What's new?
New Kling video and image models: You can now generate native 4K video (3840x2160) using three new Kling models across text-to-video, image-to-video, and reference-to-video, as well as new Kling image models with element reference and frontal image support, giving you more precision and higher output quality across both formats.
30+ new audio and LLM models: You can now generate music, sound effects, and voiceover using a wide range of new audio models including Stable Audio 2.5, MiniMax Music, the ElevenLabs suite, Lyria, Gemini TTS, Qwen 3 TTS, Kling Video-to-Audio, CassetteAI, and Mirelo SFX, so your full creative pipeline from visuals to audio stays in one place. New LLMs including Claude Opus 4.7, GPT-5.4/5.5, Grok 4.20, and DeepSeek V4 are also available.
Bug fixes & improvements
Faster and more reliable MP4 exports: Fixed a flaky download button issue; MP4 metadata embedding now bypasses re-encoding entirely, making exports faster and more consistent.
Reference-to-video pricing fix: Adding a reference image no longer silently collapses output dimensions or drops the CU estimate, so what you see before generating is what you get.
Audio model duration pricing fix: A missing capability flag on some audio models was causing output duration to be priced incorrectly. This has been resolved.
Reference-set model auto-pick fix: Modern image models including Seedream, FLUX Kontext, Imagen, GPT Image, and Qwen were being silently excluded from auto-selection, with Kling O3 winning by default. All models are now considered correctly.
A note from the team
You may have noticed it's been a while since our last release article. That's intentional. We've been heads-down on something we think is going to meaningfully change how you use Layer every day: faster to get to what you need, more intuitive to navigate, and frankly a lot more enjoyable to work in. We're not ready to share details just yet, but if you've ever had a moment of friction in the product and thought "this could be better," chances are we've been thinking the same thing.
Stay tuned. It's coming soon.