Quantel Paintbox History
quantelpaintbox.com
quantelpaintbox.com
For a few evenings in 1984 or maybe 85, Jerry Garcia rented the whole place out and played on the Paintbox.
We were also one of the first facilities to have a Quantel Mirage. We hired a math guru out of Berkeley, Eddie Elliot, to program the Mirage (Pascal). Jerry Garcia rented that, too. He had a bunch of analog graphics created on an Aurora system that he wanted to run through the Mirage.
His Paintbox stuff appeared to be just to play around with the system. He’d show up with a portfolio of sketches on paper and disappear into the Paintbox suite for an hour or so at a time. His manager, a hard-nosed guy, would sweep the place to make certain no remnants or artifacts remained in our hands after the sessions.
One night at about 3am a fist-fight between client guests broke out in the sunken client area in front of the edit suite console. A Jefferson Starship music video was being edited. Fun times.
This systems could go from Luxury Yacht to Anchor shockingly fast. I visited Broadway Video in NY (a very high-end post house) in the later 90s and they had a disconnected Quantel Mirage (~$750,000) literally propping a door open.
The earliest systems had such astronomical prices because they were based on huge amounts of discrete components and they knew the lifetime sales would likely only range from a few dozen to hundreds of units. The original Paintbox only had a 68000 CPU yet had to handle mixing two full frame composite video images with alpha channels at 60fps with 1 FPS UI lag. That required dedicated hardware to achieve most of its functions. I doubt the CPU ever touched the actual video data in most operations. The hardware processing pipelines in video effects systems of this era usually digitized, processed and manipulated video in its native YUV or YIQ encoding. Decoding to RGB would have been very costly, even more bandwidth and memory intensive - and rather pointless since general purpose CPUs couldn't work nearly fast enough to process it in real-time.
Reading patents from the high-end broadcast video effect systems of the 1980s is eye-opening. I was blown away by the 1981 Ampex ADO patent. https://patents.google.com/patent/US4472732/en
In the late 1970s these studs made discrete hardware to remap live video sources onto almost arbitrary quadrilaterals with alpha channel, light sourcing and shadows in real-time. This was basically implementing GPU texturing and lighting functions out of Atari 2600 era 74LS-type logic. The designs had to use decadent amounts of the fastest static RAM available to cache and buffer every line of video, and even then only barely kept up. It's no wonder these crazy racks of gear could cost upward of a million dollars. Yet by 1990 the march of Moore's Law would allow clever designers to achieve similar effects with the Video Toaster, a sub-$2000 add-in card for an Amiga. The Toaster won an Emmy award for technical achievement and unleashed the real-time desktop video revolution which put this kind of tech in everyone's hands.
There was a very recent post about someone restoring an Amiga 1000 that had me reminiscing about my days with an Amiga 2000 with a video toaster. I commented there about how that device replacing all of the existing gear was how I decided what career path I was taking. Maybe that shell shock of new replacing old at a young age allowed me to accept old equipment being no longer of any value where the older coworkers just had not had to accept that just yet.
So, basically using an esoteric multi-hundred thousand dollar piece of gear as a time base corrector. Cool. I recall abusing gear in similar ways back in the day!
when I was first starting, I would spend down time patching/routing the output of one thing into the input of another thing just to look at what the signal looked like before/after. that hacker ethos reveals itself in many forms
Hmmm, maybe the TBC was just re-timing the sync but the ADO was fully stripping and regenerating everything outside the active picture area. A lot of early TBCs only had 8 or 16 lines of memory, whereas the ADO had to have had a full frame store.
> I would spend down time patching/routing the output of one thing into the input of another
Brings to mind a story. My first exposure to any real gear was at a tiny video studio that only had minimal older gear to make industrial quality stuff on 3/4" tape. Late one night when no one was around, I disconnected the main production switcher and re-wired it into a different configuration using our bottom-of-the-line Chyron VP-1 as a full frame graphics generator. The VP-1 was only a very basic lower-third generator. It was "cheap" because it had no interface at all, just an RS-232 port and the user had to supply a separate computer to send it text strings with simple formatting codes for font, color and spacing. Basically, like a Hayes modem for titling (ATDT "Co-Starring"). Being the "computer guy", I got the assignment of writing our titling software using a Kaypro CP/M machine to send codes.
Since I'd written the software, it was easy to change it to generate strangely formatted stuff that still complied with the API. So that night I formatted a "lower-third" that started at the top of the screen and filled every line with maximum font-sized periods with zero kerning and zero line spacing. Because I chose a font where the period symbol was square, it filled the screen with a solid color. Modifying some of the lines to change certain period dots to have a foreground color of {transparent}, I was able to cut two side-by-side squares out of the center of the solid colored frame. Making the background color of the transparent period dots black, gave these inset squares soft-edged black drop shadows over the color field. Adding a one period-wide border to each square in an alternate foreground color was then a simple addition.
I wired the switcher to make this a final layer with two cut-out windows to whatever was underneath. I then fired up two studio cameras and framed up distinct shots in each. Using the switcher's effect buss I set up a vertical wipe splitting the screen between the two cameras. The overlaid Chyron graphic covered the split line and I carefully framed each camera to compose a nice shot. Finally, I added a text title centered underneath each inset square. And recorded the final output before unhooking everything and re-wiring back to normal operation.
The next day everyone was mystified when I showed them the recorded shot because I'd made something that looked exactly like a high-end network dual squeeze box digital effect with multiple busses of keyed graphics (in those days, shadows required a separate buss). All in a studio that had no ability to generate any graphics or any way to digitally reposition video. By rewiring components in odd ways and abusing various functionality it was possible to get some really interesting and unexpected results. Good times... :-)
One of the oddities or interesting parts about paintbox and later systems (including harrys and henrys, they were in use right up until ~2010) is that they didn't have a "framestore" that is, you painted one image, saved it to tape, loaded in a new one etc etc. There was no real "local" storage for keeping frames, as the article pointed out.
I had heard that there were disk based framestores, but I'd never seen one. There are a few people on mastodon that specialise in restoring Quantel kit like this.
Ofcourse David Hockney used a piantbox to produce some bollocks, as he always does, but that doesn't do justice to the epic artistry that paintbox operators had.
I think that was true for the early generations. The Paintbox only had memory for a foreground, background and alpha channel, so 2.5 frames at most. However, I recently watched a demonstration video for the Harriet which followed only a few years later and it said it could store 12.5 seconds of video to RAM - which would explain it's high price. The video showed recording video clips to this memory by simply playing them in real-time. It also mentioned being able to store up to 30 seconds of video offline on a removable Winchester disk pack (although I'm sure transferring video frames to that over SCSI required quite a while).
That said, there was definitely a long period of time where a lot of computers capable of displaying one frame at a time of video output where used for animation by pairing them with an edit controller and an industrial or professional-grade editing video recorder. The computer would be hooked up via RS-232 (or 422) to the edit controller which would control the VTR to rewind to a certain point, roll forward and at the precisely correct moment record a single-frame insert edit onto the tape. Then the controller would stop tape and wait for the computer to signal it had loaded or rendered another frame for recording and repeat the process all over again. Not counting loading or rendering time, just recording each frame to tape probably took 10 or 15 seconds, so a decent length clip could take hours just to record. This sort of animation recording was common on Amigas (often with a Video Toaster installed), Macs and other computers with decent graphical output. Quantel's Harriet and other systems capable of real-time video playback in this era were the ultra luxury version that avoided trashing magnetic tape around to record animation and affordable only by the extreme high-end. Although even the "low-end" was far from cheap. As I recall, just the edit controller cost around $5,000 and VTRs capable of doing single-frame edits started at over $10,000 (all in 1980s dollars).
Adrian Wilson
Honestly seems underpriced for what it was at the time. They should have looked at e.g. Cray Inc for pricing inspiration. They delivered something conceptually similar; a time machine into the future.
I was at a color facility when DaVinci was bought by Black Magic. The software+hardware to run the old version was well over six figures, and the dongle for the software was the hardware as the software was useless without it. Then BMD released the software only version...for free and it was not useless. Getting the licensed version was $1k (which now comes free with other hardware purchases from BMD). So even if you trick out a system with multiple GPUs, you're maybe pushing four figures. The new system is pretty much resolution agnostic as long as you have the patience for it while the old systems were very much resolution limited.
So yes, things are very much cheaper and more bang for that buck today. This is how the world works. For the people that spent all of that money, it was a very upsetting realization that the industry had become democratized where the cost of things was no longer a barrier to entry. They always cried over the money spent while ignoring how much money was made from it. It's just weird to see the olds crying over deprecated gear that still works but is absolutely useless in modern day.