github.comPublished
ai-avatar-system
Summary: Open-source, self-hosted AI avatar platform for real-time talking-head lip-sync conversations. Upload a photo, clone a voice from 10 seconds of audio, and converse with photorealistic video streaming at ~30 FPS on GPU. Supports multi-LLM (Claude, GPT-4o, Ollama), zero-shot voice cloning in 23 languages via Chatterbox, token-streaming pipeline with barge-in, and production-grade auth, rate limiting, and observability. Deploys locally (CPU/GPU) or on AWS (g5.xlarge, ECS, RDS, CloudFront).
Description: AvatarAI is a production-ready digital human system. The pipeline: Whisper STT → LLM (streaming) → Chatterbox TTS → MuseTalk lip-sync → video chunks over WebSocket. First video chunk arrives in 2–4 seconds on AWS GPU. Features zero-shot voice cloning, 23 languages, barge-in interruption, 100% local mode with Ollama, JWT + httpOnly cookie auth, per-user rate limits, Postgres + Alembic, S3/CloudFront, Prometheus, CI, and a full test suite. Architecture: Next.js 14 frontend (Avatar Studio, Voice Studio, Chat Interface) ↔ FastAPI backend (WebSocket manager, sentence-split streaming, MuseTalk worker) ↔ PostgreSQL, Redis, Celery, local FS/S3. MuseTalk models ~9 GB. Docker Compose for dev; docker-compose.prod.yml adds GPU reservation, float16, persistent volumes, log rotation. One-command AWS deploy script (scripts/deploy-aws.sh) or Terraform ECS path (infrastructure/). REST API for users, avatars, voices, sessions, messages; WebSocket for real-time chat with token/video chunk streaming.
AI-generated
Interactive React UI Components & Micro-interactions
Bencho is a digital library providing interactive React UI components and micro-interactions. The platform features live demonstrations of various UI blocks, allowing developers to explore functionality and copy code to integrate directly into their own applications.
Top 5 Claude Skills for Cold Email Acquisitions
This guide presents a selection of the top five AI skills designed to streamline the cold email acquisition process for agency owners. Ranked by installation volume from skills.sh as of November 3, 2026, the recommended skills cover the entire sales funnel from initial lead discovery to conversion. The resource includes specific terminal commands to integrate these tools into AI environments such as Claude Code, enabling users to automate complex marketing tasks like drafting, prospecting, and managing multi-step email sequences. The primary focus is on leveraging AI capabilities to reduce the manual effort associated with client outreach and sales follow-up.
A coding-agent skill for multi-phase security audits with independently verified, machine-readable findings - cloudflare/security-audit-skill
This GitHub repository provides a coding-agent skill for conducting structured security audits. It operates through six phases: reconnaissance, coverage-led hunting, candidate validation, structured output, independent record verification, and target-neutral reporting. The system utilizes isolated agents and coverage critic feedback to find security vulnerabilities, ensuring findings are independently verified by agents separate from the discoverers. The skill employs a verdict system using labels such as 'confirmed', 'needs_validation', and 'rejected' based on source trace and observed impact. It supports repeated runs to iteratively improve coverage, with prior analyses used to target gaps and revalidate changed sources. Installation and execution are managed via the Skills CLI, requiring Node.js for validation and an OS-enforced sandbox for target code execution. The skill is noted as the initial prototype that seeded Cloudflare's broader vulnerability discovery harness.