If we wanted a quiet end to 2025 in the AI creative space, it has not happened.

Major drops have appeared:

Kling O1
Kling 2.6 Video
Seedream 4.5
Google Nano Banana Pro
Runway 4.5
LTX-2
Flux 2

And there are probably still more to come before the ball drops on December 31st 2025.

Here's a run down of a couple of these.

Kling Goes BIG

Big is an understatement. They launched an entirely new video model with video EDITING abilities.

Yes, we can modify images with a few text prompts, but imagine this with an animated AI video.

From removing objects, adding new ones, changing camera perspectives...Kling O1 gives you creative control over all your existing creations, and the power to make more.

Kling O1

Kling O1 is an all-in-one editing suite powered by AI where you can upload images and video to make modifications.

Whether you use text prompts or use images to power the changes, this is all possible now in the Kling interface.

This video gives a comprehensive rundown of how O1 (or Kling Omni) works, and how you can get started.

Kling 2.6

Kling 2.6 is their upgraded video model, but there are some key new features that take this to the next level.

Hyper-realism, faster generation times.
High quality outputs.
Native Audio and SFX

Yes, your videos now contain audio and SFX natively built in for the output, and it includes dialogue and voices.

I cover this in my latest tutorial about Kling 2.6, but they have reached the standard (and maybe beyond) that is presented by Google Veo 3.1 in terms of quality.

Any video model released that doesn't have native audio is going to be left behind moving forward.

Seedream 4.5

The fantastic AI image model builds on everything that made it so popular in the first place.

Great prompt adherence.
Powerful image modification tools.
High-quality generations at 4K.

This combination sets Seedream at such a high standard because it combines quality with speed. We need more models that can give us this so ideation is faster rather than looking at a loading bar for minutes.

Here's a couple tests that focus on writing text, but also making image modifications where we shift the camera perspective.

LTX-2

LTX continues to impress me with how they have created an more open, accessible and impactful model that is fast to generate but also involves video modification.

In particular, their Reframe feature lets you edit the start/end of scenes but even alter dialogue whilst keeping the video consistency.

Here is my full tutorial of how this looks in practice:

Actions to Take

Stay on your toes as these models being pushed out late in the year will set the floodgates racing for the start of 2026.

We can expect to see many model updates in the coming months as competitors try to get in front of the curve.

My recommendations are to stay aware and to try to access some of these tools as soon as possible to try.

My recommendation is to use Freepik as all of these models are now being added into their existing plethora of tools.

Here's a link to give you a bit of a discount when registering for Freepik.

AI Art and Video Updates in December 2025