Catnip

Open the gate to social world model.

[Catnip @catnips_ai]

🥇MaineCoon: From Passive Video to Real-Time AI Presence

The first unlimited-duration interactive audio-visual model.

Most AI products today still feel like they live behind a screen.

You type. It answers.
You speak. It replies.

The interaction is still mostly turn-based.

Mainecoon is built around a different idea: AI should not just respond to you. It should feel present with you.

🔗Learn more
Website ↓
https://t.co/SFpsMsvLs9
Blog ↓
https://t.co/nkc1KT7bT5

[Catnip @catnips_ai]

Love seeing users sharing their MaineCoon experiences on X 🙌🙌

[Madd Fictional @Madd_Fictional]

Taking Mainecoon tech (@catnips_ai) out for a test run.

[Catnip @catnips_ai]

this doesn’t feel like video generation anymore… it feels like someone is actually there.🔥

[Ebonymanley_25 @EBONYMANLEY_22]

Created this with MaineCoon — real-time audio-visual character generation

Great experience using it great and fast character generation.

@catnips_ai

#AIArtWork #AIArt #AIArtists #AIArtCommunity

[Catnip @catnips_ai]

@TiffanyFong learned

[Catnip @catnips_ai]

@EHuanglu black mirror, but the cat is somehow the most emotionally stable one in the room

[Catnip @catnips_ai]

Breaking news from the blocky universe.
MaineCoon turns a simple character into a real-time AI news anchor.😃😃

[Catnip @catnips_ai]

Appreciate this.
MaineCoon is built for real-time interaction — not just generating clips, but keeping expression, voice, motion, and context in sync.

[François Chollet @fchollet]

MaineCoon is the first video model that focuses on social interactions: facial expressions, emotions, fluid conversation, audio-lip sync, etc. Really impressive inference specs: 22B params, 47.5 FPS on a single H100. Generates in real-time at <$0.001/sec.

They achieve this with an agentic streaming inference framework with 3 different auxiliary models to manage the cache and lookahead buffer. Super cool work.

[Catnip @catnips_ai]

This is the key shift:
From prompt → wait → output
to real-time interaction.
MaineCoon generates audio and video together, keeps perceiving while it responds, and runs fast enough that the experience feels live

[Ole Lehmann @itsolelehmann]

you've only ever known generative AI (you prompt, it generates text/media, you wait, it stops)

but this is the first thing that made the next phase click for me:

interactive AI.

a real-time audio-visual model that perceives you while it responds.

it's super interesting how it works:

most AI makes the video first, then dubs the sound on top. this one does both at the same time. so the voice and the mouth actually match.

no prompting, rendering, or waiting. it reads your face, your voice, your timing, and reacts as you go.

it's a 22-billion-parameter model, but it still runs on a single GPU at up to ~47 frames per second, faster than the ~25 fps you actually watch at.

that's why there's no loading bar.

this opens up a whole category we've never really had, like:

> AI livestreams you can talk to.
> video calls with a character on the other end
> live hosts and tutors that respond while you're still mid-sentence.

it's early and still a little rough, but it's the first time using AI felt "live" instead of loading.

[Catnip @catnips_ai]

@TiffanyFong lol lovely dinner with cat🐈 thanks to MaineCoon

[Catnip @catnips_ai]

Most AI video today is still:
prompt → wait → watch a clip.

MaineCoon is built for something different:
prompt → talk → interact in real time.

In our vision, the character is not a fixed video clip that just waits for your input. It keeps generating voice, expression, and motion on its own.

That is why AI video starts feeling less like content — and more like someone you can actually hang out with.

To meet our goal, the first step is Mainecoon, a real-time interactive audio-visual model built for streaming generation to interact with you.

1⃣Up to 47.5 FPS on a single H100 GPU
2⃣Audio-visual generation cost below $0.001 / second
3⃣Long-duration streaming generation for 1000s+ seconds
4⃣Continuous audio, motion, expression, and visual alignment
5⃣SOTA performance on SocialVideo Bench
From passive video to real-time AI presence.

Want to try MaineCoon?
Learn more and apply for early access: https://t.co/SFpsMswjhH

Share a great MaineCoon video on X and @catnips_ai , get 2 extra codes.

[Catnip @catnips_ai]

@madoka_saikou yeah it's gonna be fun in games

[Catnip @catnips_ai]

@CruiseAtuekong @TiffanyFong try talk to cats on Maincoon😺

[Catnip @catnips_ai]

@mobailabs @bozhou_ai yeah，it's hard, but we can make it. tech report release soon

[Catnip @catnips_ai]

@sunway_chan @bozhou_ai yeah, thanks you like that

[Catnip @catnips_ai]

@berryxia @bozhou_ai yeah, it's MaineCoon. follow me and apply the code

[Catnip @catnips_ai]

@HarriesSteele @bozhou_ai It's MaineCoon.

[Catnip @catnips_ai]

We made an AI-generated Trump speech with MaineCoon — real-time audio-visual character generation, built for characters that can talk, react, and keep going. @realDonaldTrump @Trump your digital campaign intern is ready.😀

[Catnip @catnips_ai]

amazing！👏👏 came for the AI girlfriend joke, stayed for the real-time interaction

[泊舟 @bozhou_ai]

卧槽，刚刚体验了1分多钟AI实时女友，没想到AI视频已经发展到这种地步了。

都已经可以直接实时进行互动，有点惊到我了，这个体验不太一样，它是可以实时互动的。

你输入一句话，角色会继续生成表情、声音、动作和画面，不是播完一段固定视频就结束。

看看我生成的御姐女友，哈哈哈，我已经想到了未来AI女友是什么样子了

[Catnip @catnips_ai]

AI went from “make me a video” to “come sit with me” way too fast😆😆 MaineCoon will make this happen

[Tiffany Fong @TiffanyFong]

Holy shit we’ve spent the last few years making AI content.

The next phase is real-time AI you can actually hang out with. 👀

[Catnip @catnips_ai]

@svpino this is the fun part. when generation gets fast enough, video starts becoming the interface.

[Catnip @catnips_ai]

@inverted_Legion lmao 🤣🤣finally, AI built for the real decision makers of the internet: cats