August 1, 2025
KREA AI: Black Forest Labs dropped FLUX.1 Krea [dev], a new open-weights FLUX model made with help from Krea AI. It's built to make photorealistic images with a unique style. No fake AI look, no blown-out highlights, just clear natural detail.
July 31, 2025
Producer.ai (Riffusion): Riffusion is now Producer.ai 👀 It's a personal AI music tool that turns ideas into sound.
It's invite-only for now. If you already used Riffusion, you're in and you get invites to hand out.
First 1000 people who use code STUDIO get access today.
July 31, 2025
Wan (Open-source): WanGP 7.5 now works with Wan 2.2. Most things from Wan 2.1 still work fine like low VRAM use, extra long videos, image making, and MMaudio. It needs more RAM now since the model’s twice as big but still uses the same VRAM.
July 30, 2025
Replicate: Replicate and Pruna AI made Wan 2.2 faster and cheaper. Now: 33 seconds, $0.06 per video (14b model, 480p).
July 30, 2025
Wan (Open-source): ComfyUI dropped a new patch for Wan 2.2. It cuts VAE decoding memory use by about 10%. The 5B I2V model got a big boost too. They also added new setup templates for the 14B models. You can grab the latest version on Git, Portable, or Desktop.
July 30, 2025
Luma Dream Machine: Luma added a new way to change stuff in Dream Machine using plain language - Modify with Instructions. You just type what you want and it makes edits across VFX, ads, movies, and design projects. Stuff like removing or swapping objects, making virtual sets, fixing characters, and restyling is coming soon for all subscribers.
July 29, 2025
Freepik: Freepik adds Veo 2 to its unlimited generations - for 2 days - for Premium+ and Pro tiers.
July 29, 2025
Wan (Open-source): ComfyUI adds native Wan2.2 support.
Available in FP16 & FP8, with 5B models running on 8GB VRAM via auto-offloading.
July 29, 2025
Runway: Runway has unveiled its new Aleph AI model for video editing, offering features like object removal, camera‑angle generation, style & lighting changes, motion transfer, and even aging actors - all via prompts. Early access begins for Enterprise and Creative Partners, with wider access to come soon.
July 28, 2025
Wan (Open-source): Wan 2.2 is out now! Featuring a faster model, open-source access, and upgraded tools from Tongyi Lab at Alibaba Group. Improved prompt following, motion control, and visual detailing make it ideal for cinematic AI creation.
