Rodrigo

AI/3D artist pushing art & tech boundaries using always open for collaboration or just a friendly chat.

[Rodrigo @rodrigon]

Recap I made of some of my work in 2023.
All the images in this video were generated using @Scenario_gg.

II'm very happy to have the opportunity to explore so many workflows using generative AI.

That was only 2023... 2024 promises much more!
@hervenivon @emmanuel_2m @araminta_k

[A.I.Warper @AIWarper]

Pretty dope.

I really need to test Video conditioning with SD 2.

[Kaku Drop 架空飴 @KakuDrop]

Seedance 2.0 のOmni referenceを使っております

１発で狙った生成をしてくれました😲

[Todd Grilliot @GrilliotTodd]

Pixel Engine v1.1 (left) vs Nano Banana 2 (right). I gave both models basically the same prompt.

Unless you've tried to create spritesheets with image models before, you probably don't realize how hard it is to get good motion. Image models don't want to animate things, they want to make images.

I trained my own animation model (Pixel Engine 1.1) to solve this problem. It's a completely novel approach, and it works surprisingly well. You can try it for free rn in the pixel engine app.

[ててつろう @craftcapitallab]

メイキング的なやつです。

今回は、基本的なアニメーションの作り方である
『pose to pose』をAIでやったらどうなるか？
という試みでした🥹

結論から先に言うと
「まぁまぁ大変だし、AI時代のスピード感にはそぐわないだろうな」です。

ある程度、予想もしてましたが🤣

今回は、コンテ、ラフ原を描いて(あるいは3Dで)、絵や動きを「ある程度」コントロールしたい人向けのフローなのかなと思います。
そういった方で、AIにまだ馴染みが無い方向けに、今度、もう少し詳細な、記事かnote書こうと思います。

----------------------------------------------
※pose to poseというは、
動きのキーとなるポーズ、主に動きの変化点などの絵を複数枚、先に用意して、後から、間の絵をうめていってアニメーションにする手法です。

キーとなるポーズや絵を、「キーフレーム」とか「原画」。間を埋めていく絵や作業などを「中割り」「動画」と言います。

----------------------------------------------
以下は、AI界隈の方向け、メイキングの簡易的な説明です。

端的にいうと、
先にラフ絵で動きのプランをつくり、
その絵を使って、キーフレームの絵を生成して、
あとはひたすらスタートエンドで生成。

です🥹

---------------
①コンテ撮をつくる
いわゆるアニメの原画よりは枚数少ないですが、
ラフな絵で、動きのキーとなる一連の絵を用意して、先に動きの流れをつくります。

---------------
②キーフレームの作成
コンテ絵と、事前につくったキャラ絵をリファレンスにして、NanoBananaPro等で、各キーフレームの絵を生成＆加工します。

私の拙いラフ絵なので、結構ガチャりましたが、もう少し丁寧に描いた、アニメーターさんのラフ原とかだと、もう少し絵のコントロールが、できるんじゃないかなと思います。

タッチ線を入れたり、変身後の衣装に変えたりなどは、さらに２段階、３段階の生成が必要でした。

---------------
③中割りの作成
キーフレームを順番に並べて、それぞれの間を、スタートエンドで、動画生成します。

今回使用したのは、
@DomoAI_ / @Kling_ai / @Hailuo_AI
の３つです。

主力はDomoAIで、3秒尺の生成ができるのと、割と単純な中割りになるのが良かったです。

逆に単純に割ってほしくない場合(例えば回転していて別のアングルが途中で入るとか)は、KlingやHailuoを使いました。

---------------
④フレームの選定(コマ抜き)
生成された各動画をつないで、元々つくっていたコンテ撮のタイミングに合わせて、大胆に絵を抜いてきます。

極端なものだと、3秒(約72枚)生成して、その間の１枚しか使わない。みたいな感じですｗ

あとはいわゆるタメツメの調整など

---------------
⑤撮影＆編集
別途用意した、イメージ流背のBGやエフェクト素材などと合わせてコンポジット。処理かけて、SEとBGM入れて完成です。

---------------

つまり、①と④の工程で、思いっきり、アニメーター的技能を使うことで、動きのキモはAIではなく、自分でやるということですね。

◆メリットとしては
- 動きや演技をあるていど自分でコントロールできる
- ある程度積み上げながらの作業なので、部分的な修正が可能
- 溶けにくい
など

◆一方デメリットは
- 他のAIの手法に比べて、相当時間がかかる
- クレジット消費も激しい
- 色やキャラが、１カット内で安定しない
（髪の毛の本数がかわったりなど、カットまたぐ分にはきにならないものでも、動きの途中で変わっちゃうので、結構目立つ）
- 実景の空間でのカメラワークとかが混ざると、多分さらに難易度が上がる
- コントロールに結局限界が(まだ)ある
などなど、、、、

時間と手間がかかるけど、商業レベルの「品質」まで対応できるわけではない。しかも手描きアニメーターの上手い人なら、下手したら同じくらいの時間でできちゃうかもしれない。。。
ゆえに冒頭の結論に至るわけです🤣

-----------------------------------------

[Danny Limanseta @DannyLimanseta]

As a hobbyist artist, I’ve found this workflow works really well.

Define the art style with your own hand-drawn art, then feed that into an AI prompt (I use Google Nano Banana Pro) to generate the remaining assets.

With this simple workflow, I’ve created hundreds of art assets, all consistent with my initial hand-drawn style.

[Rodrigo @rodrigon]

Another workflow I created using @Scenario_gg nodes. This workflow is a complete suite for character creation. It allows you to generate characters, edit them, create variations and even 3D models, and much more.

[Emm | scenario.com @emmanuel_2m]

Check this out, this is gold 🏅

(full resolution image below)

Full end-to-end character creation workflow on @Scenario_gg Workflows - brainstorming to final output - all in one.

Input a style, character description, and ref images. From there, everything runs automatically...

[Rodrigo @rodrigon]

I'm really enjoying creating workflows in @Scenario_gg . Having control over creations by visualizing the entire process is much better than generating them individually.

[Emm | scenario.com @emmanuel_2m]

This was made in one. single. click.

- No install
- No dependencies
- No broken nodes or library issues.
- 100% flexible (image, video, audio... and more)

We're so back.

[apolinario (poli) @multimodalart]

I've built a custom camera control @gradio component for camera control LoRAs for image models 🎥🧊

Here's a demo of @fal's Qwen-Image-Edit-2511-Multiple-Angles-LoRA using the interactive camera component

[Tencent Hy @TencentHunyuan]

✨We are excited to open-source Tencent HY-Motion 1.0, a billion-parameter text-to-motion model built on the Diffusion Transformer (DiT) architecture and flow matching. Tencent HY-Motion 1.0 empowers developers and individual creators alike by transforming natural language into high-fidelity, fluid, and diverse 3D character animations, delivering exceptional instruction-following capabilities across a broad range of categories. The generated 3D animation assets can be seamlessly integrated into typical 3D animation pipelines.🎮🎥

Highlights:
🔹Billion-Scale DiT: Successfully scaled flow-matching DiT to 1B+ parameters, setting a new ceiling for instruction-following capability and generated motion quality.
🔹Full-Stage Training Strategy: The industry’s first motion generation model featuring a complete Pre-training → SFT → RL loop to optimize physical plausibility and semantic accuracy.
🔹Comprehensive Category Coverage: Features 200+ motion categories across 6 major classes—the most comprehensive in the industry, curated via a meticulous data pipeline.

🌐Project Page: https://t.co/IIGGQj25Pg
🔗Github: https://t.co/L4642SrQoW
🤗Hugging Face: https://t.co/Gmgv0O0CKl
📄Technical report: https://t.co/nq0nSgeR6I

[Andy Smith Art @andysmithart]

On the left is my artistic rendition of Cable and on the right is my utilization of Gemini AI to render it in a realistic style. I think it’s highly impressive and I have no reservations about leveraging AI as a tool to play with my art.

[Valephore 🌙♦️ @ValephoreVT]

Concept Artist here, the "AI hate" directed at Larian Studio is 100% ignorant and bad faith virtue-signaling.

Artists are NOT against AI usage as a tool, we're against Generative AI that was trained on real artworks and mass-produced "slops" that imitated our hard work and then marketed as our replacement.

In pre-production, the art team would grab random pics off the internet and other games to create a general mood board for aesthetics and "vibe" that fit the art direction (There are more nuances than this, I'm describing this in the most simplified way possible)

A mood board of a character may include several pictures of other videogames or anime shows, dozens of object with "texture references" for parts of their attire, multiple reference pictures of relevant objects such as animals or historical weapons or even other character artwork made for other games; ALL of which are randomly drawn from the internet and put together for the purpose of creating an overall aesthetic for a design that will become original in its own way inspired by its references and the artist's own creativity and originality to visualize it when they create the design by doing ART.

Larian Studio mentioned using AI to streamline this process; an actively menial labor of "looking for relavant and cool pics on the internet" in essence.

There is zero reason why AI speeding up this process is a bad thing; it is essentially the same as someone asking Grok for facts they could have spent a few hours reading through Wikipedia pages.

None of the pictures grabbed by either the AI or the art team in this process are being exploited for any monetary gain since they're simply inspiration and visual direction, this has been standard practice in the industry for decades.

[Morph @Morph_VGart]

I'm a man of my word, since you guys really needed it spelled out this way:

@IsThisA3DModel called us out so now we have to add the following disclaimer to this post: This is not a 3D model or gameplay footage in any regard. It is merely a dumb AI generated video clip. This post was sponsored by SocialKettle.

Would you play a game like this?

[Swen Vincke @where? @LarAtLarian]

Holy fuck guys we’re not "pushing hard" for or replacing concept artists with AI.

We have a team of 72 artists of which 23 are concept artists and we are hiring more. The art they create is original and I’m very proud of what they do.

I was asked explicitly about concept art and our use of Gen AI. I answered that we use it to explore things. I didn’t say we use it to develop concept art. The artists do that. And they are indeed world class artists.

We use AI tools to explore references, just like we use google and art books. At the very early ideation stages we use it as a rough outline for composition which we replace with original concept art. There is no comparison.

I talked about how we use ML here if you would like to know more:
https://t.co/41KPNjaqI7

We've hired creatives for their talent, not for their ability to do what a machine suggests, but they can experiment with these tools to make their lives easier.

[A.I.Warper @AIWarper]

539 frame SCAIL test in ComfyUI

704 x 1280 resolution

Prompt executed in 01:25:27
Max allocated memory: 23.214 GB

[DailyPapers @HuggingPapers]

Huawei's MoCapAnything: Unified 3D Motion Capture for Any Skeleton

Imagine animating anything from a monocular video! This new framework from Huawei tackles Category-Agnostic Motion Capture, producing high-quality skeletal animations & cross-species motion retargeting across diverse rigs.

[Scenario @Scenario_gg]

Meet the new way to explore Scenario: 99+ one-click Apps that transform your images instantly.

No setup. No workflow building. Just drop something in… and see the result in seconds. ✨

Try them out: https://t.co/iHZTYYbU8U

[Rodrigo @rodrigon]

@RobMullenders @3vilbasterd @NHPUKOfficial I agree with you, but maybe it's because he lives in London? And thieves like these, when caught, are beaten up in Rio de Janeiro and the rest of Brazil.
Not all thieves here are heavily armed gang members.

[CyberRobo @CyberRobooo]

Heavy humanoid robot worker

Japan’s Tsubame Industries built ARCHAX:a 4.5 m piloted humanoid robot that’s designed for real-world heavy-duty work, not just Gundam cosplay.

You climb into the cockpit, get a full 360° view through nine cameras, and feel the exact grip force through force-feedback controls. With 26 degrees of freedom, it can lift heavy loads, tear down structures, or delicately search rubble in disaster zones.

It works remotely too,no one has to be on site. Battery lasts all day, air-conditioned cabin, and it can drive itself to the job in vehicle mode at 10 km/h. When the task is done, disassemble it and haul it away on a truck.

What is it used for? disaster search & rescue, nuclear decommissioning, large-scale demolition, and ground testing for future lunar base construction.

ARCHAX isn’t just a toy. It’s the first serious attempt at a human-scale robot that can replace people in the most dangerous environments.

[vijay verma @realvjy]

I create most of visual assets in Figma, but exporting everything is always tricky. So, made this little plugin to exports all parts sepratly with JSON, which can use to animate easily.

[Radiance Fields @RadianceFields]

🚨Google and UCSD just introduced Radiance Meshes, a new radiance field representation that produces watertight meshes and renders faster than 3DGS. Code and demos are available now.

Code: https://t.co/8qpQIJYo87
Demos: https://t.co/yvqh1Mhe82