Hi everybody:
Here's the thing. I've been focus on creating subtitles for old series and movies that do not exist.
Step 1
First convert to audio mp3.
Clean the noise.
Delete music etc.
Upload it to a free video service to get a transcription (you all know which one)
Step 2
Then. Take the transcribed srt. Send it to a spreadsheet to fix the overlaping karaoke thing.
Fix capitals... etc. That service also tends to insert the first one or two words that should belong to the next line at the end of the previous.
So, It is kind of tricky and time consuming.
Then I reupload a corrected srt. Get the translation from the same service. And upload to OS, the eng and spa srt. One transcribed by a machine, and the other machine translated.
Step 3
Then I fix the transcription in english, listening to the video. and yada yada. Reupload.
Get another translation. That translation this time is fixed and checked by a human.
Upload the two files to OS again
---
All these boring story is because I think that the world changed a lot. Everything goes at the speed of light in many aspects of our lives, and so with this.
So, some great help to get subtitles really fast, without sacrificing quality, is producing better transcriptions.
And for that we need. Recorders.
Recorders can create an mp3 with clear english. Modulated. No noise. background music, paused for the break lines, etc. This can be done by watching the movie and repeating the dialogs on the phone or computer with microphone headphones. 10 minutes one guy, 5 min another one
Result is clearly a better transcription from that service, that can be refined ridiculously fast by Transcribers. In order to get a first version of the subtitle in the original language, before *cc's are available.
And therefore better translations.
Can we discuss this?
*cc's, vtt, CEA-708/608 etc whatever.
Also. I've got a serious concern about the space (energy wasted because datacenter, and all of those cut trees, etc) So, as a result of my terrible tendency to ramble, when setting the google spreadsheet I thought in some crazy ways and data model to have at most 2 subs per language. But this is not directly related to the post. End).