a looking-glass book

"7 years from now...

...everything you do will be recorded"

the trailer for 'a scanner darkly' is up.

and id say seven years is maybe about right actually. as for the whole sides of the brain fighting...well, one can hope.

first thoughts on ffmpeg-php

1. this, esp as a php extension, is not made to pull 300+ images out of a video file on the file. it takes upwards of 6+ minutes to get through those 300.

2. it seems to slow the entire server to a crawl when it runs through the process. this obviously bears more investigation (and im sure marientina will let me know if im already destroying the speed).

3. this is all after compressing the movie to a smaller size - taking a 90 minute / 130 meg movie down to 30 megs. this was done by compressing the quality, the size and dumping the audio.

4. a solution would be the following: get the frame number. see if there is a file existing already for said frame. if so, output the existing frame. if not, create a new file and output the frame into it. then display that new file. the problem with this solution is this: its no better in the long run than dumping every frame out into a directory to start with. this new way requires some file system checks which might be faster than outputting the frame. but they system is predicated on the weighted good moments coming to the surface before you save out every frame of the movie. if we do it this way even for 1 frame a minute, the entire movie could be saved out in as few as 900 reloads. if we do it for 20 second intervals, then its 300 reloads. i am not convinced this is worth it. esp when you end up with 80,000 images in the folder. filesystems dont seem to like this.

so what do i do? well, for now my design shifts slightly and i see what happens with 1 frame per minute and letting the page take a little longer ot load. its a balance and if everyone complains, then i keep trying to find a solution. but this is one of those 'in 5 years...' kind of issues that doesnt seem truly within my scope this year to solve. i have bigger fish to fry right now than making lightning fast webpages. unless everyone starts complaining.

in the meantime, back to rewriting code....

success!

this might not look like much.

but it is frames of a movie, a qt file, pulled out dynamically on the server.

this is the culmination of over two months of trying to get this working on a server. this could not have been done with help from the developer, todd kirby, and marientina.

this bodes very well for much of what i want to do.

in looking at my production schedule, i am a little behind in coding (about a week) but i think that is made up for in movies. i have missed only 2 days since about the start of this month. i have 81 movie files, 99% of which will be published in the coming weeks. i do not know how many hours this is, but i do not fear hitting the magic 200 hour mark.

much of my backend code needs to be rewritten in the next week to allow for these new thumbnails. this will speed publishing and once this new code is in, things will move very very quickly - ill be able to publish a day after filming with very little post (basically compressing the movie and uploading it to the server - a much easier process than the current one).

even if i lag behind by a week, with spring break and gdc approaching, i will catch right back up to my production schedule with no worries. i am glad i gave myself some production wiggle room then. of course, there is still much to be done. but isnt that always the case?

db updates and onunload

bear with me as i sidestep this absurd ffmpeg-php installation and all its woes and think about 1 more piece. im just throwing this up so i can find these notes later; avoid this if you dont care about me thinking through a small little block of (complex) code.

so.

what i want is to keep track of when people click a thumbnail on the page. because if they click it, they want to watch that segment. which means they found it interesting somehow. so i can create an array in javascript of timecodes. actually, it can be a string. everytime you click an image, the 'onclick' is fired, calls a javascript function that adds that timecode to the string.

then, when 'onunload' is fired, another javascript function is called. this one loads a php file in a hidden iframe that updates the database. all we want is for each timecode to include a counter. (the db will have a primary key, the movie id, the time code and a counter. this could get pretty large.)

so why?

because then we can display the thumbnails based on interest. say a scene has 5 clicks registered with it. then we want to display that thumbnail over one with 1. though this sounds dangerous over time if weighted unfairly, i think thats just a balancing act and a set of ratios.

this way, we can track overall habits/interests of the viewers and turn around use that data invisibly to provide automated 'highlights'.

(between this bit and the ffmpeg-php, looks like the entire system will require a rewrite in a week or so. or basically as soon as ffmpeg-php is running correctly. whee iteration!)

no rest for the wicked

or the healing.

picked up the repaired sensecam down in long beach this morning. props to my sister for driving my pain-killered butt down there to get it. expect multitudes of pictures now - jim said in his test run last week, he took over 1400 in a day.

working with marientina on getting ffmpeg-php working on the server. this has been a much more frustrating experience than i would have guessed. she and i spent a couple of hours on it yesterday and then more time today. and this was after spending a good deal of time on it before i got sick.

does look like itll be fully functional in a few days at most, so thats good. i just hate the fact that even though im spending a lot of time on thesis work right now, its just installing some software. its one of those victories where you dont have a lot to show for the time spent. though once it is 100%, ill be able to pull out movie frames on the fly. this should be cool enough to warrent the time spent.

now i just need to figure out how to:
update a database on a click or on page unload.

im not sure this is actually possible.

e.r.

so tonight's episode of e.r. has a ton of pov video footage in it. i was hoping for something a bit more experimental, where we didnt leave the character's pov for the entire episode, but within the first 4 minutes we have already spent more time outside the character and jumped back and forth several times.

cant win them all, but it is interesting to see a recent example of pov video in such a commercial program.

back

im out of the hospital and back on my feet. mostly. if you want more info about the last week and half, check my site (roxy and carter posted a lot about it).

the upside is that i filmed every day i was in save one. not every movie is great, but at least they are unique. at least i feel like i was somewhat productive, even while stuck on my back, having morphine with a tube up my nose pumping my stomach.

bad engineering

the last couple of weeks i have been recording a lot.

or, rather, trying to.

problem #1 was the sensecam. it just didnt like me. so it is on its way back to england and i should have another one (or this one repaired) in the next week.

i have 1 days worth of pictures from it, though it didn't really do what it should have. i need to write a php file that will allow you to see all the images from that folder. thats on the table for this weekend hopefully - depends on how long the blog is down.

problem #2. this one was much larger. my powerbook was acting up. i would slide it into my bag to film a movie and *zoink*. out like a light. not asleep. not on. not on to the point it would reset my system clock. thats some bad news.

the failure rate for making movies was going through the floor. i usually can hit at about 70% success rate. meaning i get the file. sometimes the system dies, sometimes the camera gets fudtzy and loses connection. but i would say i get about 70% of what im trying to get. sadly, i usually lose it when im moving, which is far and away the most interesting stuff.

i went to the apple store today to see what they could tell me.

long story short, turns out it was the firewire cord for the isight.

note to everyone:
1. apples firewire cords blow. out of the two i have owned, this is the second that has gone bad on me.

2. thanks to some wise engineering, when that wire cuts out, the camera doesnt skip frames. it doesnt disconnect. it doesnt even crash the app using the cam. it doesnt put the computer to sleep. no, it cuts all power to the machine. all power. wiping the time/date and everything.

who was the ad wizard that came up with that campaign?

but, that said, im back in business. lets make some movies this weekend.