Forest C

Digital Artist |CPP/ Kling 😃LumaAI😀Hailuo 😃Runway😀Pika😀PixVerse😃Vidu😃Ideogram😀ImaginArt😃Pollo😀Nim😀Wan😃HeyGen😀Based Labs😀Viggle

[Voyz @voyzlab]

300 Netflix shows have already gone through generative AI - and there's a good chance you watched one and didn't notice.

Here's what's behind scenes like that, down to a single shot. The source is one photo of a soldier on a battlefield with a blue route line drawn over it. Then 22 seconds without a break: running, explosions, planes, armored vehicles, climbing a concrete wall. At he pulls himself up over the top - and the whole battlefield opens up in front of him all the way to the horizon.

The route line isn't decoration. It's a shot plan drawn before generation - where the character goes, what happens along the way, where the endpoint is. AI isn't inventing the story from scratch - it's building out an already defined path. video by animatorx83

Framework:

PATH - draw the route right on the photo before generating
BUILD - AI fills that route with events and a world
REVEAL - the scene arrives at a final point instead of just cutting off

Here's what that looks like on the inside - step by step.

How to build a route like this from scratch - I wrote about it in the article below.

The most interesting thing here @voyzlab

[Andy🤖 @andy_neon_]

HER FACE DOESN'T CHANGE. HER EYES DO. THAT'S THE HARDEST THING TO FAKE

One character. One rifle. Almost no scene movement at all

Micro-expressions break most models before a single explosion does.

Claude structured the emotional beat-by-beat sequence first - the exact moment hesitation enters, where focus drifts, when the eyes shut. Seedance 2.0 rendered it frame by frame from that timing map

6 minutes of render time. Under $2 in tokens

A performance like this would normally need an actor, a director for 3 takes, and a full crew just to catch the right blink.

No actor. No director. No reshoot

Is it possible to fall in love with AI?

[Rina @irinatoxi]

https://t.co/yvIZK7PRfk

[多肉｜AI 内容增长架构师 @leo_xiaolei]

https://t.co/pHGXKctLEU

[Leti Liu @letidotae]

Never knew I could be a water bender until Gemini Omni made it happen!

My prompt below:

[林悦己Cheer @cheerselflin]

https://t.co/EbEvJy9qFR

[林悦己Cheer @cheerselflin]

https://t.co/zHp4e9xDZj

[波妞PONYO @ponyodong]

AIGC波妞的2026上半年干货总结帖

AI视频这条路我摸了大半年——从离开Peppa Pig，一头扎进AI视觉叙事和漫画IP孵化，陆陆续续发了上千条实测、教程和踩坑记录。制作、策划孵化了近百部 AI 广告、AI 漫剧作品。

很多朋友说想找我早期的干货但翻不到，所以这次把账号里互动数据最高、最值得看的内容按主题理成了一份清单：从方法论到工具实测，从爆款案例拆解到版权避坑，你可以直接跳到自己需要的部分看，少走弯路，也少踩坑。建议收藏，需要的时候直接翻。

一、AI视频/漫剧制作方法论
1. 玩AI视频角色老是换脸？3个月才搞明白问题出在哪儿（上篇）
https://t.co/TAoFDRRMRZ
2. 为什么放弃九宫格分镜，改用六宫格故事板
https://t.co/8jEZG7W63q
3. 大半年AI漫剧踩过的坑：多人戏一换机位就崩
https://t.co/O1KFQoXEAL
4. AI漫剧最大难点是场景调度，不是画质
https://t.co/0muciT0pvH

二、AI工具实测评测
1. 视觉锚点才是小红书封面的杀手锏（附爆款封面Skill）
https://t.co/mGnlDJ0ziP
2. Lyria 3音乐模型评测，附20多个提示词模板
https://t.co/GyclVdwLW4
3. 猜图测评：ChatGPT Image 2 VS 即梦Seedream 5.0
https://t.co/8DdX1ykZrn
4. Nano Banana 2深度测评
https://t.co/nXdUqgJi8V
5. Fable 5限时回归：全球IP导演风格图谱Skill
https://t.co/7PKOFasuDl
6. SkildArt实测：数字藏品展馆+数字人换装短片
https://t.co/7LvNWSzWc5
7. ChatCut插件：一句提示词用Codex剪视频
https://t.co/5CecUn6mdO

三、爆款现象级内容拆解
1. 【喜剧爆款视频教程】
https://t.co/4UKJzFRvnC
2. 如何用SeedDance 2.0拍摄闺蜜私房视频
https://t.co/3d7rP25pTf

四、版权与合规干货
1. 别等被甲方追责，才想起AI作品也有版权风险
https://t.co/1cKpo0XlBG
2. 首个「AI作品版权自查Skill」上线
https://t.co/RIi1XjDmKe

五、IP设计与创意系列
1. 卡牌设计第四弹：七龙珠概念海报
https://t.co/zD3zlrDPtj
2. IP设计《银月狐灵》视觉开发全流程
https://t.co/n2NFs1zPHF
3. AIGC新作《天���记2070》机甲美学
https://t.co/BS4sP0b5wo

六、个人故事/账号成长
1. 从离开Peppa Pig团队到成为AI创作者
https://t.co/Mel37Uj9uK

[1p(1mpal) @impalementd]

리센느 프리티걸 챌린지 커버하는 방법

등장인물이 많아? 감정선이 필요해?

준비물
1. 원하는 틱톡 챌린지 다운
2. 커버 영상의 Depth 영상을 추출한다(클로드 해줘!)
3. 레퍼런스 비디오에 넣고 시댄스로 제작

제작할 때 사용한 프롬프트도 아래에 남길게!
> Made with @get_artcraft

[Forest C @ForestChenfeng]

@markh_app 好人做到底～把几张参考图发进来吧谢谢～

[探路AI @TanLuAI]

尝试用ChatCut 将我之前做的AI视频画面，给整理成了一个古装/仙侠运镜教学。

全程我只用文字与Codex交流，最后在它的编辑器进行了简单的微调，生成出来的效果还挺好的。

具体流程：
1.打开Codex，输入这段提示词：“阅读 https://t.co/smvcgn93b9，帮我安装 ChatCut 插件并创建一个新任务。”
2.等待插件安装完毕，将你的素材 ���名，然后发给Codex。
3.输入你的需求，我发的需求是：“做一个古装/仙侠影视剧的运镜教学合集 加字幕和旁白讲解，并加一个背景音乐，以及一些半透明的动画效果”
4.生成完毕后，可以先预览。
5.通过提示词修改大问题：“第一遍生成的画面，没有做开头，部分画面描述有点问题，我直接把问题和优化需求告诉了它：需要一个开头的旁白 介绍，可以取几个画面，做个开场动画和旁白。 背景音乐的声音加大，原来素材的声音也要保留。移镜的文案修改为 纵向移动 带出环境”
6.第二遍生成基本没问题，一些字幕和音频衔接有点小瑕疵，我手动调了下 就搞定了

流程非常简单，后续我会把全操作过程整理为一篇图文并茂的短文

[TSUBAKI @AI__TSUBAKI]

Seedance 2.0 on @Hailuo_AI

=====================

Prompt:

Style: Photorealistic behind-the-scenes smartphone vlog footage. Faux vertical smartphone footage naturally cropped into 16:9. Authentic handheld phone aesthetic. No cinematic color grading, no HDR look, no 3D render, no game engine, no beauty filters. Feels like genuine behind-the-scenes content uploaded by a Korean idol.

Lighting: Bright summer afternoon sunlight through large apartment windows gradually transitions into warm evening apartment lighting. Natural exposure fluctuations and auto white balance typical of smartphone cameras.

Color: Neutral consumer smartphone color science with slightly soft contrast, mild highlight clipping near windows and natural indoor color shifts.

Camera: Simulated smartphone camera filmed by another female idol member. Gentle handheld movement, natural walking motion, subtle autofocus breathing, slight rolling shutter during quick pans, occasional imperfect framing, brief focus hunting and realistic phone compression. Camera behaves like a friend casually filming, never professionally stabilized.

Skin: Extremely realistic skin texture with visible pores and natural imperfections. Light everyday makeup. No skin smoothing.

Acting: Gentle, elegant and soft-spoken. She smiles modestly, speaks calmly, occasionally becomes slightly shy after noticing the camera, and naturally thanks the person filming. Her personality is graceful, feminine, polite and slightly reserved. She rarely laughs loudly, instead showing warm smiles, subtle amusement and delicate facial expressions. Every movement feels refined, relaxed and effortless.

Physics: Hair, clothing, towels, food, steam, water droplets and household objects all move naturally with realistic weight and momentum.

Composition: Casual smartphone framing with occasional imperfect composition and spontaneous reframing. Feels like authentic dorm footage rather than planned content.

Continuity: Maintain the exact same face, platinum blonde long straight hair with bangs, body proportions and identity from the reference image throughout every scene. No identity drift.

Technical: Authentic 24–30fps smartphone video look with subtle compression artifacts. Final output 16:9.

Audio: Natural apartment ambience only. Air conditioner humming, distant Seoul traffic, birds outside the window, cooking sounds, boiling water, frying food, dishes, footsteps, running water, toothbrush brushing, hair dryer, refrigerator hum and soft room ambience. No background music. No subtitles.

Overall Mood: Warm, peaceful, elegant and comforting. Similar to a quiet behind-the-scenes dorm vlog uploaded by a Korean idol. Everything feels sincere, calming and intimate without trying to be entertaining.

Character:

A fictional Korean female idol in her early 20s.

Long platinum blonde hair with straight bangs.
Fair skin.
Soft cat-like eyes.
Slim figure.
Graceful and feminine.

She is known for her gentle personality, polite manners and calm speaking voice. She has just returned to the dorm after school and a schedule and quietly spends the rest of her summer evening.

Wardrobe Continuity:

CUT 1
She arrives wearing the exact sailor-style school uniform from the reference image, including the cream knit vest, blue sailor collar, blue pleated skirt, white socks and black loafers.

CUT 2
She changes into an oversized white graphic T-shirt and loose black athletic shorts while cooking.

CUT 3
She now wears a soft pastel blue sleeveless lounge top with white lounge shorts while eating dinner.

CUT 4
The same lounge outfit while brushing her teeth.

CUT 5
After washing her face she changes into luxurious white silk short pajamas featuring elegant gold piping along the collar, cuffs, chest pocket and shorts hem. The pajama set consists of a relaxed short-sleeve button-up shirt with matching silk shorts. Her hair is loosely tied back while doing skincare.

CUT 6
Fresh after showering, still wearing the same white silk pajamas with gold piping. Her hair is slightly damp with a white towel resting around her neck while drying it.

CUT 7
Relaxing on the sofa wearing the elegant white silk pajamas while hugging a plush toy and drinking iced tea.

CUT 8
Before bed she lets her hair down naturally, waves softly toward the camera while wearing the elegant white silk pajamas, turns off the bedroom light and closes the door.

Scene:

A beautiful modern luxury Korean idol dorm apartment during a peaceful summer afternoon and evening. Cozy, bright and elegant interior with tasteful furniture, skincare products, plush toys, snacks, fresh flowers, books and neatly organized living spaces.

CUT 1 — Coming Home

Continuous handheld shot.

She unlocks the front door, quietly enters the dorm wearing the school uniform from the reference image, removes her loafers neatly, places her school bag beside the entrance, smiles softly toward the camera and bows her head slightly.

Dialogue:
"다녀왔어요."

CUT 2 — Cooking
The camera follows her into the kitchen.
She washes vegetables, prepares simple Korean home cooking, fries eggs, boils ramen, tastes the soup with a spoon and smiles gently.

Dialogue:
"오늘 많이 배고프네요."

A female member is heard faintly from another room.

Female Voice:
"언니~"
She smiles softly.
"왜?"

CUT 3 — Dinner
She sits quietly at the dining table.
Before eating, she gently places her hands together.

Dialogue:
"잘 먹겠습니다."

She enjoys the meal peacefully while occasionally smiling toward the camera.

CUT 4 — Brushing Teeth
Bathroom mirror shot.
She brushes her teeth slowly before smiling shyly after making accidental eye contact with the camera.

Dialogue:
"이제 양치해야지."

CUT 5 — Skincare
Now wearing elegant white silk pajamas with gold piping.
She carefully applies toner, serum and moisturizer while calmly speaking to the camera.

Dialogue:
"오늘 피부가 조금 좋아 보이�� 것 같아요."

CUT 6 — After Shower
She walks out of the bathroom with slightly damp hair and a towel resting around her neck.
She quietly dries her hair while looking out the apartment window at the summer sunset.

Dialogue:
"샤워하니까 정말 개운해."

CUT 7 — Relaxing
She sits comfortably on the sofa hugging a plush toy while sipping iced tea and slowly scrolling through her phone.
A female member speaks softly from another room.

Female Voice:
"영화 볼래?"
She smiles warmly.
"좋아요."

CUT 8 — Good Night

Late evening.
She walks into her bedroom carrying her phone charger and plush toy.
She turns around one last time, smiles gently at the camera and gives a small wave.

Dialogue:
"오늘도 함께해줘서 정말 고마워요.
좋은 꿈 꾸세요."

She switches off the bedroom light.
The bedroom door slowly closes while the handheld camera quietly remains outside for another second before the recording naturally ends.

[楊哥 | Yang Onchain @YangOnchain]

山间巴士 × 流动绿光 × AI视频提示词分享：

这次想做一个很安静的日系胶片片段。

不是摆拍写真，
也不是普通车窗人像，
而是一个女生坐在老式山间巴士里，
窗外整片深绿树林在玻璃上流动成拖影。

耳机、单人座椅、透明车窗、低饱和绿光，
整个车厢像被夏天的绿色包起来。

以下视频为 AI 生成。
完整提示词见评论区第一条👇

这种「日系森系胶片短片」要不要继续测？

[波妞PONYO @ponyodong]

想做 AI 漫剧？先搞懂最大难点：场景调度

你有没有发现，AI 做多人打斗视频的时候，最容易乱的不是招式。而是人物一多，镜头就不知道该跟谁，空间也开始对不上。

这条视频我想讲的就是一个很实用的方法：
先不要急着写“飞身、挥刀、闪避”这些动作词。
先把场景画成一张 3D 俯视动线图。
谁从哪边进来，镜头从哪里推进，危险点在哪里，人物最后要往哪边走，都先在图上定清楚。

这样 AI 生成的时候，它拿到的就不是一句模糊的“打一场很燃的戏”，而是一张能拍的导演地图���
多人打斗想不乱，第一步不是加动作，是先控场。
这条可以当成一个小方法，以后做动作戏、舞蹈戏、多人调度都能用。
#AIGC #AI导演 #即梦AI #AI短片 #场景设计 #分镜设计

[KeenTools @keen_tools]

Our Images-to-3D-Head API is here

Converting ordinary photos into production-ready 3D heads – with MetaHuman topology & 360° textures. No hardware, depth sensors or manual cleanup. Fully automated, built for scale

Test it now: https://t.co/qalz27AUNR

#medtech #gamedev #vfx

[タナベ | 動画・音声生成AI解説 @tanabe_fragm]

復活したFable 5を使って、プレビズ作成ツールを��ってみました❗️

以前も作ったことがありますが、今回は細かいところまで調整できる、かなり使いやすいツールになったと思います😌
ただし、Claude（Pro）の利用分の制限にはすぐ達してしまいますね…💦

私が実際に使ったプロンプトを公開するので、よければそのまま試してみてください！

今回は、Mixamoからダウンロードしたモーションデータを参考データとして使用しています。
事前にMixamoで必要なデータをダウンロードしておくと、スムーズに試せると思います😊

Fable 5で使用したプロンプトはこちら👇
-----------------
Seedance 2.0などの動画生成AIに参照させるプレビズを作れるツールを作りたい。
要は、人間の細かい動きをブラウザに表示し、忠実に再現できるツールを作りたい。
「mixamo」でダウンロードできるようなデータを自分で作れるようになればOK。

要件は以下。

・人の動きを細かく指定できる
・単なる動きだけでなく移動（歩く・ジャンプ・飛ぶ等）も指定できる
・人間のよくある動きをプリセットできる（歩く、走る、ジャンプする、踊る、握手）
・マネキンの向きやカメラワークも指定できる
・1つ��画面に複数の人間を登場させることができる
・画面内で表示される動きはmp4で��存可能。
・男性・女性でマネキンは色分けして
・動きは段階を踏んで指定できるようにして（1、肩を上げる 2、肩を上げたまま肘を曲げる みたいな）

何か不明点があれば必ず聞いて欲しい。
そして、作るアプリの仕様が決まったら、それを書いて私に教えて。
私の許可を得てから作ってください。

mixamoからダウンロードしたデータを貼っておきますね。

[VoxelPlot @voxelplot]

Seedance 2.0 - Advanced Workflows Series

12. Cinematics with Unreal Engine and Claude Fable 5
Fable 5 is an absolute beast model that can take control of Unreal Engine and create videogames.

We can build scenes as a little video game with physics and assets and configure a playable character to create a cinematic scene.

In this example I created a scene of an astronaut lost in space floating between asteroids.

I created a zero gravity environment and made assets like cubes and spheres representing asteroids interact in this environment.

I recorded the gameplay and later place a camera to build the cinematic.

Workflow + Prompts 👇

[Rootport🔥 @rootport]

AviUtl2で動画生成AI「LTX 2.3」を動かせる拡張機能を作っている。バックエンドはかなり���成してきた。公式のLTX desktopにはない動作トレース機能「IC-LoRA」を組み込んだよ。