Rare find

AvatarAI. Turn any photo into a talking person with a cloned voice.

github.com/PunithVT/ai-avatar-system
by @PunithVT

Vaya's read on this project

Problem, audience, market, and the verdict — sign in to see it.

Updates

July 2026
  • fix(musetalk): ship musetalk_worker.py so lip-sync actually works (#5)
June 2026
  • ci: make deploy.yml manual-only (ECS target does not exist yet)
  • fix(backend): reap real temp dirs; harden MuseTalk worker + celery te…
  • fix(frontend): video errors no longer hang playback; release media on…
  • fix(db): support SQLite for local/self-contained dev (#4)
  • feat(tts): Edge-TTS neural fallback chain; accept legacy coqui provider
  • feat(llm): local LLMs via LLM_PROVIDER=ollama and OPENAI_BASE_URL
  • docs(readme): refresh to current stack; add comparison, What's New, FAQ
  • fix(avatars): metadata PATCH silently never persisted
  • feat(onboarding): demo avatar seeding + prebuilt GHCR images
  • docs: add MIT LICENSE; align README/description with search terms
  • feat(seo): rich metadata, JSON-LD, robots.txt and sitemap for discove…
  • fix(frontend): unmount cleanup leaks, send guards, baked NEXT_PUBLIC_…
  • fix(config/middleware): 429 responses, env list parsing, CI secrets
  • fix(pipeline): WebM mic audio, playable S3 chunk URLs, real WS close
  • fix(db): cascade deletes for sessions/avatars; safer avatar delete order
  • fix: STT ignores session language; conversation message_count is stale
  • fix(frontend): make "Open" from history actually resume the conversation
  • test(ws): end-to-end turn through the real ASGI WebSocket endpoint
  • fix(avatars): crash on missing thumbnail_path; add end-to-end journey…