How behavior actually maps to media.
Public notes from the lab. Themes, not pipelines. Read the work. Then read the spec.
Active threads
Behavior anchored to the recording, not the transcript
The transcript is a translation. The signal — pace, hesitation, push, pivot — lives in the audio. We anchor the behavioral profile to the recording, so meaning survives every transcript edit.
Audio plus video, on one timeline
A transcript cannot carry where someone looked, when they leaned in, when their breath shifted. The .0wav adds those cues to the same second-by-second line as the words.
Skip the transcript when the question does not need words
"Did the room get quieter when she answered?" does not need text. We support behavioral queries that ride the file without ever decoding to characters.
Storage that shrinks without losing meaning
Bytes get smaller every year. Behavioral signal does not. .0wav is the layer that survives — pick the codec your storage budget allows; the file still answers.
License
Notes published as we go. Apache 2.0 on the format and the reference reader. Patent pending on the layer that makes follow-ups free. Deeper technical references ship to design partners and to the public repo at v1.0 launch.