Long loading times - Web Inquisit experiment with videos

Back

Home
»
Millisecond Forums
»
Inquisit 5
»
Long loading times - Web Inquisit experiment with videos

Long loading times - Web Inquisit experiment with videos

Author	Message
lalani98	lalani98 posted 7 Years Ago Topic Details Share Topic
Group: Forum Members Posts: 4, Visits: 9	I have programmed an experiment that, among other things, calls one of four batch files, each of which calls on 6 videos stored on a university server. We are experiencing lags in loading the files that can range up to 8 minutes. With the plan to launch the study to Mturk, this is of course not ideal. We have made efforts to reduce file size, but any further reduction results in lost quality. Links to the server-hosted videos are in the attached iqx files. PC videos are in wmv format and Mac videos are in mov format. Each video is approximately 20 seconds. Total video file sizes are as per below: PC "S" condition: 117 MB PC "NS" condition: 119 MB Mac "S" condition: 64.5 MB Mac "NS" condition: 56.7 MB For what it's worth, I used https://tools.pingdom.com/ to test the download speeds and got 2-3 seconds for each. Any suggestions on how to optimize this would be immensely appreciated! -Lisa Attachments mds2b_vp_sns.exp (628 views, 99.00 KB) mac_ns.iqx (597 views, 367 bytes) mac_s.iqx (620 views, 361 bytes) win_ns.iqx (617 views, 367 bytes) win_s.iqx (621 views, 361 bytes)
	Reply Like 80
lalani98	lalani98 posted 7 Years Ago ANSWER Post Details Share Post
Group: Forum Members Posts: 4, Visits: 9	Web link here: http://research.millisecond.com/unsw_williams/mds2b_vp_sns.web
	Reply Like 94
Dave	Dave posted 7 Years Ago ANSWER Post Details Share Post
Group: Administrators Posts: 12K, Visits: 98K	+x lalani98 - Monday, January 2, 2017 Web link here: http://research.millisecond.com/unsw_williams/mds2b_vp_sns.web A couple of things: #1: The download speeds reported by tools such as https://tools.pingdom.com/ should be taken with caution. They are rarely accurate, nor would they account for various factors such as the overall load level on the university server, the speed of the individual participant's internet connection, etc. #2: Particularly the PC versions of the videos have substantial size -- based on that alone, I would estimate the download time on the order of minutes rather than seconds. #3: If there is no way to reduce size further, there is not much you can do to speed things up. What you can do, however, is opt for "streaming" the videos at run-time rather than downloading them in full before the experiment starts executing. This will lead to reduced time for the experiment to start up, but runs the risk of videos not playing back smoothly if data can't be retrieved quickly enough at runtime: https://www.millisecond.com/support/docs/v5/html/language/attributes/stream.htm If you want to go down that route, set the /stream attribute in the script's <video> element to true as in <video video> / items = videos / select = noreplace / halign = center / size = (59.5%,35%) / vposition = (40%) / playthrough = true / stream = true </video> Hope this helps.
	Reply Like 80
lalani98	lalani98 posted 7 Years Ago ANSWER Post Details Share Post
Group: Forum Members Posts: 4, Visits: 9	Dave - Thank you for your very prompt reply! (and happy new year) I had figured the speed tests were dubious. We will revisit file size and do some testing with /stream = true.
	Reply Like 84
lalani98	lalani98 posted 7 Years Ago ANSWER Post Details Share Post
Group: Forum Members Posts: 4, Visits: 9	+x lalani98 - Monday, January 2, 2017 Dave - Thank you for your very prompt reply! (and happy new year) I had figured the speed tests were dubious. We will revisit file size and do some testing with /stream = true. FYI - we ended up recreating the videos from the source files, sizing down and reducing resolution. Then the trick was to upload to YouTube and download from YouTube - this process reduced the file sizes substantially. We opted not to use stream, as this would introduce variable timing within rather than at the start of the experiment.
	Reply Like 48