Built for LIVE Creators & Their Core Patrons

Honoring stream moments.
Presence, not spend.

Riffs transforms multi-hour livestream interaction archives into private, cinematic narrative tribute reels. Sent directly to top patrons to acknowledge authentic presence — without predatory gift-farming or vanity metrics.

⚡ Native iOS Client • 🔒 End-to-End Private DM • 🧠 PydanticAI Structured Reasoning

The Anti-Gimmick Protocol

Social connection shouldn't be predatory.

Presence Over Spend

Existing live platforms reward aggressive gift-farming. Riffs detects and celebrates consistent conversational engagement, stream time, and authentic community stewardship.

Private DM Tribute

Tribute reels are generated privately and delivered directly from creator to patron. No algorithmic vanity feeds, no public leaderboard shaming, and zero surveillance ads.

Nuanced Story Arcs

Multimodal LLM analysis parses multi-hour livestream logs to extract genuine shared jokes, emotional milestones, and community stories into tight, cohesive narratives.

Intelligent Pipeline

Autonomous storytelling from raw livestream feeds.

Processing live video and chat at scale requires deep contextual understanding. Riffs uses frontier reasoning models to turn chaos into high-fidelity emotional cinema.

→ Event Stream Ingestion: Ingests live chat events, timestamps, and key interaction clusters.
→ Storyline Extraction: Claude 3.5 Sonnet structures storylines via Pydantic schema validation.
→ Native Reel Assembly: Automated cutting, localized Portuguese (PT-BR) narration synthesis, and dynamic typography.

pipeline/spec_builder.py

class TributeStoryArc(BaseModel):
    creator_id: str
    patron_handle: str
    dominant_emotion: str
    story_beats: List[StoryBeat]
    script_pt_br: str = Field(
        description="Authentic, culturally-grounded narrative script"
    )

# Autonomous assembly via Claude 3.5 reasoning
arc = await agent.run(stream_events, result_type=TributeStoryArc)
✓ Story synthesized: 4 narrative beats, 42s target duration

Provenance & Ethics

Committed to AI Safety & Provenance Standards

Riffs adheres strictly to ethical data provenance. We do not crawl unauthenticated data, we do not train generative clones without explicit biometric release, and we believe AI should strengthen real human appreciation, not substitute it.

Self-Hostable Architecture • Zero Behavioral Tracking • GDPR / LGPD Aligned

Honoring stream moments. Presence, not spend.