Sound is interpreted as word

[Alexthurwood]

Hi bro,

I loved the package and while exploring I have encountered couple of challenges.

1. If a video has a timer with sound of 8 seconds and when I apply pycaps on top of it. Output video shows the word beep for every second. I am trying to avoid that word showing on the screen but could not. Could you advise how can I stop the word being shown for the timer sounds please

2. I have set min and max characters and max and number of lines to 1 to have one line of captions in the video but for some reason the captions are showing in two lines. Could you advise how to tweak the configuration to show captions one per line please

If you need additional information let me know please.

It's a wonderful tool. Thanks a lot for your contribution to the community. Much appreciated.

[linn33]

I also encountered the same issue. I set the maximum number of lines to 1 so that the video would only have a single line of subtitles, but for some reason the subtitles are displayed as two lines, or even three.

[francozanardi]

Hey @Alexthurwood @linn33

Sorry for the delay in the response.

Are you setting the lines/characters using a template (json file)? Could you guys provide an example?

There're two different steps in the pipeline:

• segment splitting
• line splitting

Here you can find more info about what's a segment and what's a line.

Segment splitting

After transcribing the video, you get a list of segments. These segments depend on the transcriber used. If you're using whisper, it normally creates a segment for each sentence (so they're very long).

The segment splitter receives this list of segments, and tries to divide each segment in shorter segments, using the configuration provided. But here's the thing: it could ignore some of your configurations if it's impossible to resolve all of them.

First of all, there're different splitters provided in the library, and you can chain them. So, I don't know exactly your configuration. However, I could put as example the LimitByCharSplitter. It accepts minChar and maxChar:

• let's say you have a segment with 40 characters, composed by two words of 16 and 24 characters each. Then you set up minChar = 20, and maxChar = 35.
• 40 is greater than 35, so we try to divide it into two segments
• after splitting it, we have a new segment of 16 characters, and other of 24 characters.
• the segment of 16 chars is invalid because of minChars = 20
• final result: the operation is cancelled, the segment remains with 40 characters, even though maxChars was 35.

Line splitting

Here's something similar. When you set max lines, it will try to split each segment in lines, using that configuration. However, it will not divide the segments again.

So, let's say you have a very long segment (a whole sentence of 100 chars). And then, you set maxLines = 1. It will create a very long line for all the text in that segment. However, there's another configuration*, which prevents the text from going away from the video. So, if this line is too long, it will be divided again, ignoring maxLines. It's because the text doesn't fit on the video using a single line, and because the line splitter can't divide the segment in two different segments (that's a previous step).

(*) I don't remember the name right now, but if you need, I could find it.

---

So, as you can see, maxLines can be ignored if your segments are too long. Probably you need to adust that first. You could read the existing templates to get an idea maybe!

Sorry for the poor format, I'm from my mobile right now.

[francozanardi]

Hi bro,

I loved the package and while exploring I have encountered couple of challenges.

1. If a video has a timer with sound of 8 seconds and when I apply pycaps on top of it. Output video shows the word beep for every second. I am trying to avoid that word showing on the screen but could not. Could you advise how can I stop the word being shown for the timer sounds please

First of all, thank for your kind words @Alexthurwood :)

If you're using pycaps for automation, there's not an easy way to ignore these sounds. The easiest option could be changing your transcriber or try to use another whisper model. For default, pycaps use the "base" whisper model, but you could use a better model like "small" or "medium".

If you're editing each video manually, pycaps has a built-in captions editor, so you could correct the wrong transcription there.

As I said before, I don't know if you're using it from the CLI, or directly as a library from Python. From the CLI, these arguments in pycaps render could be useful:

• --transcription-preview: Stops the rendering process and shows an editable preview of the transcription
• --whisper-model: Whisper model to use, example: --whisper-model=base

[Alexthurwood]

Thanks for your response. I am using the package, but I shall try the cli mode.