Sunday, January 31, 2010

Windows media encoding and virtualdubmod batch jobs

Investigating batch converstion with Windows media encoder, found this resource which includes info on the command-line encoder. This page has the info in a more readable form - just
cscript.exe c:\path\to\wmcmd.vbs -loadprofile encoding-profile.prx -input "F:\input-folder" -output "F:\output-folder"
This page explains how to get the prx files from the WMEncoder, Properties -> Compression -> Edit -> Export button. But of course, it did not work for me,
.Error occurred in transcoding: Error Code = 0xC00D002F
since my WME does not work with Xvid encoded files, needs a nominal crop of a couple of pixels before it works. Probably then it uses a different input codec or whatever.

But checking out cscript.exe wmcmd.vbs /? found that it supports WMEncode wme session files too! So next time I can try that.

Virtualdubmod also has a batch encoding option with input and output folders, as given at this page and I used it this time: File -> Job Control, and Edit -> Process Directory from the Job control window.

machine translation of Indian languages

This IIT-built automatic translation engine is interesting, pointed out to me by BK. If you don't prefer the Hindi keyboard, you can use other tools like gmail's or google's or others like writeka, then copy-paste into the duur sampark page. The attached sentence took nearly a minute to translate, and it seems to be highly Hyderabadi Telugu! (Click on the picture below for larger image).

Saturday, January 30, 2010

voddler

Voddler seemed a good service, for watching ad-supported movies. But it really needed a 2.5 Mbps connection, so not suitable for us - too much time spent buffering. The interface is a bit quirky too, since they have made it "remote-friendly" for people running media center PCs or whatever, but PC unfriendly, not allowing mouse-clicks! And officially only open for people in Sweden, of course.

disabling TRACK and TRACE

Testing our web servers with the OpenVAS tool (fork of Nessus) at hackertarget.com, it pointed out that TRACK/TRACE need to be disabled. On IIS, it turned out that actually TRACK and TRACE are disabled, verified with telnet as given here.

telnet media.radiosai.org 80
Trying 66.249.27.178...
Connected to media.radiosai.org.
Escape character is '^]'.
TRACK / HTTP/1.0
Host: foo
A: b
C: d

HTTP/1.1 501 Not Implemented
Content-Length: 0
Server: Microsoft-IIS/6.0
MicrosoftOfficeWebServer: 5.0_Pub
X-Powered-By: ASP.NET
Date: Sat, 30 Jan 2010 04:30:44 GMT
Connection: close

telnet media.radiosai.org 80
Trying 66.249.27.178...
Connected to media.radiosai.org.
Escape character is '^]'.
TRACE / HTTP/1.0
Host: foo
A: b
C: d

HTTP/1.1 501 Not Implemented
Content-Length: 0
Server: Microsoft-IIS/6.0
MicrosoftOfficeWebServer: 5.0_Pub
X-Powered-By: ASP.NET
Date: Sat, 30 Jan 2010 04:30:44 GMT
Connection: close


On krishna, had to disable TRACE - for doing this, googled a bit, finally howtoforge showed the simple way of just adding
TraceEnable Off
to /etc/apache2/apache2.conf and restarting the web server.

PS. Initially I had tried with the code for re-write as given at many places, like here
# disable TRACE and TRACK in the main scope of httpd.conf
RewriteEngine On
RewriteCond %{REQUEST_METHOD} ^TRACE
RewriteRule .* - [F]
RewriteCond %{REQUEST_METHOD} ^TRACK
RewriteRule .* - [F]

But this did not seem to work. For enabling mod rewrite, had to do sudo a2enmod rewrite as given here.

Sunday, January 24, 2010

Tips for clean audio

A "request for sermon" about tips on getting cleaner audio, and I obliged.

While recording,

1. Record with minimum ambient noise - switch off a/c etc. Don't use mics like wireless mics which give a hiss.

2. Have both shotgun mics and lapel mics for both guest and host in case of video interviews, or have dynamic mics close to speakers' mouths in case of audio-only interviews.

The rest of the stuff has to be done in the post-processing.

1. For each track recorded, do noise reduction using Adobe audition or cool edit pro or whatever, my settings in CEP 2.0 are as shown below, after doing a "Get Profile from Selection" using a section with only noise.



2. After the NR, doing a suitable noise gate, like the one below, is recommended.



3. Only after these steps, if the different tracks (interviewer and interviewee) are merged, and then normalized, we can have best results.

Wednesday, January 13, 2010

RAID failure on mp3server

The Melbourne uplink had a couple of RAID failures this week. My first suspect is of course the power supply for the external RAID enclosure...

duplicate entries in search results

A listener pointed out the duplicate entries which were being thrown up in the search results, which we knew about, but had not yet fixed. The initial entries to the files table had white-space before the filename, this was making them "distinct" from the newer entries of the same files without the whitespace. Trying to find them using this technique,
select c1.fileid, c2.fileid, c1.filename
from `radiosai_file_master` c1, `radiosai_file_master` c2
where c1.fileid < filename =" c2.filename
did not work, since the filenames were actually not exactly the same: this was throwing up filenames which had .mp3 and .MP3 extensions. Then tried with a wildcard before the filename, found it matched. Then, making queries of the form
 SELECT *  FROM `radiosai_file_master` WHERE
`fileName` LIKE CONVERT(_utf8 '__SPECIAL%' USING latin1) COLLATE latin1_swedish_ci
(note the two underscores - meaning two characters) found that there are two white-space chars before the filename. CR and LF is my guess. Couldn't find the way to properly search with an SQL Like statement for "Like 'CR+LF+%', since
 SELECT *  FROM `radiosai_file_master` WHERE
`fileName` LIKE CONVERT(_utf8 '__%' USING latin1) COLLATE latin1_swedish_ci
returned
SELECT *  FROM `radiosai_file_master` WHERE 1
(ha ha!). So, did individual searches for __BV%, __TALK% and so on and deleted those files. Hopefully all are covered now.

Friday, January 08, 2010

withdrawing from some activities

Have stopped day-to-day involvement with some of the sites given on the left-hand side. Slowly other people are taking up various responsibilities. All for the best. The college website sssu.edu.in has a new CMS-based site, and the alumni site saistudents.org has also been relaunched...

application framework on javascript

Interesting project on freshmeat - WAJAF
Demos seen on left-hand side links work, including drag and drop, animation etc.

mounting remote volume over ssh

Briefly thought about remote mounting, googling found this link, but did not implement for video encoding of remote files etc it since my fiber media convertor frequently breaks the connection.

email with attachments on the qtek mobile phone

I think the only way to send email with attachments is to use the built-in pop3 client on the WM5 mobile phone. There was some configuration issue, so I followed the posts at howard forums which linked to the wiki giving the POP info. Mainly,
Messaging -> Menu -> Tools -> Accounts tab -> tap hold on an account to delete.

reducing table size on mysql

As a follow-up to this post, This is what I did in October: after deleting, also did
OPTIMIZE TABLE `radiosai_stream_daily`
That reduced the disk usage from 52 MB for that table to 9 MB. So, Optimize table is what reduces disk usage.

blog becoming useless? Or not?

Now, site search using Google as well as the search bar on top left give no results for many searches, even for recent posts. For example, to find this post, I searched for Delete From Where but it did not find any matches. Gmail's internal search is finding matches, so maybe I will have to log in to gmail and search there! Bing and Yahoo also do not index this blog much, I think.

Edit: Googled, found this thread, and tried cache:hnsws.blogspot.com in google search. Showed last indexed in Dec 2nd week. That's not so bad. Then, found Delete From Sql site:hnsws.blogspot.com does return the relevant post. So, maybe the <pre> tag contents are not indexed?