First question runs the model.
from zerowave import open_0wav audio = open_0wav("q3.0wav") audio.ask("churn anxiety in CFO?") # → pause spike 14:32, pitch -23 Hz
Media with a memory.
One file. Every question. Forever. Ask any recording anything — pay for the first question only.
earnings_q3.0wav · python
1from zerowave import open_0wav
2
3audio = open_0wav("earnings_q3.0wav")
4audio.ask("did the CFO sound nervous?")
5
6# → "Pause spike at 14:32. Pitch drop -23 Hz."
7# → confidence: 0.91 speaker: CFO
How it works
Ask. Re-ask. Compare.
Pay for the first run. Every question after that reads from the file — cents drop to zero, latency drops to milliseconds.
Audio + video, one file.
clip = open_0wav("keynote.0wav") clip.ask("when did eye contact break?") # → gaze drop @ 02:14, frame 3,372
Cross-modal in milliseconds.
from zerowave import compare compare("q3_call.0wav", "keynote.0wav", on="voice_x_gaze_arc")
Pay for the first question Rest live in the file.
$.ϴϴ5 the first question. Pay once, model runs once.
$ every question after. The file remembers.
<1s every answer — cached profile, in-file lookup.
Built in Las Vegas because Lake Mead can’t afford the AI industry to keep wasting compute.
read the Mojave story From the wiki
All wiki entries Read the format.
Concepts, format internals, product reference. The technical and philosophical source of truth.
Research
All research Why this matters.
Position pieces, market analysis, technical deep-dives. Sourced and dated.
Ship behavioral media on day one.
Request a design partner slot → Three slots open, Q2 226.