Acoustic Tagging
Recognize environmental sounds, instruments, and speech events with calibrated confidence scores.
CactusBrain Signal / Developer Preview
On-device audio intelligence from live microphone streams.
Tag acoustic events, enhance noisy speech, and process live audio buffers with microsecond latency. Audio frames never leave the device hardware buffer.
MICROPHONE STREAMPCM Buffer
↓SIGNAL RUNTIMEQuantized Acoustic Model
↓ACOUSTIC TAGSConfidence scores
01 / Implemented SDK behavior
PCM Streaming
Pass audio frames directly from CoreAudio / AudioRecord without file writes or intermediate disk serialization.
Speech Enhancement
Filter background noise and isolate vocal frequencies using lightweight neural filters.
Sample Rate Agnostic
Inspect required capture geometry dynamically with SDK format contracts to avoid conversion artifacts.
02 / Swift API
Clean geometric audio input contracts.
Signal adapts to your audio pipeline. Provide raw PCM samples or file URLs and get structured results immediately.
import CactusBrainSignalSDK
let session = try await CactusBrainSignalSDK.configure(
projectKey: Secrets.cactusBrainProjectKey
)
// Tag sounds in real-time from microphone PCM or audio files
let format = try await session.captureFormat
let result = try await session.tags(for: .file(recordingURL))
for tag in result.tags {
print(tag.label, tag.score)
}03 / Edge Efficiency
Engineered for battery life and low CPU overhead.
Acoustic processing is optimized with NEON vector instructions and Apple Neural Engine acceleration, ensuring your background audio tasks consume minimal power.