THE COLLECTION

EVERY SIGNAL, SORTED.

Search the whole shelf by topic, tool, term, or plain old curiosity.

SHOWING 2 OF 619 STORIES
ISSUE 666MORNING

China, Meituan LongCat, LongCat-Video-Avatar 1.6, Open-Source Video, Audio-Driven Avatars, Long Video Generation, Lip Synchronization, Identity Consistency, Human Evaluation, Consent, Creator Tools, GPU Inference and Media Provenance

LongCat made ten-minute AI avatars. Track the drift before the face slips

Meituan released an open audio-driven avatar checkpoint that it says can generate videos longer than ten minutes with less accumulated error. The longer the clip, the more chances identity, timing, motion, rights and compute have to wander.

#video#open-source#evaluation
ISSUE 657MORNING

China, Singapore, ByteDance, BytePlus, ModelArk, Dreamina Seedance 2.5 Flash, Generative Video, Video Editing, Text to Video, Image to Video, Multimodal References, API, Production Testing, Acceptance Benchmarks, Provenance, Rights and Responsible AI

Seedance 2.5 Flash opened its API. Benchmark the edit before you price the speed

BytePlus has made Seedance 2.5 Flash available through ModelArk for video generation, editing and extension. The fast model can produce clips up to 30 seconds and combine images, video and audio references. That is enough flexibility to make a handsome demo and enough complexity to make a production test absolutely necessary.

#video#api#evaluation