DeepSeek V4.1 Flash 1017: a fresh snapshot is live
DeepSeek V4.1 Flash 1017 is now available in the chat, the Code section and the public API: an October 17 snapshot of the fast, affordable model with tool calling and web search. How to try it.
Vidu Q4: cinematic video with synced sound from a single photo
A new video model in NeuralSpace: one photo or a set of references becomes a cinematic scene with dialogue and synced sound, up to 4K and 16 seconds.
$0.02 per second of video in MiniMax H3
Generate cinematic AI video in MiniMax H3 for just $0.02 per second. Custom HeyGen fine-tuning, realistic motion physics, and photo, video, and audio reference support until November 1, 2026.
Whisper API and MCP: how to add speech recognition to your project
Whisper large-v3: capabilities, the /v1/audio/transcriptions endpoint, MCP tools and getting a key.
AI for Video: Text-to-Video, Photo Animation, and AI Dubbing
How modern AI video models work: generating clips from text prompts, animating still photos, and multilingual dubbing with lip-sync. Model breakdown for SeeDance, Kling, and Grok, real latency benchmarks, and a step-by-step guide on NeuralSpace.