Model drops, day by day.

V4.1 Flash updated to version 1017, now behind the deepseek-flash API name, with weights on Hugging Face.

5.6B open-weights image model for text-to-image, editing and multi-reference with up to 10 images, native 4K.

For film: Runs locally on consumer GPUs, and bounding boxes place elements where you ask.

Upload a video and get an original score plus sound effects placed on the timeline, in ElevenCreative and the API.

For film: Effects come as separate layers you can retime; the score can follow existing audio.

Krea's new image editing model: most edits under 8 seconds at native 2K, up to 6 reference images per edit.

For film: Keeps the same character across poses, locations and light for a consistent set.

Kling 4.0 fully live on the Kling site: native 30-second takes, Omni Reference with up to 15 references and 10 keyframes.

For film: One-pass long takes and keyframe control for continuous scenes; API access is promised soon.

Open-weight multimodal model, 816B total and 42B active, 1M-token context, built for coding and agentic work.

Voice cloning that turns a recording, a script and a target image into speech shaped for the scene on screen.

For film: Delivery and room acoustics match the shot, from studio to street to a moving car.

Video-to-sound model that replaces a clip's audio with synced effects, foley and ambience while keeping dialogue.

For film: Keeps dialogue while adding sound; Studio and API take clips up to 10 minutes.

Fish Audio's most emotive text-to-speech model, live on fish.audio, the Fish Audio app and the API.

For film: Follows direction: a whisper stays a whisper and a laugh lands where you put it.

Google's frontier model now generally available in the Gemini API, AI Studio and the Gemini app for Pro and Ultra.

Anthropic's fastest small model yet, around 75% cheaper to run than Haiku 4.5 on average.

Preview of the next Vidu model: image and reference-to-video with native audio, 3 to 16 second clips, 2K/4K out.

For film: Up to 15 image and 3 voice references, with multi-shot camera moves in one take.