Why does face recognition eliminate the alpha channel?
If I start with a PNG image with transparency for the background, generating a transformation using :crop => :thumb and :gravity => :face turns the transparent pixels into black pixels.
Is this a bug? Is there a work-around that maintains the transparent pixels?
Sure! How about this:
Original image: https://res.cloudinary.com/hb3rxe9xf/image/upload/v1390949516/akkbaypalx3pesi7srzq.png
After face detection/cropping: https://res.cloudinary.com/hb3rxe9xf/image/upload/c_thumb,g_face,h_100,r_10,w_100/v1390949516/akkbaypalx3pesi7srzq.png
Note the transparent background in the original image and how the background turns black in the cropped image. If the url doesn't specify g_face, the transparent pixels remain.
The issue is now fixed:
Thanks again for reporting about this.
Post is closed for comments.