Syclus123 / Awesome-Interactive-Model

# Awesome Interactive Model
A curated map of voice / Omni interactive models organized by how interaction is realized —timing, think/speak scheduling, and effect scope—not by a cascade→E2E→dual-system “evolution ladder”.
 540 curated entries (🔓 open weights: 155, 🧩 code / runtime: 145, 🔒 closed API / product: 75, 📄 paper only / unreleased: 165). Verified sources as of 2026-10-19. 
---
## 📰 News & Updates
- **2026-10-19** 🎉 **Our work, Jarvis**, gives coordination a brain of its own through a dedicated harness that keeps realtime dialogue and background reasoning in sync for full-duplex Audio/Omni agents. Paper · Code · Demo
- **2026-10-15/16** — OpenAI **GPT-Live-1.1** rolls out with new voices; API adds native **Decisions** delegation and **prompt-to-voice**.
- **2026-10-14** — xAI **Grok Voice 3.0** (`grok-voice-latest` now routes here; 2.0 deprecated).
- **2026-10** — Research wave: RUBATO, D2, DiffuPlex / Diffuplex, SteerablePlex, DuplexAgent-RSI, DuplexAct, VoxTurn, Tyger, ELLSA 2, PERSIST, and more.
- **2026-10-06** — Sesame agentic voice assistants with updated CSM; Hume announces EVI/TTS API sunset (access ends 2026-11-13).
- **2026-10-05** — Amazon **Nova 2.5 Sonic** GA (speech-to-speech realtime).
- **2026-09-10** — OpenAI **GPT-Live** API GA; **gpt-realtime** / **gpt-realtime-mini** announced for deprecation.
- **2026-09** — SeedRealtime, Grok Voice Think Fast, Hibiki-Zero, MiniCPM-o 4.5, PersonaPlex, and many duplex / dual-system papers.
---
## 📑 Contents
- 📚 Surveys & Position Papers
 - Spoken Language / Speech LLMs
 - Full-Duplex & Turn-Taking
 - Omni, Streaming & Proactive Agents
 - Position & Perspective Papers
 - Influential Blogs & Announcements
 - Tutorials & Primers
 - Related Awesome Lists
- 🎙️ 1. Turn-based Speech & Omni LLMs
 - Early Speech–Language Models
 - End-to-End Spoken Dialogue Models
 - Omni Models with Speech Generation
- 🔄 2. Full-Duplex Interaction Models
 - 2.1 Engineered Sync (Controller / Cascade)
 - 2.2 Learned Sync: Multi-Stream / Dual-Channel
 - 2.3 Learned Sync: Time-Division / Chunk / State Tokens
- 🧠 3. Brain–Cerebellum Dual Systems
 - 3.1 Concepts & Frameworks
 - 3.2 Open Dual-System Models
 - 3.3 Fast–Slow Cascades & Speculative Handoff
 - 3.4 Closed Dual-System Products
- 💭 4. Think-while-Listening / Speaking
- 🛠️ 5. Agentic Interactive Models & Harness
 - 5.1 Native Tool / RAG / Action Channels
 - 5.2 Interaction Harness & Runtimes
 - 5.3 Voice Agent Frameworks & SDKs
 - 5.4 Memory for Interactive Agents
- 👁️ 6. Streaming Omni & Proactive Vision
 - 6.1 Streaming / Duplex Omni Assistants
 - 6.2 When-to-Speak Video LLMs
- 🤖 7. Embodied Interaction
 - 7.1 Full-Duplex Embodied Models
 - 7.2 Dual-System VLAs with Interjections
- 🔬 8. Cross-Cutting Technical Themes
 - 8.1 Turn-Taking, Backchannel & Interruption
 - 8.2 Interaction Alignment & RL
 - 8.3 Expressive, Empathetic & Persona Control
 - 8.4 Simultaneous Speech Translation
 - 8.5 Multi-Party Interaction
 - 8.6 Low-Latency Decoding & Systems
- 🏢 Ind