Ateme launches the X2000 encoder to compete with AWS. The product supports AV1 and HEVC codecs.
The real-time AI avatar category is fragmenting along architecture and scope lines: complete platforms (D-ID, HeyGen), specialized real-time specialists (Anam), character-driven enterprise plays (Soul Machines), and modular rendering layers (Simli). The key competitive gap is that Tavus remains developer-first with an API-centric approach, while D-ID and HeyGen are pushing toward business-user-accessible workflows spanning both live and scripted content. D-ID's V4 technology, with sentiment-aligned delivery and sub-0.5-second latency, represents the company's bet that multi-format support—not raw rendering quality—will determine enterprise adoption. Watch whether Tavus adds visual creation tools for non-technical teams or doubles down on its API-first developer moat.
D-ID launched V4 Expressive Visual Agents on March 16, 2026, positioning them as a unified platform for both real-time LLM-connected conversations and scripted long-form enterprise video. The launch announcement confirmed availability to 1,500 enterprise customers and millions of subscribers, with more than 800,000 visual agents and 300 million non-interactive avatars created on previous D-ID models. The platform uses a diffusion-based model trained on performances captured from real actors, delivering sub-0.5-second conversational latency and up to 4K resolution. D-ID also acquired simpleshow, a Berlin-based AI explainer video company, in September 2025, expanding its enterprise content capabilities across more than 70 countries.
D-ID's own internal benchmarking places its V4 technology first in lip-sync accuracy among competing real-time avatar platforms. The company's published tech specs show D-ID scoring 7.16 LSE-D compared to Anam at 7.84, HeyGen at 7.93, and Tavus at 8.42, claiming a 10 to 15 percent performance advantage across SyncNet metrics. However, these are self-published figures from D-ID's own testing methodology, not independent third-party validation. The comparison methodology used a fully synthetic, reproducible test across real-time avatar streaming from all four platforms.
The V4 Expressive Avatars first became available in D-ID Studio on February 4, 2026, before the full agent launch in March. The Studio integration allowed users to select specific sentiments for avatar delivery and generate performances for marketing, training, and product content. The real-time visual agent capability followed shortly after, combining the expressive avatar layer with an LLM brain, WebRTC streaming via the D-ID Client SDK, and optional camera-based sentiment awareness that feeds nonverbal cues into both the LLM response and the avatar's delivery.
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source