Blur and silence everyone who has not agreed to be filmed
Make a copy of a video with every face blurred and every word silenced except those of the people who agreed to be filmed, before it is posted or shared
Give video, a video in the person's files, and agreed, pictures of each person who agreed to be filmed (a clear picture of their face; the person themself included). Every face in every frame is found; a face that matches someone who agreed stays as it is, and every other face, including one it cannot be sure of, is blurred past recognition. Give agreed_voices too, a recording of each of them speaking, and every word said by anyone else is silenced, judged half a second at a time; without it the sound is copied unchanged. The original video is never changed. The copy is kept in the person's files under Made by tools, named by name when given. Returns how many faces were seen, blurred and kept, how much speech was silenced, and any agreed picture in which no face was found. Faces are found with YuNet and recognized with SFace (OpenCV Zoo) and voices told apart with CAM++, on the tool's own computer; nothing leaves it.
What it asks for
- video (Needed, file): The video, by its identity in your files
- agreed (Optional, list): Pictures of each person who agreed to be filmed, by identity in your files; with none, every face is blurred
- agreed_voices (Optional, list): Recordings of each person who agreed speaking (ten seconds or more each); every word by anyone else is silenced, and with none, every word is
- voice_match (Optional, number): How alike speech must be to an agreed voice to be kept (default 0.5); higher silences more
- match (Optional, number): How alike a face must be to an agreed picture to be kept (default 0.363); higher blurs more
- name (Optional, words): What to call the copy, like Serve practice without strangers
Use it from your AI
Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this tool by what it does and runs it.
Say to your AI: “Use Trillion to make a copy of a video with every face blurred and every word silenced except those of the people who agreed to be filmed, before it is posted or shared”
What your AI calls: run_tool {"tool_id":"tool:blur-faces-not-agreed"}