Skip to content Skip to navigation

Meeting notes: Subtitling online video and NAB conference debrief

edcorner.stanford.edu - Web project providing subtitled videos in multiple languages
- presented by Forrest Glick on May 1st, 2008 - d.school conference room
~~~
Overview:
Total of over 1,200 video available on the site - all free.
Current subtitle languages include English, Spanish, German, Serbian, Arabic.
Preparing to "crowd source" translation of materials.
Tool for translating written in Java.
Videos are used in other countries, and we are excited about possibilities for translations.
All videos are progressive downloads.
Podcasts are broadcasted the same night. 2,000-5,000 downloads ever day. 1.9 TB in a day - high water mark, so far.
87% through iTunes. 1,000-2,000 video plays per day. So far podcasts are audio only.
Latest version of JW flv player does H.264 - considering implementation.
40% of hits are from outside the US.
Site also allows users to embed videos on another site. New interface will allow the creation of a playlist of videos to be embedded.
To download copies of videos, user needs to create an account. Simple registration process.
Development process of Test server, staging server, production server using Subversion.
Entire platform is built with Open Source tools.
Please get in touch for more info.
See an example of a video with multi-language subtitles at:
http://edcorner.stanford.edu/authorMaterialInfo.html?mid=1716

Process:
Capture video/audio of a weekly lecture series that features guest speakers
Podcast mp3 file the same day - available via iTunes, hosted on Leland servers
Human-generated transcript produced within a week. Transcript is used for editing purposes, select clips from transcript rather than scrubbing around the video. Canadian service produced transcript - $0.90 per minute.
Video is edited down to key moments. Each video is generally 1-5 minutes long.
Edited video is encoded to Flash, Windows Media, and QuickTime formats.
Video and metadata (title, description, transcript, etc) uploaded to server
Transcript is sync'd to video - user can click in transcript to jump to portion of video.
Once English subtitles are available, our tool allows for the translation to other languages.
Translator logs into tool, submits translation, then assigned to another translator for verification.
Subtitles are output as .srt files and made available to users with video
Flash-based player is used for web-based playback of video - JW FLV player provides support for captions.

NAB Overview
~~~
Annual Las Vegas conference. Originally TV, radio - now video and multimedia.
Lot of realtime usage. Do the tools really work? And, what not to do.
Format wars - are they over. Adobe moving to H.264
Silverlight - Microsoft's new toolkit. Being used for the Olympics. XML, but XAML (their own version).
MP3 recording devices, cameras.
Great opportunity to see new hardware, ie Capture cameras for classroom. Panasonic, Sony.
New light systems. LED, too. Nice light, not hot. Lighting just on speakers.
Telestream server engine. 12 GB raw files to H.264 - 8 core servers.
Display walls - 4,000 x 4,000 pixels - 4K projectors.
Ultra HD - 7K by 4K pixel screens from Japan with 12 channel Dolby surround sound.
RedCam 4K is $40,000.
Sony V7 CF disc and tape at same time. Chips last 30 minutes.
Current expectations for production is HD - 16x9
Monitors aren't catching up to the cameras.
Kinglsey - display wall 5x5 (in Med School) 5-30 inch monitors. Slide show only, not video.
All stations in town are broadcasting HD. Jack downconverts to SD. Looks better. In Japan, all 16x9. Pretty soon 4x3 will look like home movies. 720p making gains on the web.
H.264 extremely processor intensive to decompress. Microsoft VC1 codec less processor intensive.
Frame grabs for Stanford Report.
Final Cut Studio demo - collaborative editing around the globe all looking at one machine in real time.