Thread by @Techjournalisto

Hours of video footage to go thru for #OSINT project? Lil thread on speedy analysis tools
Imaging: you analyze riots of ydays Covid protests in Viennas #w1812. Rioters attacked police. Searching YTube for footage >20 min recorded livestream of “LIVE: Demo in Wien”, 29k views
(0)

(2) We download the video via ur Youtube downloader of choice (list below). Our video is nearly 3 hours long, worth checking the tool MSS #VideoIndexer Azur Media Analyser for Media @Video_Indexer. Open free account. Upload hour long-footage free videoindexer.ai/media/library

(3) MS #VideoIndexer packs a punch: search ppl in video, OCRing & item recog. & trace ppl paths. Detect, group & recog ppl, perfect for cross-refcing of right extremism joining corona demos. The label tag "law Enforcement" reveals scene of barrier breaking

(4)We can also compare footage of with Demo on 11.12.2021, where rightwigs Rutter or convicted rwing Küssels were present & where journalists were attacked. We find one aggressive attendant via MS video tool

https://twitter.com/PresseWien/status/1472620788402601993

https://twitter.com/PresseWien/status/1472621315039416324

(5) Audio: event detection classifies audio effects in the non-speech segments of the content. Caustic events inc “Gunshot, Glass shatter, Alarm, Siren, Explosion, Dog Bark, Screaming, Laughter, Crowd reactions sa cheering, clapping and booing & silence”
techcommunity.microsoft.com/t5/azure-media…

(6) There is also a livestream chat in YouTube video, during the Vienna demo. Use Python3 lib Chat-replay-downloader via “pip3 install chat-replay-downloader” & export the full chat via CSV: “chat_replay_downloader '' --format_file csv -o chat-replay.csv”

(7) if u aim really high on analysing livestream chat content, a free account of MeaningCloud allows machine learning to extract topics, analyze sentiments, identify lang, cluster/categorize text. Just set up a Google sheet, use license key, and Boom, analyze text.