v2.1.1

• More type annotation
• Avoid changing types of variables in runtime.
• Avoid changing values of parameters of methods(they stay the same throughout their life).

v2.1.0

• The videohash code is now type annotated.
• The black bars in videos are removed before calculating the videohash value.
• The extension of the downloaded video is set to .mkv if the video lacks an extension because the Matroska container can contain all codecs.

v2.0.4

What changed?

Nothing. The purpose of this release is to fix the page on PyPi and the project now has a logo. (#45 )

v2.0.3

• BUGFIX: On Windows, FFmpeg and youtube-dl/yt-dlp could fail because of the space in the directory name. This issue was fixed by quoting the paths of these programs.
• CHANGE(non-breaking): A new attribute bitlist in the VideoHash class, the attribute is python list containing the bits of the hash value.
• CHANGE(non-breaking): The package now XOR's the bitlist instead of calculating the Hamming distance the usual way.
• Retired Python 3.5 is no longer supported.

v2.0.2

• All exceptions are now available directly from the videohash import. (imported exceptions in videohash/init.py).
• Explicitly set collage_image_width=1024 in class MakeCollage, this changes nothing as it was set to 1024 while creating the object in VideoHash class.
• shlex quote output_dir for POSIX in videohash/framesextractor.py

v2.0.1

• v2.0.1 is incompatible with the version one of videohash(1.x.x). Sorry!
• The version 2 usage wavelet-hash(based on Discrete Wavelet Transformation) for hashing the image. wavelet-hash as implemented by Dmitry Petrov in the image hash library. From now on, the whash of the collage is the videohash for the video.
• Supports both youtube-dl and yt-dlp.
• Operators such as '-', '==' and '!=' are now supported by the VideoHash class instances. The instance can be compared with other VideoHash class objects and also python string objects of the 64-bit binary hash and its hex representation.
• The generated collage is now accessible to the end-user and not deleted right after its use to calculate the hash. In fact, the instance doesn't delete the files created by itself unless delete_storage_path() method is called on the instance.

v1.0.7

• Unbreak video paths containing spaces

v1.0.6

• Standardize the videos before calculating the hash.
• It may be assumed that the standardization will increase the time to calculate the hashes but we are also decreasing the size of file by downscaling and reducing the frame rate.

v1.0.5

• Windows OS support added.

v1.0.4

• Make the collage as close to a square close as possible. This change decreased the hashdiff of partially similar files by 15.33%(mean), tested on 164 partially similar NASA public domain files.