Feature Request: Load waveform from image

I would like to suggest a possible space-saving feature: allow user to upload image of waveform (instead of only generating it from wav files).

Rationale: I’m working with mp4 and mp3 data (I know wav is better, but I have no control over the data generating process), and I need to convert these to wav and keep them around in order to see the waveform, which has become a key cue during data transcription. This causes logistical problems as data sets grow exponentially due to the size differences of these files. I understand that maybe it’s not possible to generate waveforms directly by the Elan source code for files other than wav, but there are some tools available to generate waveform images from a variety of a/v formats (ffmpeg).

Why not give the user an option to enjoy the features of a waveform without eating unnecessary storage?

Thanks, I can add this idea to the wishlist. I understand the rationale but I also see several problematic sides to this approach. A main problem is that of resolution: should ELAN simply assume or require a specific time resolution in the image? Otherwise it will be difficult/impossible to keep the image and the crosshair synchronized (which is necessary to be a cue for transcription). Scaling and zooming will be problematic. ELAN would need to offer a way to indicate that an image is to be treated as a waveform image etc.

Already on the wishlist is to support visualization of the waveform directly from any supported media file (the best solution to the problems you mention). The problem is this has to be implemented for all native media frameworks we currently support/use on the supported operating systems.

Thanks a lot!

I thought about this too. Maybe a hacky solution, but I thought about exporting an image with the same width resolution as the number of samples in the audio file…or at leas some floor-divisioned ratio of img resolution to frames in the audio.

This is indeed a better solution!

Thanks again, Han.

I’m not sure if I understand correctly, but if you would save an image with a width based on one pixel per sample ratio, even a short audio file would result in an image with enormous width. That won’t save storage space, I guess. So, you’ll have to combine many samples into one pixel width and that has to be known by the waveform viewer (with all problematic sides attached to it, as already mentioned).