search

CCExtractor / Subtitle-Resync

A tool to automatically generate in-sync subtitles of different versions of the same base media (such as with edits)
GNU General Public License v3.0
23 stars 3 forks source link

Stuck at finding different segments #1

Open Johndirr opened 4 years ago

Johndirr commented 4 years ago

Running

$ ./resync -o lowres.wav -m highres.wav -s blub.srt

I get the following output. But it won't go past this state and never finishes resynchronizing.

[11-30 18:22:40][Debug] Extracting audio fingerprints from: lowres.wav [11-30 18:22:40][Debug] Begin reading WAV file [11-30 18:22:40][Debug] Opening mode chosen: readFile, proceeding [11-30 18:22:40][Debug] Trying to read from file : lowres.wav [11-30 18:22:40][Debug] Reading file data [11-30 18:22:44][Debug] File data read and stored in buffer [11-30 18:22:44][Debug] Processing data and extracting samples [11-30 18:22:44][Debug] Checking chunkID, should be RIFF [11-30 18:22:44][Debug] Wave File chunkID verification successful [11-30 18:22:44][Debug] Begin decoding wave file [11-30 18:22:44][Debug] File format is identified as WAV [11-30 18:22:44][Debug] Finding FMT and DATA subchunks [11-30 18:22:44][Debug] FMT index : 12 , DATA index : 70 [11-30 18:22:44][Debug] PCM : True [11-30 18:22:44][Debug] MONO : True [11-30 18:22:44][Debug] Sample Rate 16KHz : True [11-30 18:22:44][Debug] BitRate 16 bits/sec : True [11-30 18:22:44][Debug] Number of samples : 65917611 [11-30 18:22:44][Debug] Reading samples [11-30 18:22:46][Debug] Successfully decoded [11-30 18:22:46][Debug] File decoded successfully [11-30 18:22:53][Debug] The audio is 4119.85 seconds in length. [11-30 18:22:53][Debug] Audio fingerprints successfully generated.

[11-30 18:22:53][Debug] Extracting audio fingerprints from: highres.wav [11-30 18:22:53][Debug] Begin reading WAV file [11-30 18:22:53][Debug] Opening mode chosen: readFile, proceeding [11-30 18:22:53][Debug] Trying to read from file : highres.wav [11-30 18:22:53][Debug] Reading file data [11-30 18:22:57][Debug] File data read and stored in buffer [11-30 18:22:57][Debug] Processing data and extracting samples [11-30 18:22:57][Debug] Checking chunkID, should be RIFF [11-30 18:22:57][Debug] Wave File chunkID verification successful [11-30 18:22:57][Debug] Begin decoding wave file [11-30 18:22:57][Debug] File format is identified as WAV [11-30 18:22:57][Debug] Finding FMT and DATA subchunks [11-30 18:22:57][Debug] FMT index : 12 , DATA index : 70 [11-30 18:22:57][Debug] PCM : True [11-30 18:22:57][Debug] MONO : True [11-30 18:22:57][Debug] Sample Rate 16KHz : True [11-30 18:22:57][Debug] BitRate 16 bits/sec : True [11-30 18:22:57][Debug] Number of samples : 66148352 [11-30 18:22:57][Debug] Reading samples [11-30 18:22:59][Debug] Successfully decoded [11-30 18:22:59][Debug] File decoded successfully [11-30 18:23:06][Debug] The audio is 4134.27 seconds in length. [11-30 18:23:06][Debug] Audio fingerprints successfully generated.

[11-30 18:23:06][Debug] Finding the different segments...

Any idea whats wrong? The audio files were created as stated in the docs.

sypai commented 4 years ago

Hey @Johndirr , sorry for the delay. Can you provide a link to the audio files?

Johndirr commented 4 years ago

Here they are: https://drive.google.com/open?id=1w0jaCcCYacrpV1yq-ogOp4w9fleqe_m9 https://drive.google.com/open?id=16PD1vzydNtoJrCeOKRC9xZtU7bcLpmQo

sypai commented 4 years ago

Ah! Unable to download the files. I'd suggest you extract the fingerprints for both the files and send them over to me.

You can get fpcalc static binaries from here: chromaprint

It's easy, just run: fpcalc -raw -json -overlap /path/to/audio/file I'll look into what's causing the program to stop.

Johndirr commented 4 years ago

Dumped the output into these files. Hope this will help. lowres.txt highres.txt

I can also upload the wav files to a different hoster if you suggest me one.

sypai commented 4 years ago

Hey! What I can tell after running the code for the two files is that these two files are almost the same. The case that resync currently handles is that of two files with changed timings commonly due to the presence of extra portions (mainly commercials). As far as I can see the current files are the same maybe with different frame rates.

Maybe I can try something but for that, I'll need the files(media and subtitle).

sypai commented 4 years ago

The google drive links seem to be broken.

can't-download-audio-files

Johndirr commented 4 years ago

Hi, you are right. The two files are the same with a different framerate. Uploaded the wavs to another site: https://filebin.net/bz6ihecdnhp1v1gs/files.zip?t=mswa7vao But I don't care that much about the difference for these files in particular. I just want to get the program to work :)

Reuploaded files

Johndirr commented 4 years ago

Ok, I gave this another try. This time it ended in a Segmentation fault

./resync -o original.wav -m new.wav -s orignalsub.srt Subtitle-Resync 0.00, CCExtractor Development.

[05-11 11:20:18][Debug] Extracting audio fingerprints from: original.wav [05-11 11:20:18][Debug] Begin reading WAV file [05-11 11:20:18][Debug] Opening mode chosen: readFile, proceeding [05-11 11:20:18][Debug] Trying to read from file : original.wav [05-11 11:20:18][Debug] Reading file data [05-11 11:20:25][Debug] File data read and stored in buffer [05-11 11:20:25][Debug] Processing data and extracting samples [05-11 11:20:25][Debug] Checking chunkID, should be RIFF [05-11 11:20:25][Debug] Wave File chunkID verification successful [05-11 11:20:25][Debug] Begin decoding wave file [05-11 11:20:25][Debug] File format is identified as WAV [05-11 11:20:25][Debug] Finding FMT and DATA subchunks [05-11 11:20:25][Debug] FMT index : 12 , DATA index : 70 [05-11 11:20:25][Debug] PCM : True [05-11 11:20:25][Debug] MONO : True [05-11 11:20:25][Debug] Sample Rate 16KHz : True [05-11 11:20:25][Debug] BitRate 16 bits/sec : True [05-11 11:20:25][Debug] Number of samples : 105009152 [05-11 11:20:25][Debug] Reading samples [05-11 11:20:27][Debug] Successfully decoded [05-11 11:20:27][Debug] File decoded successfully [05-11 11:20:39][Debug] The audio is 6563.07 seconds in length. [05-11 11:20:39][Debug] Audio fingerprints successfully generated.

[05-11 11:20:39][Debug] Extracting audio fingerprints from: new.wav [05-11 11:20:39][Debug] Begin reading WAV file [05-11 11:20:39][Debug] Opening mode chosen: readFile, proceeding [05-11 11:20:39][Debug] Trying to read from file : new.wav [05-11 11:20:39][Debug] Reading file data [05-11 11:20:45][Debug] File data read and stored in buffer [05-11 11:20:45][Debug] Processing data and extracting samples [05-11 11:20:45][Debug] Checking chunkID, should be RIFF [05-11 11:20:45][Debug] Wave File chunkID verification successful [05-11 11:20:45][Debug] Begin decoding wave file [05-11 11:20:45][Debug] File format is identified as WAV [05-11 11:20:45][Debug] Finding FMT and DATA subchunks [05-11 11:20:45][Debug] FMT index : 12 , DATA index : 70 [05-11 11:20:45][Debug] PCM : True [05-11 11:20:45][Debug] MONO : True [05-11 11:20:45][Debug] Sample Rate 16KHz : True [05-11 11:20:45][Debug] BitRate 16 bits/sec : True [05-11 11:20:45][Debug] Number of samples : 85014528 [05-11 11:20:45][Debug] Reading samples [05-11 11:20:47][Debug] Successfully decoded [05-11 11:20:47][Debug] File decoded successfully [05-11 11:20:56][Debug] The audio is 5313.41 seconds in length. [05-11 11:20:56][Debug] Audio fingerprints successfully generated.

[05-11 11:20:56][Debug] Finding the different segments...

Segmentation fault

@sypai If you would be so kind to have a look at this. I uploaded the files to mega: https://mega.nz/folder/KUgjmCpB#ailTQWAXfl_bpCBu-LnP5g If you can't download from there please tell me a host and I'll reupload.

sypai commented 4 years ago

Hello, @Johndirr Apologies for the delay.

mega doesn't seem to work :( Can we try G Drive once more?

Johndirr commented 4 years ago

Here you go: https://drive.google.com/open?id=15EwO6LTaSWBm8HLtt6CwhoDvB3WkCxiz EDIT: Accidentally closed the issue...

interlark commented 3 years ago

I tried resync and have the same segmentation fault.

interlark commented 3 years ago

The problem is somewhere in slice function https://github.com/CCExtractor/Subtitle-Resync/blob/82e7673c4452056598d9edd46db685a17ea1b9cd/src/utils/align_fingerprints.cpp#L321 According to backtrace, there are calls: main -> Tool::sync -> AlignFP::brum_brum -> AlignFP::segment_it -> AlignFP::next and after it AlignFP::slice have been recursively called like ~70k times until the program died. UPD: That's funny, but it seems like every recurrent call of function slice is called with the same arguments of ub ("upper bound") and lb ("lower bound"). lb=-20, ub=24883. Is it legal for lb be negative, because otherwise it's wrong lb to be compared with type of size_t here: https://github.com/CCExtractor/Subtitle-Resync/blob/82e7673c4452056598d9edd46db685a17ea1b9cd/src/utils/align_fingerprints.cpp#L327 I think because of this wrong type comparison we have segmentation fault here. Forcing with negative values infinity recursive calls.

shootlazer commented 1 year ago

Same thing here. Different audios tried, different subtitles, same result. Shame because I haven't found any other tool that does exactly this.