Gemini 3.8 Live adds real-time video avatars, now GA on the Live API

Google's speech-to-speech model now renders a lip-synced face in 97 languages; OpenRouter halves batch token prices and GPT-6 lands in Codex CLI.

Nowline SEP 26 7:00 AM banner

Top AI stories from the last hour

Top AI stories from the last hour

Copy markdown

  • A lip-synced face on your voice agent

    Gemini 3.8 Live is now generally available with Live Avatar — native speech-to-speech that also renders a talking, lip-synced video avatar, recovers from interruptions without losing context, calls tools mid-conversation, and handles 97 languages with live camera and screen input. Build it through the Gemini Live API and the open-source ADK using the curated avatar library; custom avatars stay allowlist-only for now, and every output carries a SynthID watermark.

  • OpenRouter batches your calls at half price

    The OpenRouter Batch API is live: POST a list of requests to /api/v1/batches and pay about 50% less per token across 70+ models, with results back inside 24 hours (median 7 minutes during the beta). It covers chat, responses, messages, and embeddings — built for evals, backfills, and bulk generation you don't need answered this second.

  • GPT-6 Sol and Luna reach the Codex picker

    Codex CLI 0.156.1 adds OpenAI's GPT-6 Sol and GPT-6 Luna to the /model picker, with a hint nudging heavy users toward Luna for rate limits. Once you update, you can drive your terminal agent on the GPT-6 frontier models directly instead of aliasing them by hand.