal TV viewing distance to see how they look – otherwise I tend to make the effects too subtle to be noticed when I come to watch the results on my television!
The next thing that will help to sell the output as film is having some film grain. Film grain is irregular in shape and coloured. In fact, I used the Colour Spots setting of the MSU Noise filter to create film grain in VirtualDub.
Kdenlive has a Grain filter, which simply creates random noise of 1 pixel by 1 pixel in size. Although technically this is not at all accurate, it can look pretty good if you are careful. The settings for film grain will vary from job to job, so some trial and error is involved.
As a starting point, these settings are good:
Kdenlive Grain settings
And will look like this:
Again, it looks odd by itself (and you can’t really see it at all on lossy YouTube videos!) but it will look fine when added to the other effects. You’ll start to notice the rendering begin to slow down a bit when you have added Grain! Incidentally, Grain is still worth adding even if YouTube is your target medium because it helps break up any vignette effect you add later.
The next thing you need to do is to add some blur – edges on 16mm film in particular tend to be quite soft. Kdenlive has a Box Blur filter which works just fine for blurring. How much blur you add depends on your source material, but a 1 pixel blur is fine as a starting point.
Colour film is printed with coloured dyes, so it has a different colour gamut to the RGB images you create with The GIMP, Inkscape or a digital video camera. In addition, it also fades over time. Therefore to make computer-originated images look like film-originated images some colour adjustment is normally required.
Luckily, Kdenlive has a Technicolor filter that allows you to adjust the colours to better resemble film.
Kdenlive Technicolor settings
The way colour film fades depends on whether it has been kept in a dark or light place. If I’m recreating a colour 16mm film that has been stored safely in a dark tin for many years, I make it look yellowish. If I’m recreating a colour 16mm film that’s been left out in the light a bit too much I make it look blueish. Both these looks rely adjusting the Red/Green axis slider – not the Blue/Yellow axis slider as you might think!
Source image faded with Technicolor
You soon begin to notice that the telecine machines used by broadcasters could adjust the colours they output to make colours that were impossible to resolve from the film. For instance, some of the blue backgrounds on ATV colour zooms were too rich to have been achieved without some help from the settings on the telecine machine. So the precise colour effect you want to achieve varies from project to project, and sometimes you will be actually increasing colour saturation rather than decreasing.
The Technicolor filter is, ironically, the filter you use to make colour source material monochrome too!
The biggest problem when trying to recreate old film is recreating gate weave – that strangely pleasing effect whereby picture moves almost imperceptibly around the screen as you watch.
MSU Old Cinema created an accurate but very strong gate weave which was too severe for recreating 16mm film. The Kdenlive Old Film filter has what it calls a Y-Delta setting, that makes the picture jump up and down by a set number of pixels on a set number of frames. It’s easy and quick (a Y-Delta of 1 pixel on 40% of frames is good) but introduces black lines at the top of the frame and is so obviously fake it won’t really fool anyone!
So there is, sadly, no quick way to create gate weave in Kdenlive. However, the good news is there is a way, provided you’re prepared to do a bit of work. You need to use the Pan and Zoom filter. The Pan and Zoom filter is intended to do Ken Morse rostrum camera type effects – it’s particularly good if you have a large image and want to create a video to pan around it.
However, what we can do is use the Pan and Zoom filter to move the frame around once per second. Firstly you zoom the image in by 108%. This means you won’t see any black areas around the edge of the frame as the picture moves around.
First of all, zoom the image very slightly
Next, you create key frames on each second:
Then add one key frame per second
Then you move the image around slightly on each keyframe – plus or minus two or three pixels from the starting position is often plenty.
Obviously, for a 30 second caption that’s 30 keyframes and 30 movements – a lot of work if done “by hand”. However it won’t go to waste, as you can save your Pan and Zoom settings as a Custom effect and resuse it again and again on different clips.
And, luckily, doing all this by hand isn’t even necessary. Custom effects are stored as simple XML files in the /kde/shared apps/kdenlive effects folder so it is possible to write a small Python script to automatically create as much gate weave as you want – something I’ll come back to.
As well as gate weave, you can also use the Pan and Zoom filter to stretch the frame, which is perfect for simulating stretched film. Again, that’s hopefully something I’ll return to another time.
Here’s an example of video moving with the Pan and Zoom filter:
The Pan and Zoom filter also adds hugely to your rendering time, so it’s best to switch it off until you do your final render.
Glow is a very important effect to add when simulating film, particularly monochrome film. Kdenlive does not have a glow filter, so if I need to add glow to a video file I have to improvise. I export the video as a PNG sequence, add glow to the PNG files using a GIMP batch script (written in Scheme), and then reimport the video file. It’s worth the effort, as it’s amazing how much glow helps to sell something as being originated on film.
Glow added using The GIMP
The GIMP glow filter tends to be rather harsh, and tends to wash out images if you use too much glow. Therefore you have to experiment a lot.
Finally, there is often uneven brightness or contrast visible across a film frame. In VirtualDub I used a filter called the Hotspot filter. The hotspot filter is actually designed to remove this effect from old film, but turned out to be just as good at putting the effect in!
However, with Kdenlive, this effect is best achieved in the GIMP when required as Kdenlive’s Vignette effect is too unsubtle to be of any real use.
So, put it all together, and you get something like this:
All in all, Kdenlive does a pretty good job of making digitally originated images look like 16mm film but although there is room for improvement. The film scratches filter needs work, there is no glow and the film grain is really just noise rather than grain. However you can still get some excellent results and I’m really pleased with it.
If you’ve been reading my blog for a while, you’ll know that whenever my attention turns to the Midlands it’s usually at the prompting of my friend Roddy Buxton.
Roddy is a lighting engineer, electrician and visual effects designer. He is now based in South Yorkshire but grew up in ATV Land. Roddy’s TV career started in the Central Television film department, working as a spark on such programmes as Peak Practice and Boon.
As with Oliver Postgate, Roddy’s branching out into visual effects design happened quite by chance when a director noticed how practical he was and decided he’d be the perfect person to knock up a semi-practical suitcase nuke for a film!
In 2009 Roddy thought he’d like to have a go at creating a working replica of the ATV station clock from the sixties. As far as I know, this clock only exists as an off-air photograph in the Transdiffusion archive.
Station clocks used to be shown at numerous times during the day by all televisions stations from the 1950s to the early 1990s. Giving the correct time was not only seen as a valuable public service to the viewer, but the clock equipment was used to sync the studio’s signals with external sources. This was vital to prevent visual glitches such as this:
Roddy asked if I could supply him with the artwork in a format suitable for printing, something I was only too happy to do. I sent him two Inkscape files, one for the clock face, the other for the hands:
Watch your face…
…and hands.
Roddy soon had other things on his mind – a new addition to his family! – and I thought no more of the model clock.
However, earlier this year Roddy started asking me questions about my Flash recreation of the BBC Schools dots. The BBC Schools dots were shown in the minute immediately preceding BBC One’s programmes for schools and colleges between September 1977 and June 1983.
In the final year the dots were digitally originated using technology similar to Richard Russell’sGNAT clock, but before that the dots were a mechanical model in the “Noddy room”.
The Noddy room was a special studio in the BBC that held various mechanical models and 12″ by 10″ captions. These were captured in black and white by a remote controlled camera that used to “nod” up and down as different ones were selected – hence the name “Noddy” room. Colour was usually added electronically to the images before they were broadcast.
Roddy discussed the lighting for the dots:
The lighting for the BBC Noddy wasn’t anything specialised. It consisted of two P38 flood lamps (available from all good DIY/Electrical stores) – these are likely to have been photographic lamps – however the only difference being is the price and box they come in. They are the same voltage, wattage and colour temp. The lamps were attached to the camera, so wherever the camera was pointing that area would be lit.
So I supplied Roddy with the dots artwork as an Inkscape SVG file and I also uploaded my latest recreation of the dots in Flash to YouTube:
In May I was delighted to receive a mail from Roddy with a photograph of this prototype dots model:
Prototype dots, courtesy Roddy Buxton
Roddy said:
The clock face is made from hardboard; though I am not happy with the results; as the dots are not that hard edged. I think I will end up using this clock face as a template to make the actual clock out of punched steel/aluminium – that way I can get hard edged dots.
Here are some more of Roddy’s pictures of making the prototype:
Holes drilled through the hardboard
Temporary captions applied to check sizing
Matt paint added to remove the reflections
I thought the prototype looked absolutely fabulous, and by now I was looking forward to seeing the finished product enormously. I didn’t have long to wait – on the 11th June Roddy wrote to say:
The artwork for the BBC “Dots” has arrived in printed form. The company I have used did all of my printing for me for £11 including P&P, and to the correct sizes – and in a matt finish too.
And on the 6th of July I finally got to see how the final schools dots model was progressing:
I showed the progress so far to my friend Rory Clark who summed it all up beautifully when he said:
Bloody hell – they’re impressive!
Not only that, but Roddy had started work on something else as well!
In Colour – the ATV Station Clock replica
Roddy assures me:
Another week and the dots will be vanishing ;-)! (Finally)
I’m very much looking forward to that. And I’m also hoping Roddy will also recreate some other models from times past – and it looks like I’m in luck!
Roddy tells me:
On the “to do list” – I have always wanted a BBC Globe – so will look at that. The “Diamond” is pretty easy to do (will definitely look at that!)
The only thing I haven’t been able to do using free software since moving to GNU/Linux in 2008 is animate. And it bugged me. Everything else – raster graphics, vector graphics, offline video editing, audio editing, font design, desktop publishing – I could achieve, but animation was the reason I’ve had WINE and Macromedia Flash 8 installed on my machine for the past three years.
When I first started playing with GNU/Linux I came across a program called Synfig Studio which could do animation, but at that time it needed to be compiled from source code. It seemed a bit too much like brain surgery for a GNU/Linux beginner! However, the other day I was banging my head trying to do some animation in Flash. I decided to Google for any free software tools that might be able to help and I was reminded of Synfig Studio once again.
Blue hair? Why, it’s Mrs. Slocombe!
I went to the Synfig Studio website and the first thing I noticed was that a brand new shiny version of Synfig Studio was available as an RPM for Fedora. In other words, all I had to do was download, double click and go. Everything worked perfectly. I found the Synfig Studio website was excellent, there were a large number of tutorials and an extensive manual and so I set about reading.
Animation programs are always off-putting to beginners due to their complexity, and Synfig Studio was no exception – partly because it began life as an in-house tool in a professional animation company and that really shows in the power and complexity of what it offers.
I learned Flash 2 back in 1998 by trying to create the ATV Colour Zoom ident as I thought it would be quite a good challenge and force me to look into the tool properly. For the same reason I dusted off one of the more challenging animations in my “TODO” list to learn Synfig – the BBC South West Spotlight dots titles.
My plan was to draw the Spotlight logo in Inkscape, import that into Synfig Studio and then animate it. The first thing I did was set up my canvas. Changing the units to pixels is very important – Synfig Studio uses points by default which seems a strange choice for a tool not centred on printed work.
When I tried importing my artwork from Inkscape it came in at the wrong size:
Imported SVG from Inkscape
The reason was obscure and not what I had been expecting. I had assumed it was the old Inkscape dpi (dots per inch) problem, but it was to do with something called Image Span which is related to the aspect ratio of the end animation. After reacquainting myself with Pythagorean theorem I worked out I needed to set the Image Span to 16 for 768 by 576 pixel artwork from Inkscape.
Setting Image Span in Synfig Studio
Then artwork comes in correctly from Inkscape. However, now I could see some problems with imported SVG:
Problems with Imported Inkscape SVG
There were two problems – the holes had disappeared in the “P” and “O” and there was a segment missing from the circle of the letter “O”.
Paths with holes are imported into Synfig Studio as two objects or “layers” (everything in Synfig Studio is a layer) – the letter and its hole. To make a letter with a hole in it you need to place the hole layer above the letter layer, and then give the hole a layer an “alpha over” blend method. As you can see, the logic behind the program is very different to Flash!
Using Alpha Over in Synfig
The nick out of the letter “O” was Inkscape’s fault. When you convert text to paths in Inkscape you often get double nodes (nodes stacked on top of each other). Double nodes also cause problems in Inkscape itself so it’s always a good idea to merge these nodes in Inkscape.
The join nodes button in Inkscape
Inkscape ellipses don’t import as Synfig Studio circles (they come in as something called Blines instead), so I redrew the dots in the Spotlight logo as Synfig Studio circles to make animation easier later. In fact to get an ellipse in Synfig Studio you draw a circle and then apply a transformation layer to it – again, a bit strange for a beginner! So, now I had the artwork imported:
Inkscape SVG imported perfectly
I discovered I didn’t actually need the background rectangle I’d drawn in Inkscape in Synfig Studio, there’s a special type of layer for solid backgrounds called “Solid Colour” that always fills the background however large your animation is. This is analogous to “Background Colour” in Flash, only in Synfig Studio you could use a “Gradient” instead.
Now I needed to colour my artwork. I found a small bug in Synfig Studio which means that you cannot use the HTML-style RGB value (a six digit hexadecimal number) to enter colours. My background colour in hexadecimal was #171a17. When I entered this into Synfig Studio I got a mid grey, instead of the charcoal colour I was expecting.
A Lighter Shade of Dark
I went into the GIMP and discovered that #171a17 is equivalent to the the RGB percentages 9% 10% 9%.
The GIMP Colour Picker information dialog
I entered the values 9%, 10%, 9% into the Red, Green and Blue spinboxes on the Synfig Colours dialog box, and I got the colour I expected. However, I also found that the HTML code displayed on the Colours dialog became 010101 – not what I expected!
In Synfig Studio, the HTML code is wrong
The ever-helpful Genete on the Synfig Studio Forums suggested that I might have a non-linear palette selected for my file, but this turned out not to be the case. So the moral of the story is, sadly, only enter colour values as RGB percentages.
Speaking of colours, it would be great if Synfig Studio could load GIMP palettes, or create a palette from the currently imported layers.
I then set about animating. This is quite different to Macromedia Flash as in addition to “keyframes” you also have the concept of “waypoints”. A “keyframe” stores every setting of every “layer” item on the current canvas at a particular point, whereas a “waypoint” just stores one setting. You also have to forget about the concept of “frames” that was so key to Macromedia Flash. Synfig Studio, in common with Swift 3D, uses the concept of time instead. As far as the time-line was concerned I am very glad that I had done some work in Swift 3D before approaching Synfig Studio.
Keyframe labels appear on the canvas too
One thing I did like is the fact you could label not only your layers but your keyframes – that saved me an awful lot of scribbling! Once you have your keyframes set up Synfig Studio really excels. There are numerous different ways of defining how the animation gets from one keyframe to another. The default was TCB which gives beautiful naturalistic movement, but for Spotlight it would cause arcing like this:
Arc caused by TCB Interpolation
When I really wanted linear tweening to give me straight edges like this:
Corrected by Linear Interpolation
Another little gotcha I found whilst animating was that the time-lines starts at “0f”, not “Frame 1” as in Flash. This caught me out when I was putting the animation together as I was getting odd blank frames!
Whilst animating I came across a niggle caused by my operating System. In GNU/Linux Alt and left-click is used to move windows around. However, in Synfig Studio Alt and left-click is used to transform (i.e. scale) objects. Fedora 15’s deskptop GNOME 3 compounds this problem by removing the “Windows Movement Key” setting that you could adjust in Gnome 2 to change this behaviour. Fortunately the wonderful Synfig Studio forum came to the rescue as “nikolardo” had a cunning work-around:
“Another workaround for the Alt issue presents itself when you realize it only happens when you Alt-click. Pressing Alt and then clicking gets picked up by the WM (openbox, in my case), but clicking on a vertex and then holding the Alt key produces the scaling behavior intended. So, next time you Alt-click and the window moves, let go, and then click-Alt.”
Whilst working I found that “Groups” were not what I expected at all. The purpose of Groups in Synfig Studio is to collect disparate items around your animation so they can be selected together. In fact, when creating the animation I never used any groups at all, although I can see how they would be useful on other animations.
I loved the fact I could enter a frame number e.g. 454 to move somewhere on the time-line and it got converted into seconds and frames. I tend to think in frame numbers and it’s great I don’t have to keep dividing by 25 and working out the remainder. This was a huge help when setting up keyframes.
Useful for creating guides at 0x and 0y
Another thing I found was I could use the Canvas Metadata window, which at first seemed useless, to adjust the guides. It would be even better if you could use pixels instead of internal units to adjust the guide positions in this window.
One thing I soon learned as I worked was that Synfig Studio’s canvas window is not always WYSIWYG, and the Preview Window isn’t always an accurate reflection of the end result either (but this is being rewritten for the next release) – you have to do a render in order to see how your final result is coming along. This is particularly true if you are using effects like Motion Blur. For instance, when the Spotlight S is rotating, this is what I get to see on the stage:
What you see in Synfig Studio…
Whereas this is what the end result looks like:
…is much more impressive when rendered!
Correction from Genete:
“That’s because your display quality settings were not set to high quality. There is a spin button on the top of the canvas that allows you to set the quality to 1 (better), instead of use 8 (worse) the default value. WYSIWYG is fully done always in Synfig Studio. The problem is that it takes some time to render complex effects like motion blur, duplicate layer, etc.”
For my renders I used a PNG sequence, and only rendered the frames I’d just worked on. One thing I noted when rendering is that the render progress bar and cancel button on the canvas window don’t work. In the future I would love it if a WebM render option was added to Synfig Studio, particularly given the popularity of YouTube.
Notice that zooms, blurs and colour corrections are layers.
As I’ve said before, in Synfig Studio everything is a layer. Not just every single shape but a whole host of other things such as colour changes, blur effects, tranforms. So, obviously the number of layers you get soon gets large and unwieldy. However you can “encapsulate” layers together into what are called “Paste Layers” and then deal with these encapsulated layers as one object.
The capsules show encapsulated layers
You may be thinking this sounds a bit like the Flash concept of having symbols, but it isn’t – yet. The encapsulated layers are still on the main canvas and therefore use the main canvas’s time-line. In order to use encapsulated symbols in a way analogous to Flash library symbols you need to “Export” the Paste Layer as a separate Canvas. It will then appear in the Canvas Browser.
The Canvas Browser
Now your capsule of layers is a canvas in its own right, with its own independent time-line and you can use it in a way akin to library symbols in Flash. As you work, you’ll find that the main canvas’s time-line gets cluttered with keyframes and waypoints, so it’s worth exporting encapsulated layers to simplify your work.
The only real downside of the Synfig Studio time-line design is shared by Swift 3D. It’s that you can’t add and remove things from your animation easily. If you want to “hide” something you have to set its amount to 0 and then you have to fiddle about with waypoints with constant interpolation in order to show it again. It seems too much work when you simply want to put things on and take things off of your canvas.
Exporting a Paste Layer after you have already done work on an animation needs some care. Key frames are not brought across to the new canvas, and the exported animation duration defaults to 5s (five seconds) which means you have to increase it to the right length manually. So, before you start work on an animation it’s better to decide upon its structure first. But that was always the case anyway!
One minor thing – I found that I could only remove things out of encapsulated layers by dragging and dropping which was not discoverable for me – I expected to find another way of doing it via a button of some kind too.
Put a space in an Exported Canvas name and…
Entering a canvas name with a space in gives a message telling you about the C++ standard type library throwing an exception – not something most cartoonists would find particularly helpful!
When adding an exported canvas from the canvas browser on your main canvas you can offset its start-point by any number of frames. However, the offset needs to be a negative number of frames to make it appear a positive number of frames later and a positive number to make it start earlier which foxed me for a bit too!
Anyway, enough moaning – these are only very minor points! What you should take away from all this is that with exported canvases I found I could work exactly the same way as I was used to in Flash.
This does the hard work in the Spotlight animation.
Meanwhile, back to my animation. I wanted to emulate some optical film effects in my animation. The first one, motion trails, was easy to do with the Synfig Studio Motion Blur layer. This gives you a huge amount of control over the appearance of your finished trail.
Software doesn’t get any more magical.
I also needed some “optical glow”. I achieved this very easily by using the Colour Correct layer. This actually had a setting for Over Exposure – the exact effect I wanted to emulate – built into it! I was absolutely amazed! And not only that, I could animate the Over Exposure setting too. Incredible.
A bit of Blur (of which there are a dazzling array) helped to sell the glow even more.
The range of effects you can add to your animations in Synfig Studio is truly overwhelming. I think I’ll be blogging for months about the huge range of things you can do in Synfig Studio. It is an enormous amount of fun.
Zoom layers are a very clever idea.
To zoom in and out I used, naturally enough, the Zoom layer. Having a zoom on a separate layer is incredibly sensible when you actually start using it, but seemed very odd at first appearance.
And, it goes without saying, moving the dots around the canvas in Synfig Studio was simplicity itself.
So, here’s the finished result:
Did I mention Craig Rich knew my Granny…
Synfig files are very small and compact. The final file size was tiny – 11.9KB. I found that utterly incredible and it compares very favourably to Flash.
I could have completed these titles in about two hours in Macromedia Flash 8, in Synfig it took me two days to learn the tool and complete the animation which I was quite pleased with.
Synfig is an excellent tool that is staying firmly installed on my computer! I really love using it and I am excited about what I can achieve using it in the future and the vast range possibilities it opens up. It is powerful, flexible, stable and rewards the effort you put into learning it a thousand times over. It also has a friendly and helpful community. Recommended.
I’ve always found that I learn far more about a piece of software through playing rather than working and that’s certainly been the case with the offline video editing program Kdenlive. I’ve also found that when you have a definite goal in mind then “the software doesn’t do that” no longer becomes an option and you use the software far more inventively in order to achieve it.
And so it proved when I decided, using Flash 8 and Kdenlive, I’d try and recreate the title sequence for ITN’s News At One. I first saw this title sequence back when the programme was called First Report – that was before the soporific Aussie soap The Sullivans pushed the programme to its more familiar one o’clock slot.
Here are the original News At One titles that someone culled from TV Ark:
As you can see, it’s quite a complicated title sequence and people who know far more about this sort of thing than me have noted how hard it would have been to achieve in the 70s. They’ve always been very, very good technically at ITN.
Although I haven’t got a clue how ITN did it, I decided on the basic approach of having two layers of video in Kdenlive for my version.
The bottom layer would be the typing hands (or paws in my case – thanks Lionel!). These were shot on my little Canon Ixus camera and then put through the Kdenlive “Threshold” filter and coloured with, of all things, the MSU Old Cinema filter.
Lionel touch-typing
The top layer of video would consist of the zoom out of the Houses of Parliament’s clock tower (replaced by the block of flats where I live) which fades out, and the photo montage of odd people and drab places (replaced by pictures of my daughters, their toys and things in the square outside).
Creating the montage for the top layer was straight forward. I loved the fact that, on the original, the outlines around the images were all different widths, and some of them were even bent out of true. Even the top of the News at One logo was at an angle!
Creating the montage in Flash
The problem I faced was how to merge these two layers. My first thought was to use a Chroma key, but this gave me all sorts of problems and some rather nasty digital fringing that looked like anything but 70s television.
Click to enlarge
What I really needed was to be able to export transparency from Flash into Kdenlive – and it turned out that I could. The first step was to export the sequence out of Flash 8 as a PNG sequence at 24 bits per pixel with alpha channel, like this:
Getting transparency from Flash to Kdenlive
Then I could import the PNG sequence into Kdenlive, and use a Composition transition to knit the two layers together.
Click to enlarge
PNG sequences are the best thing ever invented for my work – everything can use them perfectly and they mean that I can transfer work between packages without losing quality or suffering odd effects.
So, here’s the finished sequence:
Now I understand keying properly I can do loads of things I thought I simply couldn’t do before. And all because I thought I’d do a little sequence to make my daughters laugh.
One of my favourite websites is Mikey Bennett’s Vintage Technology page. I love looking at all the vintage television sets Mikey has lovingly restored back to full working order. If your old Bush has lost its colour or your horizontal hold is ruining your enjoyment of your Rank then Mikey’s your man.
Mikey on an experimental 1969 3-D television
A few years back Rory Clark created a very entertaining DVD to demonstrate all the sets in the South West England Vintage Television Museum collection that Mikey curates. The DVD featured a range of test cards and tuning signals from the very old up to the present day accompanied by a selection of tones and music. Although it has given sterling service since then, Rory wanted to create an updated and expanded DVD for Mikey.
One of the cards Rory wanted to include this time was a prototype Test Card F featuring a rather different picture in place of Carol Hersee. Here’s the original:
String vests have never photographed better
Unfortunately the surviving scan of the card wouldn’t really show off Mikey’s television sets to best effect as it has faded quite considerably – the grey linearity squares had a distinctly reddish cast and the green castellations in the reference generator area had almost gone black. Therefore Rory asked me if I could recreate the card.
To do this, I used Inkscape as I only draw items in Flash now if it’s completely unavoidable. This is what I came up with in Inkscape:
She’s gone. Was it something I said?
The hardest job when recreating the card was doing the hand lettering on the caption. I did experiment to see if I could get away with using Benguiat Condensed, but it simply didn’t look close enough. In the end, the lettering took as long as the whole of the rest of the card.
It’s interesting to see the differences between this card and Test Card F. A good place to go to find out what’s missing is Alan Pemberton’s Pembers’ Ponderings website. He has two clickable Test Card Fs which will tell you exactly what each part of the card does.
I can’t wait to see how Rory “distresses” the my Inkscape drawing to make it look like a real transparency on the finished DVD.
I’ve kerned A–Z, a–z and 0–9 of the Washington typeface so that it’s useful for most English language applications. Version 2 of the typeface is available to download here. The version 2 package includes the FontForge source file, along with a PDF showing the available glyphs. Windows users will probably need to download the 7-zip utility in order to decompress the archive file.
