Guess what.
I have made subs for the first episode of Mariken!
Mariken is one of the movies (or so I thought it was movie) that intrigued me a couple years ago, and through my enjoyment of many other coming-of-age and childhood-focused adventure tales, it became one of the movies that led me here to FLM.
For those that just want to get to the point, here they are!
Mariken (2000).S01E01.[HEVC.aliptes].en.srt
Synced and created from popdrome's upscale of the series,
shared here.
Episode 2 in the works.
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
[Image]
[Image]
[Image]
[Image]
[Image]
[Image]
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Some details of my adventure with these subs, and other stuff for nerds.
popdrome wrote: ↑Sat Jun 11, 2022 11:35 am
Yes the translation is so difficult because it contains Nether-dutch, Limburgish, Antwerp-dutch, and brabantish, which are all dutch / english forms but very hard to translate.
...
The biggest challenge in this project is actually, trying to understand what is said on screen! Some characters in this series don't just speak dutch, but like I explained, either an older version of it, or a dialect. (Unlike the very extensive variations and long-life of English), Some of these languages are long dead!
Nether-dutch? Wouldn't that be like Nether-Netherland language? Brabantish? How could this be done?
Well, the answer is, RESOURCES. I smothered myself in them and even got a little dazed from it all. It was great. I had a lot of fun doing this. My resources included:
- basic whisper transcription of original first episode (via SE)
- secondary whisper transcription using the v3 whisper CPP model (via SE)
- third transcription of problem areas, cut with LosslessCut and uploaded to freesub.ai (for some reason, freesub can help with tough areas and does well with timings)
- fourth transcription of the same areas using whisper again (via SE)
- fifth transcription of movie version with German dub - Hmmm, can dubs be used to assist? If so, the value of a dub that exists for a movie with no subtitles could be great. And IT DOES work!
- ... if that's the case, can those pesky Russian voice-overs be used? Surely, having audio from two languages being played will disrupt a whisper transcription too much, right? As a proof of concept, I used the Russian VO for the Mariken movie version. And IT WORKED well, incredibly well! Sixth whisper transcription using this source was referenced.
- think we're done? seventh reference, the original Mariken book, translated with help from Mr. ChatGPT, who I chatted with extensively. He is pretty friendly. The ChatGPT translation may be better than what DeepL can do, though many passages come out nearly the same. I referenced the passages of the book that corresponded to their respected scenes, which actually lined up closely.
All transcriptions and the book laid out in notepad docs, all spread across the screen. Cross referencing them all at the same time became dizzying and often led to me getting easily distracted. But it was so much fun. Mariken was more enjoyable than I was expecting, so I wanted this to be my best work. I went all out in making this into the best sounding English I could muster.
For those areas that used archaic or dialectic language, well they had to be interpreted by both the German dubbers and the Russian VO guy, and I was able to use those in close examination of the best original Dutch transcription as I could create with whisper. And many lines were lifted right out of the children's novel, sometimes with slight modification. These all assisted in figuring out these difficult passages.
The part that I worked on the MOST and combed over and over, was the song played during the credits, sung by the main girl Laurien Van den Broeck herself. It is quite a cute little ditty that references Mariken's observations and the ideas of understanding and interpreting knowledge. The song required just as much cross referencing between all my whisper transcriptions, it was almost painful in parts, trying to line up the lyrics one line at the time to seek out differences. Some slight differences in sound, like "ieat" vs "niet", made great differences in meaning.
The song is fairly simplistically written and plain in vocabulary, so it translated pretty directly, that is, once I had the proper lyrics set straight. However, like many song lyrics, they maintain a rhythm, consistent syllable length in cadence with the beat of the song, and a rhyming scheme. Hmmm, could I translate DIRECTLY while maintaining some of these song characteristics? What a challenge! I hope you enjoy the work I put into it (ChatGPT assisted with some ideas for vocab and phrasing). Direct translation was kept and I added no extra splashes to get a loose rhythm and rhyme.
I will post more on the experience tomorrow. I want to touch on some details I've noticed on Russian VOs,
You do not have the required permissions to view the files attached to this post.