Gephex is bril­liant. Prob­a­bly a great way to build a really capa­ble vision mixer (with some good real-time cap­ture hard­ware) on a shoe­string bud­get. I’m sick of drop­ping $120 and trekking over to Artar­mon every time a few sources need to be strung together! Actu­ally, if it weren’t for the fact that hire was locked in for an immi­nent evening, I’d prob­a­bly have can­celled and spent my $120 on another cap­ture card, instead. It’s nearly 10 frames behind real­time but that’s on a reg­u­lar Win­dows box run­ning as an un-prioritised process… on a ded­i­cated *nix machine I reckon that would drop back to about 4 frames, which is totally a fair deal (you nor­mally lose ~2 to genlocking/keyers any­way, and more if there’s a chain of mix­ers involved). Oh, yeah, and it does myr­iad effects and key­ing, too. Need to fig­ure out how to link net­work streams in, but its pretty much per­fect already. This is totally tak­ing prece­dence as my spare-time hard­ware project — it’s just call­ing for some proper gear to be built. Time to buy that book on micro­con­trollers methinks.

There are other hard­ware projects I’ve got cook­ing, yes, but none so imme­di­ately use­ful or eas­ily imple­mented. The great thing about this is the hard work (read:software) is essen­tially done already. At worst I’d need to hack some kind of inter­face dri­ver, but, really, it’s pretty much func­tional as is. And, because it’s already been ported to Linux and BSD, it’s really triv­ial to build a bare­bones sys­tem upon which to base it all. Pre­serv­ing key­board + mouse input is a totally nec­es­sary design para­me­ter any­way (for rea­sons of net­work stream inte­gra­tion, titling(!!), etc.) so hard­ware can be peri­od­i­cally switched on as it becomes avail­able. I’m tempted to pull apart my lan­guish­ing Athlon XP, but it feels too pow­er­ful for the task (not even kid­ding… this thing is light­ning fast) and I wouldn’t know what to do with the rest of the RAM in it. My biggest con­cern is track­ing down cap­ture hard­ware that’s Linux or BSD friendly. Ide­ally there’ll be a secu­rity cap­ture card that does PAL at full frame rate and has 4 inputs, because essen­tially that means it’d be triv­ial to add a few extra cards and, all of a sud­den, it’s quite fore­see­able to have a 12 input vision mixer that will key and title away til the cows come home.

One con­cern I have is that the mixer com­po­nent only takes two sources… which is much the same as on any hard­ware mix­ers I’ve used (two buses: select source on A + B bus, mix buses), but it feels really inflex­i­ble. I’d chain them together but think that might neces­si­tate extra gen­lock­ing time and increase over­all latency. I can’t actu­ally think of a usage sce­nario for this one, though, so it’s not a big deal. Because key­ing exists inde­pen­dently of mix­ing it’s not a con­cern of 2 sources + keyed source, and that’d be the main sit­u­a­tion in which such a thing would be at all necessary.

The other cool thing about this is you can mix dig­i­tal and ana­logue sources with impunity. Need SDI? Sure, get an SDI cap­ture card and add an input source. Firewire? Done deal. Same goes for out­put: because you can out­put via FFMPEG, your “vision mixer” poten­tially also encodes an IP-distributable stream simul­ta­ne­ously with real­time out­put to a monitor.

This is an ines­timably cool piece of soft­ware, but the most bril­liant thing is it isn’t really any­thing new. I dis­cov­ered it because I was look­ing for EffecTV which I’d last used in a pro­duc­tion con­text over 12 months ago… Gephex uses exist­ing open-source fil­ters and pro­cess­ing solu­tions and just pro­vides an excel­lent means of chain­ing them together. You can cre­ate some excel­lent motion art­work with it, but the most excit­ing thing for me is that it enables use of cheap and dis­pos­able x86 hard­ware in place of hideously expen­sive and pro­pri­etary (read: more expen­sive, but also inex­ten­si­ble and not par­tic­u­larly flex­i­ble) solu­tions that the ‘pros’ use.

Increas­ingly I’m dis­in­ter­ested in ‘pro­fes­sion­al­ism’ about this sort of thing, because that’s way out of my price league and, to be hon­est, the most com­mon place I wish this tech­nol­ogy were applied is in church and Chris­t­ian event con­texts, where (even if there is money) no-one is inter­ested in effec­tive com­mu­ni­ca­tion through applied tech­nol­ogy. So we con­tinue to try and push for­ward with no money and a bunch of inno­v­a­tive and irrev­er­ent (to the pros) solutions.

Ulti­mately, it’s about achiev­ing excel­lence in the qual­ity and nature of the work done to share the gospel and build up the body of those who fol­low Jesus — but excel­lence can be attained with­out even a smat­ter­ing of ‘professionalism’.

That said, I’d still love to own an MX-70.