-
Notifications
You must be signed in to change notification settings - Fork 1
08. Development notes
Some misc thoughts and to-do items.
The module pact.bulktranscription contains working code, but it's
slow. Perhaps in the future parts of it can be applied to
auto-transcribe things, or even as a kind of "offline script" that
might be interesting.
Possible future improvements.
I tried using ttk.scale for better styling, but it was garbage. The slider handle kept jumping out of the scale, and not respecting the from_ and to values of the scale (e.g., for from_=2500 and to=4500, a value of 3500 (right in the middle) would be shown about 75% along the scale, and for higher values it would disappear completely).
The current design is good enough -- I can define clips reasonably quickly and easily.
I had tried to use a separate slider to define the "start and end" of the clip, but it was a hassle. The matplotlib chart of the waveform didn't line up accurately with the slider position, and it just felt inaccurate. Using buttons and going by feel was better.
Ideally, the design/UI for this would be something like Audacity's "clip selection", where the user clicks and drags a range on a plotted chart of the audio waveform. I tried various versions of this, but couldn't get it to work reasonably:
In the plot() method, you can convert the bookmark clip bounds to the corresponding axis positions on the chart, and then use axvspans:
# To shade a time span, we have to translate the time into the
# corresponding index in the signal array.
def signal_array_index(t_ms):
span = self.to_val - self.from_val
pct = (t_ms - self.from_val) / span
return len(self.signal_plot_data) * pct
cs, ce = self.bookmark.clip_bounds_ms
if (cs is not None and ce is not None):
shade_start = signal_array_index(cs)
shade_end = signal_array_index(ce)
self.axv = plot1.axvspan(shade_start, shade_end, alpha=0.25, color='blue')
I was hoping to use this to do on-the-fly updates of the chart as the user dragged a slider bar, but the performance was terrible.
Per https://stackoverflow.com/questions/40325321/python-embed-a-matplotlib-plot-with-slider-in-tkinter-properly, it appears that we can use the spanselector in tkinter, but when I tried using the bare minimum code in this app's windows, it didn't work ... the spanslider couldn't be selected. Perhaps this is due to grid being used, rather than pack ... not sure, didn't bother looking further.
Originally in this project, mp3 playback was handled by pygame, but I had problems with pygame not having accurate time when it comes to mp3s. Per https://www.pygame.org/docs/ref/music.html#pygame.mixer.music.play :
"For MP3 files the start time position selected may not be accurate as things like variable bit rate encoding and ID3 tags can throw off the timing calculations."
This time skew was very pronounced in some files.
To swap out pygame, I used a wrapper and substituted the VlcPlayer wrapper in its place, which still exists in the code (even though that wrapper might not really be necessary now).