


Your team’s AI, finally in one place.


Lovely to see Rabbit coming to life 🐰 The launch creative was made with Endless and we love how it turned out. Huge congrats to the whole team! 🧡
Excited to introduce Rabbit: the creative space for kids 🐰🧡 @heyrabbitapp is a 3D world where kids create their own games, characters, stories and music with AI. With @juliandeangeIis and @0xKoller, we're on a mission to turn passive consumers into active creators. We don't want kids to spend more time on screens. We want them to do more with the time they already spend there. Waitlist in the reply

señoras y señores en breve se viene testflight

@franciscoparata turned his cat Ponyo into Enzo Fernández, and we all wanted to do the same thing. So we decided to see how four video models would handle the exact same edit.
Endless @makingendless Oct 30 Article Best AI model for video editing: Seedance 2.5 vs Kling 4 vs Wan 3.0 Prime vs MiniMax H3 The most expensive AI model is not the best at editing a video. We edited the same clips with the four Endless models that can swap someone inside an existing video. Seedance 2.5, the most expensive


In our article we break down how we got these results, with every prompt and the credits each run cost. And if you want to try it yourself, all four models are available on endless.app.

Our last try was Kling 4. It looks good for the most part, but things go wrong at the end🤨 The last shot feels like a completely different take, and the jersey disappears.

MiniMax H3 was the cheapest option at 118 credits, but it also changed the ending quite a bit. The hug happens too early, and the salute is gone.

Wan 3.0 Prime got almost everything right and even kept the original audio. We ran it on our canvas, and it cost 244 credits.

This is the one that started it all. @franciscoparata made it with Seedance 2.5, with Enzo's 24 jersey and the exact same salute at the end. Pretty hard to beat.

Here's the original clip we used. Pay close attention to the ending: the salute only lasts 0.3 seconds, so we made sure to mention it in every prompt to see if the models could get it right.

A mi viejo (carnicero) una agencia le quería cobrar una fortuna por manejarle las redes. Le tiré un prompt a Endless y el agente analizó su marca, investigó qué le funciona a la competencia y armó un brand-kit y un ZIP con una semana de contenido.

So, is GPT Image 2.6 better? Our take: 2.6 pulls ahead on stylized and cinematic scenes, but it's close on the rest. Run your own tests with both models on Endless: endless.app

A still from a 1950s mafia movie. Both turned the reference photo into a solid noir frame. 2.6 goes harder on contrast and drama, the look to camera hits stronger, and the face stays closer to the reference. 2.5 is softer and warmer but 2.6 nails the noir look.

Our next test: turning a photo into an authentic 1950s print. Both kept the same man and pose. 2.5 goes darker and grittier, with more scratches and dust. 2.6 is softer and more faded, like an old print that's been sitting in a drawer. Two different takes, both convincing.

This is where 2.6 pulls ahead. We made a candy glass scene, and in 2.6 the whole world feels more put together and more saturated. The vehicle, the thermos, the robot and the shop actually look like they belong in the same place. 2.6 clearly won.

A regular phone pic of a birthday party. Both got the unposed, at-home vibe right. One person looking at the cake, two looking at the camera. 2.5 feels warmer. 2.6 light is harsher and a bit uneven. No real winner here.