Rumi provides a free Chrome extension that runs AI inference locally on desktop video streams to produce timestamped text annotations of visual objects, audio cues, and scene metadata. Only these short annotations are sent over encrypted channels, preserving privacy while users earn compute‑based points redeemable for rewards or cryptocurrency. The anonymized annotation dataset is licensed to interactive media and contextual advertising platforms for real‑time content enrichment.
Funding
Funding not disclosed
Founders
Product
Problem
Consumers watch video content on desktop browsers without any incentive to contribute compute resources, while AI developers lack large‑scale, privacy‑preserving annotations of visual and audio streams needed to train contextual models. Existing data collection pipelines often require screen recording or upload of raw media, creating privacy risks and limiting user participation.
Solution
Rumi delivers a free Chrome extension that runs entirely on the user’s machine while they stream on supported platforms. The extension performs on‑device inference with proprietary AI models that index visual frames, audio tracks, and title metadata, then emits concise text annotations (e.g., character names, scene descriptions, emotional tone) to a decentralized network. Because only short textual metadata leaves the device, raw video and audio never leave the user’s computer, preserving privacy by design. Users earn points proportional to their device’s compute capacity and streaming minutes; points can be redeemed for rewards or exchanged for cryptocurrency. The aggregated, privacy‑filtered knowledge base is made available to interactive media services and contextual advertising platforms, enabling richer, real‑time experiences without compromising user data.
Target Audience
The primary users are desktop streaming consumers who watch video on supported platforms and are interested in earning rewards for their compute contribution. Secondary customers include interactive media providers and contextual advertising networks that require privacy‑safe, high‑granularity content annotations.
Features
- On‑device AI inference pipeline that extracts visual objects, audio cues, and timestamped scene descriptors from permissioned video streams.
- Zero‑knowledge data flow: no screen captures, no raw audio/video upload; only short text annotations are transmitted over encrypted channels.
- Compute‑based reward engine that calculates earnings from CPU/GPU contribution and streaming duration, issuing redeemable points in real time.
- Multi‑mode analysis: (1) viewership ranking, (2) audio emotion and language parsing, (3) content summarization with key characters and settings.
- Seamless integration with major desktop streaming services via a lightweight Chrome extension; automatic activation on supported tabs.
- API endpoints for advertisers and interactive media partners to query the aggregated annotation database for contextual ad placement and dynamic content linking.
- User‑controlled privacy dashboard allowing data export, revocation, and instant uninstall of the extension.
- Compliance with industry‑standard encryption (TLS 1.3) and data‑handling best practices to meet GDPR and CCPA requirements.