觉明

十年SRE工程师
探索如何用AI转行创业

[Tibo @thsottiaux]

Reminder that you can use the Codex App, CLI and SDK with any open source model, not just with OpenAI models.

https://t.co/spPifB4ck3

[傅盛 @FuSheng_0306]

给《丧尸清道夫》追着发offer的好莱坞AI大神PJ Ace，分享了9个AI做电影的技巧：
1. 角色先定死，正面背面近景远景各来一张
2. 场景生成360度环绕，截图就有所有机位
3. 每条prompt结尾贴同一套灯光+摄影描述，锁定风格
......
完整版见视频，大家可以试试。

[Emily @IamEmily2050]

My GitHub skill for Seedance V2 are free. I keep updating it all the time. No need to pay anyone; use it as the base for your own skill building. It also works with other video models. I will have an independent one for Kling and Grok Imagine later when they get to the Seedance V2 level.

[Cia0 @Cia0_exe]

Zui X Lonelia

Seedance 2.0 implement fast battle without any glitches in 10 seconds.

[Anima @Anima_Labs]

Today, a new cartoon-style test with Seedance 2!

Short cartoon film made with @invideoOfficial

We wanted to try a style quite close to French children's comic strips. I think we've all experienced that scene where our nephew (or child) tries to smoke their first candle, but there's more drool than air, messing things up 😂🎂

Made with Invideo's Agent One, the AI ​​agent that lets you manage everything in one place, from the script and character sheets to video generation, experimentation, and even part of the editing!

[OscarAI @Artedeingenio]

Fantasy worlds don't always need darkness and battles to feel epic:

--sref 1203290485

This Midjourney V7 style reference blends delicate ink linework, luminous pastel palettes, fairy-tale architecture and cinematic composition into a look that feels like a modern illustrated storybook.

Dragons, castles and young heroes are presented with a sense of wonder rather than danger.

What makes this style stand out is its balance between concept art and children's book illustration.

Every scene feels like a frame from an animated fantasy film, with soft lighting, elegant character design and vast landscapes inviting the viewer into the adventure.

[OscarAI @Artedeingenio]

Midjourney prompt:

anime character reference sheet of a female warrior with shoulder armor and leather straps, full pose and close-ups of hand, helmet, and profile view, technical cel-shading, blueprint composition, Japanese text annotations, soft post-apocalyptic landscape in watercolor --ar 16:9 --raw --profile jen8y7u --v 7

Seedance 2.0 prompt (opening sequence):

15-second continuous single-shot. No cuts. No scene transitions.

Style: epic anime gladiator duel, hand-painted watercolor and ink illustration brought to life, classic 1980s OVA aesthetic, Studio Ghibli character design mixed with historical fantasy action, cinematic choreography, realistic character motion, detailed Roman architecture, dramatic sunlight, sketchbook concept art aesthetic, dynamic camera work, emotional storytelling.

Sound design: roaring crowd, distant horns, footsteps on sand, leather straps creaking, metal scraping against metal, shields rattling, wind moving through the arena. No music. No dialogue.

A massive Roman-style coliseum packed with thousands of spectators. The young brunette warrior stands on one side of the arena wearing blue armor and carrying a short sword and round shield. Opposite her stands the young silver-haired gladiator wearing a bronze helmet and light armor, armed with a curved sword. Dust swirls across the golden sand as they prepare to fight.

0–3s

The camera descends from above the coliseum.

The crowd erupts in cheers.

The two gladiators slowly circle each other.

Dust blows between them.

3–6s

The silver-haired gladiator charges first.

The camera rushes forward alongside her.

She swings her curved blade.

The brunette warrior raises her shield and blocks.

Sparks burst from the impact.

6–9s

The brunette counterattacks.

The camera circles around both fighters.

They exchange rapid sword strikes.

Sand scatters beneath their feet.

The crowd grows louder.

9–12s

The silver-haired gladiator leaps onto a fallen stone block.

She launches herself forward.

The brunette pivots and narrowly avoids the attack.

Both fighters slide across the arena floor.

Dust explodes around them.

12–15s

The camera sweeps around the duel.

The two warriors charge simultaneously.

Their blades collide in a brilliant shower of sparks.

The impact freezes them for a brief moment.

The crowd stands and roars.

Final frame:

Epic cinematic wide shot of both gladiators locked blade-to-blade in the center of the coliseum, surrounded by swirling dust, golden sunlight, and thousands of cheering spectators.

Negative prompt:

No fatalities, no gore, no blood splatter, no dismemberment, no modern clothing, no firearms, no magic powers, no floating characters, no walking on air, no exaggerated anime physics, no dialogue, no subtitles, no text overlays, no logos, no extra limbs, no distorted anatomy, no camera cuts.

[OscarAI @Artedeingenio]

Midjourney prompt for the Titan:

design sheet of a humanoid giant with exposed muscle system and steam vents, drawn in anime mecha design sheet style, full body and head angles, close-up of hand and facial structure, technical breakdown layout with handwritten Japanese notes, watercolor background of ruined cities --ar 16:9 --raw --profile jen8y7u --v 7

Prompt for the swordsman:

blueprint style anime character sheet of a young scout with maneuver gear, aerial pose plus static front and side views, gloves and sword detail, technical cel-shading, soft Japanese watercolor background with rooftops, design sheet layout with labels and kanji --ar 16:9 --raw --profile jen8y7u --v 7

For the first animation, I only used the Titan character sheet as the reference image. I then used Seedance 2.0’s Extend feature and added the swordsman character sheet. After that, I extended the sequence two more times without uploading any additional images, because Seedance had already learned and recognized both characters.

Seedance 2.0 prompt (first sequence):

15-second continuous single-shot. No cuts. No scene transitions.

Style: post-apocalyptic anime action, ultra-detailed biomechanical giant, hand-painted watercolor and ink illustration brought to life, sketchbook concept art aesthetic, Studio Nue and classic OVA influence, Akira-inspired destruction, Mad Max energy, dynamic cinematic camera, realistic mechanical animation, industrial dystopian atmosphere, dust, smoke, sparks, and debris.

Sound design: heavy mechanical footsteps, hydraulic pressure releases, metal groaning, collapsing concrete, distant explosions, steam vents, electrical crackles, debris falling, roaring wind through ruined streets. No music. No dialogue.

A colossal biomechanical titan with exposed synthetic muscles, glowing reactor veins, armored fists, and a skull-like face charges through the ruins of a dead megacity. Heat radiates from cracks in its body. Every movement feels impossibly powerful.

0–3s

The camera races low through a devastated industrial street.

Smoke and dust fill the air.

Suddenly the giant crashes through a collapsing building ahead.

Concrete explodes outward in every direction.

3–6s

The camera flies backward while facing the titan.

The creature sprints directly toward the viewer.

Each footstep sends shockwaves through the ground.

Cars, debris, and metal containers bounce violently into the air.

6–9s

The titan leaps.

For a brief moment the enormous body hangs above the city.

The camera rotates around him as fragments of concrete, sparks, steel beams, and dust swirl through the air.

The glowing veins across its body pulse like a reactor overload.

9–12s

The titan crashes back into the street.

A massive shockwave erupts.

Windows shatter across nearby skyscrapers.

Steam bursts from its shoulders and forearms.

The camera dives beneath one gigantic punch.

12–15s

The titan drives its fist into the ground.

The entire street fractures.

Buildings begin collapsing in sequence toward the horizon.

The camera pulls rapidly upward into a huge aerial shot as dust clouds consume the city.

Final frame:

Epic wide cinematic view of the biomechanical titan standing amid a collapsing megacity, steam rising from its body, glowing reactor veins burning through the haze while entire districts crumble around it.

Negative prompt:

No static poses, no standing still, no dialogue, no subtitles, no text overlays, no logos, no cartoon motion, no low-quality animation, no extra limbs, no distorted anatomy, no camera cuts, no frozen action.

You can now enjoy 25% off Seedance 2.0 on Mitte 🎉

Here’s the link if you’d like to give it a try:

https://t.co/3358y75ldz

[Loriel.AI @ou_zhen599]

Design a striking black-and-red prison-themed character poster featuring a live-action cosplay version of Tifa Lockhart from FF7. The subject of the poster is an East Asian adult woman in her 20s, embodying the refined aesthetic of Japanese and Korean cosplay, blending the elegant charm of Japanese cosplay photo books with the figure proportions of a Korean idol. Please retain Tifa Lockhart’s iconic features and translate them into believable, high-end cosplay attire, rather than cheap cosplay costumes. Style direction: A high-saturation, dark-themed Japanese-style game character profile poster, featuring a black-and-red profile panel on the left and a stark, gray-white concrete prison wall on the right; a 3D promotional image in an anime-realistic style; High-end gacha character card cover aesthetics; glossy, luminous CG skin with subtle sweat marks; striking vintage-style typography; densely detailed graphic profile panels; oversized SSS-class rating; FF7-related logos; cold white commercial interrogation lighting; sharp, anime-style beauty face; strong foreground-background separation; not ordinary realistic photography, not low-saturation documentary portraits, nor low-saturation mugshots. Focus on depth and impact: strong foreground-background separation. Tifa Lockhart must visually occupy the space in front of the left-hand filing cabinet and the stark gray-white concrete height measurement wall, rather than blending into the background. The black filing cabinet on the left and the light-colored concrete wall on the right serve as flat background graphic layers; Tifa Lockhart is the primary subject in the foreground—she is larger, brighter, sharper, and closer to the camera. The black prisoner tag is the object closest to the viewer, standing out strikingly in the foreground and casting a faint shadow on her. Utilize clear occlusion, depth of field, parallax effects, and sharp highlights along the edges of her hair, shoulders, hands, jewelry, and the tag. The character occupies 75%–85% of the poster’s height, with a close-up shot extending from the head to the upper thighs, creating a highly impactful cover composition. Poster layout: A high-impact, vertical prisoner profile poster. On the left is a striking black vertical profile panel, occupying approximately 25%–30% of the frame. The panel features character-related insignia, the bold white title text “Tifa Lockhart,” the smaller Japanese text “Tifa Lockhart,” the prisoner file, prisoner code FF7-0TF, “High Risk,” the SSS danger level, a red classified seal, a barcode, a warning label, and realistic file typography. The text should convey a dense, visual, collectible, and aged feel, but should not contain excessive small text that must be clearly legible. Right: A close-up shot shows Tifa Lockhart standing in front of a cold, off-white concrete prison height measurement wall. The wall features a light gray-white cement texture with black height markings, fine cracks, scratches, dust, faintly visible Shinra-style patterns, Shinra-style graffiti integrated into the background, and high-contrast, cool-toned shadows. In contrast to the black archival panel on the left, the wall must appear bright, cold, and austere—slightly overexposed yet retaining its rough texture. The composition extends from her head to the upper thighs, presenting three strong visual focal points in sequence: her delicate anime-style beauty, the oversized black prisoner ID tag, and the contours of her waist and abdomen. Her arms are crossed across her upper body, holding an oversized matte black prisoner ID tag. Outfit: A high-end black strapless bandeau bikini featuring a strapless design. The bikini top is completely covered by the black prisoner ID tag, revealing only her shoulders, arms, waistline, abdominal contours, leg proportions, and low-rise bikini bottoms. Absolutely no straps, thin straps, halter straps, neck straps, bra straps, or transparent straps. Red-and-black patchwork fingerless combat gloves, along with understated FF7-style accessories. Pose and Expression: A regal prison intake pose, with her chin slightly raised, eyelids half-lowered, a cold and commanding gaze, and a calm yet contemptuous expression. Shoulders relaxed, posture upright and elegant, hands clenched tightly around the name tag, exuding calm confidence—as if undergoing a mugshot for prison registration yet still commanding the entire room. The black name tag is the primary visual element in the foreground: a matte surface with a thick black border, featuring clean, crisp white text reading “Tifa Lockhart,” with slightly worn edges; the grip of the hands and the pressure of the fingers should appear realistic and natural. The nameplate must clearly cover the upper body, and the composition must remain restrained and sharp: no nipples visible, no nudity, no see-through clothing, no sexual acts, and no cropping of body parts. Lighting and atmosphere: cold white interrogation lighting. A strong, cool-white key light from the upper left illuminates the subject, making them brighter than the background and casting clean, bright highlights and cool-toned shadows on the face, collarbone, waistline, edges of the sign, Shinra accessories, and concrete elevation markers. A cold contour light from the right edge separates the hair, shoulders, hands, accessories, and the sign.

[李岳 @liyue_ai]

GPT Image 2生成3D CG 插画：雨中灵姬。

提示词：
9:16 竖版，高精度 3D CG 东方幻想女性角色写真，3D CG oriental fantasy beauty portrait，anime-style 3D CG character art，semi-realistic 3D character render，镜头为大腿及上半身构图，画面主体是一位明确成年的年轻东方幻想系女性，视觉年龄约 20–26 岁，整体气质清冷、空灵、精致、安静，带有雨中水系精灵般的神秘感与高级感。整体不是平面插画，而是高完成度 3D CG 角色渲染，具有精致角色建模、真实材质表现、电影级冷调柔光与高级虚拟角色海报质感。

人物拥有精致的东方美型脸，小巧流畅的鹅蛋脸，皮肤冷白细腻，带轻微通透感与柔和皮肤着色，肌肤表面有细腻水珠与湿润反光。眼睛细长清澈，瞳色为冰蓝绿调，瞳孔有通透玻璃感与细致高光，眼神微垂，安静、疏离、略带脆弱感。睫毛纤长，眼妆干净克制，鼻梁秀气挺直，嘴唇柔软，唇色为低饱和裸粉色，嘴唇微启，神情冷静而迷人。

发型为短款蓝黑色渐变发，主色为浓黑与深海军蓝，发尾带冷调蓝色高光，发型为短层次波浪感短发，一侧额发自然垂落遮住部分脸颊，顶部带编发结构，发丝湿润、轻盈、细腻，具有高精度发丝建模与柔顺光泽。耳部佩戴精致蓝色水晶几何耳饰与金属耳骨夹，增强东方幻想与水元素气质。颈部佩戴白色高领装饰项圈与青蓝色宝石流苏细节，精致而高级。

服装为精致的东方幻想水系礼装，上半身穿着白色轻薄、湿润感的缎面的贴身长裙，布料柔软垂坠，带细腻高光和微透感，胸口与躯干线条以克制优雅的方式表现。外层披着一件宽松白色衬衫式轻纱外搭，自然滑落至手臂与腰侧，形成层叠褶皱与飘逸感，增强随性与梦幻氛围。腰臀处可见黑色蕾丝边短裤细节，作为整体造型中的轻哥特点缀，使画面更具层次与精致感，但整体保持高级、克制、不低俗。

人物姿势为优雅侧身站立，身体呈现流畅的 S 型曲线，肩颈舒展，背部与腰线自然延展，头部微微低垂侧转，视线向下，整体姿态安静、轻盈、克制。手臂自然下垂，手指修长，姿态如同在雨中静立的一瞬间，突出侧脸、肩颈、背部、腰臀线条与服装湿润材质的精致表现。

背景调整为精致的东方幻想水境场景，整体以冷白、冰蓝、淡灰为主色调。背景为朦胧的雨幕、轻雾与若隐若现的东方幻想建筑轮廓，可融入远处模糊的亭台、石阶、水池边缘或空灵宫殿轮廓，营造高级幻想空间感。人物周围有动态水花、水流弧线与透明水晶般飞溅效果，像由水元素凝结出的装饰性波纹，背景适度虚化，既通透空灵，又不喧宾夺主。

光线采用冷白主光与柔和逆光结合，整体为雨天高调低饱和电影感打光。面部、锁骨、肩背、手臂与腿部有细腻湿润高光，发丝边缘与水花边缘有轻微轮廓光。白色布料、蕾丝、金属、宝石、水珠与水花都具有真实清晰的材质反馈。整体画面干净、通透、冷艳、梦幻、精修感强，具有高端收藏级 3D CG 东方幻想角色海报质感。

high detail 3D CG render, Unreal Engine quality, Octane render look, cinematic lighting, PBR materials, subsurface scattering skin, realistic wet hair strands, translucent wet fabric, crystal water splash effects, delicate lace details, elegant oriental fantasy styling, aquatic spirit atmosphere, volumetric light, cinematic depth of field, soft bloom, ultra detailed, polished anime realism, premium virtual character portrait

[Kiki @Mayz1169]

swimmer interview 🏊‍♂️

made with GPT Image 2 + Seedance 2

prompt⬇️

[Kiki @Mayz1169]

1️⃣character sheet:

high-detail anime character sheet,
professional character design reference,
full body standing pose,
multiple angles and expressions,
swimwear reference,
commercial anime art style,
light novel illustration style,
masterpiece, best quality, ultra detailed, 8k

ASPECT RATIO:
16:9 (can be wide to fit multiple poses)

CHARACTER:
female competitive swimmer,
long silver/white hair,
purple or violet eyes,
athletic and toned physique,
tall and elegant body proportions,
confident but calm expression,
professional racing swimsuit (navy blue with white/purple stripes),
swimming cap with goggles on head,
wet skin highlights and subtle reflections,
realistic anime anatomy,
anime idol sports aesthetic

POSES:
main front full-body standing pose,
side view, back view, profile view,
leaning, sitting poolside, stretching, starting position,
hands on hips, neutral expression,
full-body dynamic and neutral positions for reference

EXPRESSIONS:
neutral, slight smile, wink, serious, surprised,
subtle facial emotions,
detailed anime expressions

DETAILS:
close-up headshot,
swim cap and goggles,
swimsuit back details, swimsuit textures,
hands and feet close-up,
hair strands, skin highlights,
light reflections on wet skin

BACKGROUND:
minimal or clean white background for clarity,
poolside elements optional in secondary poses,
character-focused reference sheet style

COLORS:
navy blue, white, purple, skin tones,
hair silver/white with subtle shading,
high contrast highlights,
anime-style smooth color transitions

CAMERA/COMPOSITION:
full-body front view main,
secondary smaller angles around main figure,
consistent perspective,
clear silhouette studies,
professional character sheet layout

OUTPUT:
character sheet format,
all poses and expressions included,
numbered sections optional,
reference sheet for animation or illustration,
masterpiece, best quality, ultra detailed, 8k

[Yamato Sakaguchi @sakaguchi_ymt]

Testing Storybords × AI
ストーリーボードを参照し、それぞれのカットを制作するテスト

※テストのためストーリーボードもAIで制作

[MayorkingAI @MayorKingAI]

New style created with Midjourney Style Creator

[Prompt] --sref 3262487664 --ar 16:9

[小互 @xiaohu]

Claude Code 升级为4.8后老是出这个

The model's tool call could not be parsed (retry also failed).

你们遇到过吗？

好烦

[Glitter Gal @GlitterPixely]

Character sheet format:
Create a clean 16:9 horizontal character reference sheet optimized for Seedance.

Layout:
- large full-body front view on the left in a neutral standing pose
- face close-up with neutral expression
- left-side profile view showing the character from the left side
- 3/4 body view
- right-side profile view showing the character from the right side
- full-body back view
- accessory/detail callout section if the character has important props, armor, weapons, tech, jewelry, or markings
- small facial expression strip with neutral, happy, angry, surprised, and sad expressions
- bottom color palette strip with labeled swatches
- short consistency notes

Character:
[Describe the character clearly here: age appearance, face shape, skin tone, eye color, hairstyle, body type, outfit, shoes, accessories, materials, colors, markings, special features, and silhouette.]

Keep consistent:
Same face, hairstyle, hair length, hair color, eye shape, eye color, skin tone, body proportions, outfit, shoes, accessories, materials, color palette, markings, silhouette, and overall design across every view and expression.

Seedance optimization notes:
Use clear readable shapes, neutral poses, clean silhouette, consistent proportions, visible hairstyle from front, side, and back, clearly separated views, and soft even lighting. Make every view look like the exact same character.

Style:
Clean production character sheet style, plain neutral background, soft even lighting, crisp readable linework, balanced spacing, no dramatic shadows, no cinematic lighting, no motion blur.

Do not:
Do not redesign the character. Do not change the outfit. Do not change the hairstyle. Do not change colors. Do not create alternate versions. Do not add extra characters. Do not add a background scene. Do not use action poses. Do not use extreme perspective. Do not make the expression strip look like different characters.

[Kiki @Mayz1169]

AI is getting scary good at fashion videos

made with GPT Image 2 + Seedance 2

workflow⬇️

[Latte @0xbisc]

OOTD transformation

made with GPT Image 2 + Kling 3

Here’s my workflow ↓

[小宇Chengzi @Chengzilhy]

以我的人物模特图作为唯一人物身份参考，严格保持我的模特本人的脸部特征、五官比例、年龄感、肤色、发型、气质和真实身材比例；不要参考示例图人物长相，只参考这张图的非身份视觉信息、穿搭风格、版式结构和视觉排版。 生成一张高级白底穿搭清单海报，整体为干净极简的时尚穿搭展示图，浅灰白摄影棚背景，画面清爽、有留白、像服装搭配教程封面。 画面右侧是一位成年女性全身模特，人物完整站立入镜，位于画面右半区偏中位置，占画面高度约75%–80%。模特正面站立，身体微微自然放松，一只手举起黑蓝红配色掌机游戏机，另一只手自然下垂，眼神看向镜头，表情清冷、克制、微酷，不夸张微笑。整体是甜酷辣妹穿搭风格。 服装搭配：黑色一字肩短款上衣，肩部微露，短袖贴身版型，胸前有小蝴蝶结金属装饰；下身是黑色低腰工装短裙，裙身有拉链、口袋、抽绳、褶皱和金属扣细节；腿部穿黑色半透花纹连裤袜，上半段微透，下半段像长筒袜拼接效果；脚穿黑色厚底机车短靴，鞋底厚重，有金属扣带和拉链细节。配饰包括银色爱心项链、蓝色蝴蝶结发夹，整体发型柔顺长发。 画面左侧是穿搭物品清单区，采用黑色细线边框分成三个横向模块，每个模块左上角有巨大黑色编号“1、2、3”。 第1组展示配饰和游戏机：银色双层爱心项链、深蓝色蝴蝶结发饰、黑蓝红配色掌机，每个物品下方标注“1x”。 第2组展示下装和鞋子：黑色工装短裙、黑色厚底机车短靴，每个物品下方标注“1x”。 第3组展示上衣：黑色一字肩短款抽绳上衣，下方标注“1x”。 左下角有虚线分隔线，下方写“TOTAL 6 ITEMS”，旁边有六个简洁线稿小图标：项链、蝴蝶结、游戏机、短裙、靴子、上衣。 顶部左侧有大号粗黑中文标题“甜酷辣妹穿搭”，下方是大写英文标题“SPICY GIRL FIT”，字距拉开，黑色极简排版。右上角有一个手绘风爱心小图标。整体排版保持高级、干净、白底、黑白主调，穿搭以黑色为主，点缀蓝色发夹和游戏机颜色。画面像小红书穿搭清单封面、AI时尚搭配教程海报、clean outfit breakdown poster。 负面提示词：不要改变我的模特身份，不要参考示例人物长相，不要二次元感，不要卡通感，不要低清晰度，不要脸部模糊，不要手指错误，不要腿部畸形，不要文字乱码，不要排版混乱，不要物品缺失，不要多余人物，不要背景杂乱，不要廉价海报感，不要过度性感化，不要暴露过度。

[觉明 @jueming0312]

@GlitterPixely Hello, I only saw prompts for generating storyboard images in this article. Did you share any prompts for generating videos?