这是提示词合集而非可安装软件,README 未给出任何安装或运行命令;提示词需自行复制到对应模型(如 GPT Image 2.6、MiniMax H3)使用,仓库本身不提供模型
awesome-image-video-prompts
ai-boostGPL-3.0活跃维护收录可直接复制的 AI 图像与视频提示词,每条都附真实生成效果和来源,还提供用 agent 编辑视频的配方。
Copy-ready prompts for the leading AI video and image models, each with its real output and source. Plus agent recipes for editing video.
收录主流 AI 图像与视频模型的即用提示词,每条附真实输出和来源,另有视频编辑 agent 配方
收录可直接复制的 AI 图像与视频提示词,每条都附真实生成效果和来源,还提供用 agent 编辑视频的配方。
核心能力
- 按题材分类整理 1081 条视频提示词与 1392 条图像提示词
- 每条提示词逐字引用模型官方指南或创作者原帖
- 附真实生成结果与来源链接,可点开看原片
- 提供 5 条用 agent 编辑视频的配方
- 按模型与标签检索,可跳在线画廊搜索筛选
容易踩的坑
暂无 release 数据(可能没有发版,或尚未同步)。
用户点评
登录用户的真实使用记录,可按版本和硬件筛选还没有人点评。第一个说出真实体验的人最有价值。
Formerly Awesome-GPTs. The original GPTs list lives on in legacy/awesome-gpts.md.
Contents
- Recently added
- By model
- Video prompts (1081 entries)
- Cinematic scenes and short films (150)
- Action, chases and sports (210)
- Sci-fi and fantasy (94)
- Anime and animation (81)
- Ads and product videos (130)
- Dialogue and talking heads (57)
- Music and dance (49)
- Phone footage and vlogs (174)
- Nature, travel and aerial (11)
- Animals and pets (30)
- VFX and transformations (95)
- Image prompts (1392 entries)
- Posters and editorial layouts (204)
- Infographics and slides (93)
- Phone and UI mockups (30)
- Product shots and e-commerce (78)
- Logos, ads and packaging (27)
- Photo realism (122)
- Portraits (273)
- Characters and figures (108)
- Illustration (118)
- Comics and storyboards (21)
- Photo to art (111)
- Architecture and interiors (24)
- Maps (15)
- Collage (103)
- Photo editing (65)
- Agent recipes: editing video (5 entries)
- Guides and skills
- 3D with agents
- Browse by tag
- Contributing
Recently added
- 2026-11-06 · Overhead POV selfie in bedroom, white tee and shorts · GPT Image 2.6
- 2026-11-06 · Deadpan Tibetan fox with label above head · GPT Image 2.6
- 2026-11-06 · Woman leaning forward in a car at night · GPT Image 2.6
- 2026-11-06 · Glowing snowman in gilded frame · GPT Image 2.6
- 2026-11-06 · Woman pushing cart looks back in Sam's Club · GPT Image 2.6
- 2026-11-06 · Woman in light blue sweatshirt before polaroid wall · GPT Image 2.6
- 2026-11-06 · Woman seated in moonlit lake water · GPT Image 2.6
- 2026-11-06 · Glowing cyan poetry and equations on wet skin · GPT Image 2.6
- 2026-11-06 · Red apple hanging from a tree branch · GPT Image 2.6
- 2026-11-06 · Woman in mirrored elevator with Keor embroidered top · GPT Image 2.6
- 2026-11-06 · Woman with bob holding red maple leaf · GPT Image 2.6
- 2026-11-06 · Chinese weekend book club poster with open book · GPT Image 2.6
By model
- GPT Image 2.6: 93 prompts, 43 of them animated GIFs, each with its output and source
Video prompts
Every prompt below is quoted verbatim from its source: the model maker's own guide or model card, or a creator's post on X with the video attached. Click a thumbnail to watch the original with sound. Grouped by what you want to make, covering the leaders of the Artificial Analysis video arena; each entry names its model, and creator posts are sorted by likes, counted on the date shown.
Cinematic scenes and short films
Twenty shot character detail montage, face revealed last
A numbered shot list with time blocks and one abrupt camera move per shot builds three connected action bursts. Faces stay out of frame until shot 20, and hard match cuts link similar shapes without morphing.
MiniMax H3 · Reference-to-video · Source: @aimikoda on X · 3,281 likes on 2026-10-28
character-reference montage hard-cuts camera-moves studio-background
Use @[character reference] for the character's identity, anatomy, costume, colors, materials and visual style. Adapt body-region cues to its actual anatomy, replacing absent features with existing equivalents. Invent no anatomy or accessories. Select signature inserts from its most recognizable details. Rigid parts rotate; flexible materials bend or stretch.
A 15-second character-detail montage: exactly 20 shots joined by hard cuts. An empty, seamless warm-white studio background; soft directional light reveals the surfaces. Shots 1-19 are tightly cropped body details and rear angles: keep the face, eyes and facial profile entirely outside the frame, including reflections. Shot 20 is the first and only face reveal. No full-body view.
Three connected bursts: a sharp head turn with the body bracing; a coiled arm opening into a strong sweep and torso twist; one short step into a planted pivot, finishing with a sudden look into camera. Carry body state across cuts without resetting to neutral. Existing hair, clothing and attachments react naturally. Keep the character active, with clear articulation and weight shifts.
Jump between distant body regions. Mix brief visual punches with longer moving-detail reads. Use one abrupt camera move per shot except locked-camera inserts 4 and 11, where the character supplies the speed. Vary height, angle and direction. Brake moves briefly for legibility while the subject keeps moving. Local whips stay tightly cropped; changes of body region use hard cuts. Connect selected cuts through similar shapes or motion vectors, without morphing. Real-time action, no slow motion or freeze frames.
0-3s, shots 1-5. A concealed head turn triggers the first burst:
1. Extreme upper-back detail: a crash push catches its surface shifting as a sharp head turn begins entirely above the crop.
2. Fingers: a short downward snap catches them curling into a fist as the body braces.
3. Rear three-quarter shoulder detail: a tight camera arc catches the shoulder drawing back behind the turning head.
4. Locked floor-level foot detail: the grounded support foot braces, its raised heel pressing down firmly without taking a step.
5. Rear neck detail: a short upward tilt catches the nape stretching with the turn; stop below the back of the head, keeping the jaw and face excluded.
3-9s, shots 6-13. The coiled arm opens into a forceful sweep:
6. A signature upper-body detail: a fast diagonal slide catches an existing seam, joint or distinctive contour responding as the arm draws inward.
7. Elbow: a sharp lateral track catches the folding joint tightening the arm's coil.
8. Supporting knee: hard match cut from the bent elbow to the knee's similar angle; a short camera drop catches the leg compressing under the same preparation.
9. Wrist: a quick shallow orbit catches it turning outward to initiate the release.
10. Upper chest: a short punch-in catches the chest opening as the arm drives outward beyond the crop. Let this moving surface read slightly longer.
11. Locked hand detail: fingers spring open and the hand sweeps rapidly across the fixed frame. Catch one crisp, readable phase of the spread fingers before they leave the crop.
12. Upper back: a sharp sideways slide continues that screen direction as the shoulder rolls through the sweep and the torso twists.
13. A distinctive torso or waist detail: a close diagonal move catches its physical response to the twist, such as an existing fold pulling taut or an attached piece swinging. Give the material response a clear moving read.
9-13s, shots 14-19. One short step and a pivot build toward the reveal:
14. Thigh of the other leg: an abrupt low tracking move catches it driving the single short step out of the twist.
15. Advancing foot: a brief downward whip catches it landing; show clear floor contact.
16. Ankle and lower-leg detail: a fast side slide catches the landed leg accepting weight as the ankle flexes.
17. Waist: a tight orbit catches the torso rotating above the planted-foot pivot. The foot supports the turn rather than sliding across the floor.
18. Back-of-head contour: a short lateral whip catches the start of the return turn from behind. Keep only the rear contour in frame, never the facial profile.
19. A signature shoulder detail: a sharp diagonal slide catches its follow-through beneath the turning head, which stays completely above the crop. Cut at the peak of this movement.
13-15s, shot 20. First face reveal:
20. Hard cut to a frontal face close-up as the character sharply completes the turn toward the lens. An explosive short push reveals the entire face clearly for the first time and stops tightening before cropping out its features. Direct eye contact lands while hair, clothing or natural body follow-through remains alive. Keep the whole face readable through the end; finish on living motion, without a held pose or another cut.
No added props, scenery, text, graphic overlays or split screens.
Carved limestone relief map reveals drama credits
Timed beats at 3, 6, 9 and 11.5 seconds give each credit its own moment and a clear hold cue, so every name stays readable. Explicit bans on mixing other civilizations and inventing glyphs, plus an audio list that names the instruments to use and to avoid, keep the output on-theme.
MiniMax H3 · Text-to-video · Source: @HBCoop_ on X · 1,032 likes on 2026-10-29
maya title-sequence typography timed-beats
Create a visually rich 15-second, 16:9 opening-title sequence for a historical drama titled THE JADE SUN, set within a powerful Classic Maya city-state.
The story concerns royal succession, drought, trade and political rivalry. Build the titles from historically grounded materials rather than conventional battle scenes.
VISUAL LANGUAGE
Use warm limestone, deep jungle green, polished jade, black obsidian, red mineral pigment, woven cotton and flickering firelight. Show corbelled architecture, reservoir maps, carved relief details, jade mosaics, cacao vessels and the shadow of a ceiba tree.
Do not mix Maya imagery with Aztec, Incan, Egyptian or European motifs. Do not generate decorative glyphs or invented ancient writing.
Display only the exact English credits below. Use monumental serif typography inspired by carved geometry without imitating Maya glyphs. Keep all letters correctly spelled, flat and readable.
0–3s: Sunlight travels across a limestone relief. Shallow carved lines form a map of reservoirs, raised roads and temple platforms. Red pigment fills one route and reveals:
A SERIES BY
ELENA MAR
Hold the credit clearly.
3–6s: Small pieces of polished jade slide across black cloth and lock into a circular royal ornament. The empty center opens to reveal:
STARRING
TOMAS REYES
The jade stops moving before the name appears.
6–9s: An obsidian mirror reflects firelight and the partial silhouette of a ruler wearing an authentic feathered headdress. Condensation retreats from the mirror to reveal:
CREATED BY
ISABEL CRUZ
Hold the text in the clean dark reflection.
9–11.5s: No text. A stone model of the city is gradually covered by the shadow of an approaching eclipse. Reservoir channels dry and turn from blue-green to dark red. One jade bead rolls into the remaining light.
11.5–15s: The bead becomes a brilliant green sun against black obsidian. Limestone letterforms rise around it to reveal:
THE JADE
SUN
Gold-green light moves once across the title, then everything becomes still. Hold through the end.
Audio: distant jungle atmosphere, a low wooden slit drum, clay flute breath, stone movement and one restrained conch tone during the eclipse. No generic chanting, fantasy music, orchestral trailer impacts or modern instruments.
Living room FPV dives into TV toast scene
The pointing finger is the guiding vector for the whole move, so the camera follows the gesture into the TV. Each stage is timed in seconds, and the screen image is written as a living scene with only subtle breathing and glass movement, which makes the portal entry match the second reference.
MiniMax H3 · First & last frame · Source: @CharaspowerAI on X · 1,005 likes on 2026-10-29
fpv one-take tv-portal reference-images
Here’s the prompt it created for this FPV sequence 👇
Use Image 1 as the exact opening frame and Image 2 as the exact destination world, creating one uninterrupted 15-second ultra-smooth cinematic FPV one-take with no cuts, no glitches, no teleportation, no jumpy motion, and no visible camera, as if an invisible stabilized micro-FPV lens is drifting naturally through a warm, lived-in living room and then physically flying into a television to emerge inside the second scene. Begin by preserving Image 1 perfectly: the man stays in the same yellow knit shirt, seated deep in the leather recliner with his boots propped up, cigarette smoke rising from his right hand, beer can still present, his left index finger urgently pointing toward the television in the direction he is already gesturing, with the same casual-but-suddenly-focused expression, the same residential living room geometry, the same pleated curtains, table lamp glow on the right, worn boot soles in the foreground, leather reflections, knit texture, and nostalgic warm low-key lighting. For the first 0–3 seconds, start almost motionless and then begin a very subtle forward FPV drift with a slight sideways arc that feels physically weighted and organic, skimming past the foreground boots and the armrest edge while the camera naturally aligns itself with the line of his pointing finger; use the gesture as the guiding vector for the whole move, letting the viewer feel the television is the true target of the composition. Between 3–7 seconds, perform a smooth curved tracking arc around the man’s pointing side, maintaining his face, hand, smoke, chair, and curtains in consistent perspective while progressively revealing the television in the room; the motion should feel like a real flying camera easing around a seated subject, not a pan or zoom, with strong parallax between the foreground boots, the midground man, and the soft curtain plane behind him. As the television becomes visible, its screen should already contain Image 2 as a living cinematic scene, not a static frame: the same blonde man in a black tuxedo jacket, white shirt, and black bowtie, with slicked-back hair, a knowing confident smile, and a champagne coupe raised toward camera, but animated with subtle, elegant human micro-motion only—soft breathing and natural chest rise, a faint head-and-shoulder shift, natural blinking and tiny eye movement, minimal facial micro-expressions that preserve the original confident smile, slight hand and finger adjustment around the glass, a gentle wrist rotation or small lift of the coupe, delicate liquid movement inside the glass caused by the motion, tiny reflections sliding across the glass surface, a subtle weight shift through the body, and barely perceptible cloth movement in the tuxedo from breathing and posture. From 7–11 seconds, gradually accelerate directly toward the television in a smooth, elegant FPV approach, keeping the man in Image 1 and the TV in a single spatial relationship, with realistic lens behavior, soft screen reflections, faint glass depth, slight perspective distortion at the screen surface, and natural falloff in depth of field as the screen grows larger and larger until it dominates the frame. The television surface must feel physically real, with subtle room reflections and believable contrast between the screen image and the surrounding bezel, and the approach must remain seamless, continuous, and stable, with no abrupt zoom, no wobble, and no transition artifacts. As the camera gets closer, the motion inside the TV should continue naturally and remain alive, with the character’s breathing, blinking, tiny gaze shifts, and champagne glass micro-movements persisting uninterrupted so the screen feels like a real moving scene rather than a frozen image. From 11–15 seconds, the camera flies through the television screen in one uninterrupted motion, crossing the glass plane invisibly so that the screen image expands to full frame and becomes the actual world of Image 2 behind it; at the exact crossing moment, perfectly match the perspective, lighting, and framing of the TV image so the transition is imperceptible, as if the television were a real window into that scene. Once inside Image 2, preserve the tuxedoed blond man exactly: his slicked-back hair shape, confident direct gaze, black bow tie, white shirt collar, champagne coupe, shallow depth of field, sparkling bokeh background, glossy highlights on glass, and opulent celebratory atmosphere, while continuing the same minimal natural motion—breathing, blinking, tiny facial changes, minute hand and glass adjustments, and subtle liquid and reflection movement—so the final image feels alive and cinematic rather than frozen. Continue a gentle deceleration after entry so the final frame settles into a composition very close to Image 2: his face sharp, the raised glass elegant in the foreground, reflective highlights crisp, background bokeh glowing softly, and the overall mood polished, cinematic, and premium. The final image should feel like a flawless Hollywood-level FPV match transition from a warm domestic interior into an opulent toast scene, with realistic parallax, physically believable motion blur, subtle smoke, rich contrast, stable facial identity, correct anatomy, consistent clothing, faithful room continuity, seamless screen-entry logic, living motion inside the TV, and a perfectly composed hero ending frame.
Zombie outbreak in a busy shopping mall
Starts in normal trading hours so the outbreak has something to break. Ends on a quiet beat instead of an explosion.
Seedance 2.5 · Text-to-video · Source: @auqibhabib on X · 1,002 likes on 2026-10-27
Create a photorealistic live action horror short film set inside a busy modern shopping mall during normal operating hours. The scene must feel like a real emergency captured by a human handheld camera operator, not a choreographed action movie. Prioritize realistic human behavior, believable physics, natural body movement, consistent character appearance, spatial continuity, practical effects and authentic environmental detail.
MAIN CHARACTER: A male mall security guard in his late 30s wearing a realistic dark security uniform and carrying a radio. Keep his face, clothing, body proportions and equipment identical throughout the entire video.
SURVIVORS: A small group of ordinary shoppers with realistic clothing and consistent appearances. Their faces, clothing and body proportions must remain unchanged throughout the sequence.
0–4 SECONDS:
Begin immediately in the food court with shoppers eating and walking normally. Without warning, an infected man suddenly attacks a nearby shopper. The attack is chaotic and physically messy rather than theatrical. The victim falls against a table, knocking over food and drinks. Nearby shoppers freeze for a moment because they do not understand what is happening.
4–8 SECONDS:
Several shoppers back away. One person briefly tries to help before realizing the attacker is dangerous. Chairs scrape across the floor as people retreat. The security guard hears the disturbance and runs toward the food court. His reaction is confused and concerned, not heroic.
8–12 SECONDS:
The infected attacker suddenly rushes toward another shopper. The crowd finally realizes something is seriously wrong. Panic spreads naturally. People run in different directions instead of moving together. Some hide behind storefronts while others run toward exits. Shopping bags and personal belongings fall naturally as people flee.
12–16 SECONDS:
The security guard reaches the food court and sees two infected people approaching. He grabs a nearby metal crowd-control barrier for protection. He initially backs away while watching them. One infected suddenly charges. The guard swings the barrier defensively and knocks the attacker off balance. The movement must look heavy, awkward and physically difficult, never like professional choreography.
16–20 SECONDS:
A few shoppers help each other retreat behind a storefront entrance. One person pulls an injured shopper away from the food court. Nobody performs exaggerated combat moves. The guard keeps backing toward the survivors while maintaining distance from the infected.
20–24 SECONDS:
One infected suddenly charges again. The guard sidesteps at the last moment. The infected crashes into the glass railing from its own momentum. The glass cracks realistically but does not explode into an unrealistic shower of fragments. The guard immediately moves the survivors farther away.
24–27 SECONDS:
The survivors move together toward a partially closed storefront. The guard stays behind them and keeps watching the infected. Distant screaming comes from another section of the mall, suggesting the outbreak is larger than what the camera has witnessed without showing an unnecessary crowd of zombies.
27–30 SECONDS:
The group reaches temporary safety behind the storefront. The guard slowly looks through the gap toward the food court. The mall has become eerily quiet. An escalator continues running in the background while abandoned shopping bags, overturned chairs and spilled food remain scattered across the floor. A distant infected figure briefly crosses the far end of the mall. The guard quietly raises one finger, signaling everyone to stay silent.Cut naturally to black.
Action, chases and sports
Seoul subway revenge comedy
Sets the camera language in the first line (handheld, fluorescent, real time), then plays the gag in clear beats and spells out what must stay visible at the end.
Seedance 2.5 · Text-to-video · Source: @AIwithSynthia on X · 10,056 likes on 2026-10-27
action comedy handheld subtitles
Create a live-action cinematic revenge-comedy video set in a Seoul subway station. Use realistic handheld camera movement, cold fluorescent lighting, teal-green accents, and real-time action.
A Korean woman in her 20s with a black bob, sunglasses on her head, a fitted crop top, high-waisted blue jeans, black boots, and a brown shoulder bag descends the subway stairs holding a takeaway coffee. A Korean man with messy dark hair and blonde tips, wearing a dark-green bomber jacket, suddenly hits her from behind and laughs as she stumbles against the handrail. Furious, she chases him onto the platform as he boards a train and mocks her through the closing glass doors.
She punches the glass, creating a dramatic crack and opening. As she confronts him, her coffee cup slips from her hand, falls to the platform, and spills visibly across the floor. She grabs his jacket, forcefully throws him through the air, and sends him crashing into a large metal rubbish bin. The bin rattles and wobbles as he lands inside, looking terrified and humiliated.
She leans toward him and says, "Don't mess around." She then calmly walks away while he remains stuck in the bin.
Maintain identical faces, hairstyles, outfits, and location throughout. Include realistic physics, expressive reactions, cinematic close-ups, subway ambience, and clear white subtitles. Keep the action non-gory, the coffee spill visible, and the final dustbin landing unmistakable.
Basketball player backflip dunk on outdoor court
The prompt splits one action into four numbered shots (dribble, jump, mid-air flip, dunk), each with its own shot size and angle. Low angles and a slow-motion close-up on the flip give Kling's physics a clear beat to simulate.
Kling 3.0 · Text-to-video · Source: @higgsfield_ai on X · 2,533 likes on 2026-10-27
multi-shot sports basketball low-angle physics
Shot 1: A side profile, medium shot of a muscular African American man in athletic wear, intensely dribbling a basketball on an outdoor concrete court towards a hoop against a high white wall under a bright blue sky.
Shot 2: A dynamic, low-angle wide shot showing the man jumping remarkably high into the air, performing a backflip while holding the basketball, with the white wall and sky in the background.
Shot 3: A close-up, slow-motion shot of the man mid-air and upside down during his flip, showing his focused expression and the basketball held firmly in his hands against the sky.
Shot 4: A dramatic, low-angle shot focusing on the basketball hoop as the man, still upside down from his acrobatic jump, powerfully dunks the ball into the net.
Woman slices wooden dummy in night courtyard
Six numbered shots, each with its own framing, move a single scene from a macro mouth detail through wide action to a closing portrait. Naming the camera angle and the physical effect per shot, such as flying wood splinters and dust, gives the model clear impact beats.
Kling 3.0 · Text-to-video · Source: @higgsfield_ai on X · 2,473 likes on 2026-10-27
multi-shot action katana macro night
Shot 1: An extreme close-up of a young woman's mouth, her teeth clenched on a wooden toothpick, creating a tense and gritty atmosphere.
Shot 2: A medium shot of the woman in a traditional Japanese-style courtyard at night, illuminated by soft lanterns. She is dressed in a white tank top and dark trousers, holding a katana in a defensive stance with a determined look on her face.
Shot 3: A dynamic, low-angle action shot of the woman jumping high into the air, raising her katana to strike a wooden training dummy that resembles a samurai in armor.
Shot 4: A close-up, high-speed impact shot showing the woman’s katana slicing through the wooden dummy, sending wood splinters and dust flying into the air.
Shot 5: A fast-paced tracking shot of the woman landing gracefully and immediately striking another wooden structure, demonstrating her speed and precision.
Shot 6: A final medium close-up of the woman’s face as she finishes her move. She is slightly breathless, her hair disheveled, with a sharp, focused gaze directed off-camera.
Two Anime Fighters Brawl on Rooftop
Character reference blocks lock each fighter's outfit, then a style and energy header sets cel shading, impact frames and constant camera motion. A four shot list gives each shot its own camera path and a chain of strikes with no pauses, and an Avoid line bans slow motion and freeze frames.
Seedance 2.5 · Reference-to-video · Source: @aimikoda on X · 1,962 likes on 2026-10-27
anime fight-scene multi-shot camera-movement character-reference
References:
@[char1 ref]: Mina. Long black hair, green striped crop top, baggy orange cargo pants with black graffiti, white sneakers, black fingerless boxing gloves on both hands, chain bracelet.
@[char2 ref]: Rue. Long pink hair, black crop top, black cargo shorts with chains and tags, bare knuckles with silver rings, skin visible on both hands, chunky lime sneakers, dog tag necklace.
Visual style: High-sakuga flat-color 2D anime, bold linework, hard cel shading, flat blue and teal backgrounds that flash hot pink or orange on impact, speed smears, impact frames, deep foreshortening.
Energy: Within every shot the camera keeps moving: it spirals around crossing limbs, dives under kicks, skims the floor, and whips between faces, often following a fist or a sole within inches of the lens. Every movement flows straight into the next strike, block, or counter. Every landed hit plays as sakuga impact: a two-frame impact frame, then the struck body whips away in a long smear, elastic recoil, immediate re-entry into her own attack. Heads snap, torsos fold, feet leave the ground, bodies skid across concrete. Each impact throws a color flash, dust, flying hair and cloth, and radiating speed lines.
Avoid: No subtitles. No slow motion, no idle posing, no pauses between actions, no freeze frames, no held two-shots.
Mina and Rue trade one relentless brawl on an empty concrete rooftop with a chain-link fence at the back and a steel vent box at center right. Mina starts screen left, Rue screen right. Whenever they close, one body is nearer the camera and the other clearly behind it with air between. No dialogue; hits, shoe scuffs, chain rattle, wind. Shot list: 4 shots, about 30 seconds, every shot packed edge to edge with exchanges.
Scene 1: 连续时间,天台。
Shot 1: 大全景起,地面高度,低角度;镜头从水泥地面猛然升起,穿过两名已经腾空扑向对方的格斗者之间,然后贴着Mina的拳套甩进中景,并绕着两人交错的手臂急速环绕。Mina的右勾拳与Rue抬起的膝盖在半空撞上,冲击帧,两人同时被震成拖影向后飞开,滑步落地的瞬间已经再次扑上;Mina左直拳,Rue下潜躲过并打出上勾拳,Mina用前臂接住,前臂被顶上自己的下巴,头向后一甩成拖影,她借着仰头的势头打出一记一二连击,第二拳砸开Rue的防守,冲击帧,Rue的头横向甩出,粉发拉成一道弧,她踉跄一步就从踉跄里踢出正蹬,踢中Mina的拳套,把Mina推得双脚在地上滑出一道痕。快速细节切:Rue眯起的眼睛,Mina收紧的拳套,狗牌甩动,然后镜头钻到Rue旋踢的下方。Rue的脚跟掠过Mina低下的头顶,Mina从内侧起身的身体勾拳把Rue整个人折弯在拳套上,冲击帧,Rue的双脚离地,她顺着折弯的势头甩出肘击,Mina双拳套挡住,拳套却被顶进自己的脸,她向后踉跄两步,Rue追上来一记戴戒指的直拳擦着镜头砸过,Mina侧滑,把一记短右拳埋进Rue的肋部,冲击帧,Rue半边身体塌陷成拖影,双脚在地面打滑,她就从打滑里顶出一膝,砸在拳套上,把Mina的胳膊震到自己胸口。
Shot 2: 跟拍起,位于Rue身后腰部高度;镜头随她冲向铁丝网,沿着网向上扫,在她蹬离时翻转,然后落到地面高度,贴着天台地面疾驰穿过两人相撞的缝隙。Rue推开Mina的胸口转身冲向铁丝网,在网上蹬两步,蹬离,后空翻越过从下方冲进来的Mina,落下的膝盖砸中Mina的肩膀,冲击帧,Mina被砸成一个拖影般的前滚翻,从翻滚里起身时顺势扫向Rue落地的那条腿,Rue跳过扫踢,撑手后空翻,落地时Mina的直拳已经到了,Rue格挡,格挡把她自己的手臂震开,Mina的第二拳穿过空档擦中她的太阳穴,冲击帧,Rue半转过身,粉发横扫镜头。快速细节切:荧光绿鞋底填满镜头,橙色工装布料猛甩,铁丝网剧烈震颤,然后镜头贴地。两人再次相撞:Mina低身滑过Rue的飞跃拳,两人在另一侧旋转成背对背,同一瞬间各踢出一记后蹬,鞋底对撞,冲击帧,两人都被踢得向前扑倒撑地,撑起后同时转身甩出旋转背拳,拳套撞上指节,粉色闪光,两人的头各自甩向一侧成拖影,Rue先回过身勾住Mina的腿,Mina跳脱的同时抡出过顶拳,擦过Rue的太阳穴,把她打得转了半圈,Rue转完这半圈的势头直接变成一记反手肘,砸在Mina的肩上,Mina的肩膀被压得一沉。
Shot 3: 高角度俯冲起;镜头向通风箱俯冲,压低绕到箱侧,然后在两张脸之间左右急甩,每一拳都跟着拳头从一张脸甩到另一张脸。Mina,戴黑拳套,跃上通风箱,膝盖朝下砸落;Rue,赤手戴戒指,侧步闪开,一记侧踢踹进箱体,箱体轰然凹陷;Mina从压扁的边缘跃起打出飞身直拳,Rue后仰躲过,扫掉Mina的支撑手臂,Mina侧手翻落地的同时勾拳打中Rue的下颌,冲击帧,Rue的头甩过去,整个人被带得转了半圈,脚步踉跄,又从踉跄里反扑一记直拳,震得Mina的下巴一晃。快速细节切:弯折的金属,断裂的电缆冒着火花,拳套在箱顶打滑,然后镜头切进两张脸之间。隔着废墟的近身对轰:刺拳,格挡,直拳,侧闪,没有一拳停在半路;Rue的膝盖顶进Mina垂下的拳套,把她推退一步;Mina的头槌砸在Rue的眉骨上,冲击帧,Rue的头向后甩成拖影,她的身体先于意识反击,戴戒指的勾拳打在Mina的拳套上;Mina的反击肘撞上Rue的脸颊,冲击帧,粉发炸开,Rue踉跄两步,膝盖一软又撑住,从撑住的那一下里直拳回敬,打得Mina的头偏过去,Mina的头偏到一半就已经转回来,两人同时下蹲。
Shot 4: 高角度大全景起,倾斜成荷兰角;镜头在两人跃起时螺旋下降,在最高点从两人之间穿过,再翻到两人侧面,并在两人上升时继续旋转。两人同时向上爆发,拳头在前,Mina的黑拳套与Rue戴戒指的拳头冲向画面中央同一点,头发向后飞扬成拖影,速度线向中心汇聚,拳头进入触及范围的瞬间背景炸成全屏亮粉色;拳头仍在逼近,头发仍在甩动,镜头仍在旋转,画面在此切断。
Sci-fi and fantasy
Lab containment breach, sci-fi horror
Asks for an immediate hook and grounded practical effects, which keeps the creature from looking like CGI.
Seedance 2.5 · Text-to-video · Source: @auqibhabib on X · 2,425 likes on 2026-10-27
Cinematic sci-fi horror-action short film set in a research laboratory, opens with an immediate strong hook, natural realistic pacing, grounded practical-effects style. Lab filled with steel equipment, glass containment tubes, monitors, specimen tables, harsh fluorescent lighting. Five scientists in lab coats work at various stations. Natural handheld camera work, muted authentic color grading, all sound fully natural and diegetic — no music, no artificial sound design until the tension score kicks in.
>
> **[0-1s]** Hard cut: a containment tube shatters violently, glass exploding across the lab floor.
>
> **[1-2s]** A small dark parasite creature skitters out fast, latching instantly onto a scientist's neck.
>
> **[2-3s]** He screams, collapsing to his knees, the others recoiling in shock.
>
> **[3-4s]** Dark veins pulse rapidly beneath his skin, visible even through his collar, his body convulsing.
>
> **[4-5s]** His eyes turn a deep obsidian black — he rises, movements jerky and unnatural.
>
> **[5-6s]** He lunges at the nearest colleague, parasite tendrils extending from his mouth toward her.
>
> **[6-7s]** She's caught, tendrils latching onto her neck, her scream cutting short.
>
> **[7-8s]** Veins spread fast across her too, body convulsing against a lab table, equipment crashing down.
>
> **[8-9s]** She rises seconds later, eyes black, joining him, both turning toward the remaining three.
>
> **[9-10s]** The three scientists scramble back, one grabbing a fire axe from an emergency case on the wall.
>
> **[10-11s]** The two infected charge fast, jerky inhuman movements, tendrils whipping from their jaws.
>
> **[11-12s]** The lead scientist swings the axe hard into the first infected — a solid impact, it staggers into a specimen table, equipment shattering.
>
> **[12-13s]** A second scientist grabs a fire extinguisher, blasting the other infected in the face, it recoils, tendrils flailing.
>
> **[13-14s]** The third scientist shoves a steel cart into the staggering infected's path, sending it crashing down.
>
> **[14-15s]** Both infected collapse motionless, twitching, tendrils retracting slowly.
>
> **[15-16s]** The three survivors regroup, breathing hard, scanning the wrecked lab for more movement.
>
> **[16-17s]** Wide shot: broken glass and toppled equipment everywhere, alarm lights beginning to flash red.
>
> **[17-18s]** A third containment tube behind them starts cracking, unnoticed by the group.
>
> **[18-19s]** It shatters suddenly — another parasite creature launches out fast.
>
> **[19-20s]** The lead scientist spins just in time, swinging the axe and slicing it out of the air mid-leap.
>
> **[20-21s]** It hits the floor, motionless, dark fluid spreading slightly beneath it.
>
> **[21-23s]** The group presses their backs together, weapons raised, scanning every corner of the lab.
>
> **[23-25s]** Emergency lights flash steadily red, alarm klaxon blaring through the facility.
>
> **[25-27s]** The lead scientist gestures toward the exit; they move together cautiously, stepping over debris and fallen colleagues.
>
> **[27-29s]** They reach the heavy lab door, the lead scientist sealing it shut behind them with a hard mechanical clang.
>
> **[29-30s]** Final shot: the three survivors lean against the sealed door, breathing hard, alarm lights flashing red across their exhausted faces. Cut to black.
>
> Natural diegetic sound throughout — glass shattering, guttural convulsions, screaming, metal impacts, alarm klaxon, heavy breathing — layered with a tense instrumental score building through the fight and easing at the end. No text overlay, no watermark. Natural handheld camera with realistic shake during action, muted clinical color grading.
A woman teaches a dragon to bake heart-shaped bread
Short and specific: colours of the dragon, one hard rule (all bread heart-shaped), and a mood word to end on.
Seedance 2.5 · Text-to-video · Source: @ElsaSofia__AI on X · 1,430 likes on 2026-10-27
Photorealistic medieval fantasy film. Warm sunlight, stone cottage, wooden table, iron grill. Colossal dragon with pale pink-white scales, pink spines and golden eyes. All bread must be heart-shaped. Realistic fire, food textures, natural physics.
0-5s: Heroine arranges raw heart-shaped dough on the grill. Dragon sniffs eagerly, then blasts enormous flames! Fire engulfs the dough. She jumps back: "Whoa! Easy! You're making bread, not dragon food!"
5-10s: A charred heart-shaped loaf smokes on the grill. Dragon looks guilty, hiding its head. She shows the burnt bread, then smiles and replaces it: "Look. Just a little flame. Can you do that?"
10-17s: Dragon carefully releases a tiny steady flame beneath the grill. Heart-shaped dough rises, forming a golden crust. Steam rises as dragon watches nervously.
17-23s: She cuts open the crispy heart-shaped bread, revealing a soft fluffy center. "You did it! Perfect!" Dragon tries to eat it all. She laughs and feeds it a small piece.
23-30s: Dragon nuzzles her for affection. She strokes its scales, shares more heart-shaped bread. Dragon closes its eyes happily, tail wagging. Wide sunset shot, warm and healing.
Cinematic lighting, detailed scales, realistic flames, expressive acting. No subtitles, narration, logos or watermark.
Celestial puppeteer rebuilds a collapsing golden star
A fixed painterly style block with explicit negatives opens and closes the prompt, then WORLD and CHARACTER blocks lock the look. Ten timed shots of three seconds each, with the shot type named in every cut, give the 30 second piece a clear arc from macro detail to hero frame. The character sheet made with GPT Image 2.5 serves as the identity reference.
Seedance 2.5 · Reference-to-video · Source: @itsPixieVerse on X · 1,127 likes on 2026-10-27
painterly-animation multi-shot character-sheet sci-fi 30-seconds
Cinematic 2.5D animation in painterly rendering, gouache concept art in motion, visible brush textures, flat posterized color blocks, matte finish, soft volumetric lighting. NOT flat 2D cartoon, NO outlines, NOT glossy CGI, NO photorealism.
WORLD: Shattered zero-G celestial observatory overlooking a collapsing golden planetary core, palette of obsidian black, ivory stone, and blinding gold energy, floating cosmic debris.
CHARACTER: Extremely tall adult woman in early 20s, black visor-hat with glowing gold glyphs, floating orbital halo, black prosthetic arms with trailing strings. Massive ballooning obsidian cloak, cream draped gown, platform heels. Elongated tall proportions, supermodel legs, painterly matte skin.
CINEMATIC SHOT DESIGN (10 Filmic Cuts across 30 Seconds):
• SHOT 1 (0–3s | Macro Detail): Macro close-up of Ophelia's black prosthetic fingers twitching; glowing gold gravitational threads unspool into the vacuum.
• SHOT 2 (3–6s | Scale & Threat): Wide shot of the dying planetary core cracking in half, blasting shockwaves of golden light through floating ivory pillars.
• SHOT 3 (6–9s | Character Entrance): Low-angle tracking shot as Ophelia glides weightlessly into the abyss, her massive obsidian cloak ballooning around her like a cosmic void.
• SHOT 4 (9–12s | Expression Focus): Close-up on her face under the visor; the gold glyphs flare brightly as she tilts her head in absolute, chilling composure.
• SHOT 5 (12–15s | Casting the Strings): Medium shot; Ophelia sweeps both mechanical hands outward, casting hundreds of glowing gold puppet strings across the entire starfield.
• SHOT 6 (15–18s | Reality Pull): Dynamic wide shot; she violently pulls the strings taut, arresting the collapse of the planet and dragging the colossal burning fragments back together.
• SHOT 7 (18–21s | High-Speed Tracking): FPV tracking shot swooping past floating debris as the planetary fragments collide and fuse into a newly ignited miniature golden sun.
• SHOT 8 (21–24s | Power Apex): Medium shot; Ophelia rotates her mechanical wrists, the floating orbital halo around her head spinning rapidly and locking the new star into orbit behind her.
• SHOT 9 (24–27s | Aftermath Echo): Extreme wide shot; the restored golden star illuminates the massive expanse, Ophelia silhouetted dead-center as the master of the cosmos.
• SHOT 10 (27–30s | Hero Poster Frame): Low-angle hero shot; Ophelia lowers her hands, gravity strings dissolving into gold dust, her massive black mantle settling around her towering silhouette. Hold final frame.
Weighty natural animation, zero-G fabric motion, 24fps filmic motion, shallow depth of field. NOT flat 2D cartoon, no bold outlines, not glossy CGI, no photorealism.
Woman and Pet Dragon Cook a Roast
A second-by-second shot list from wide shot to extreme close-up builds a full comedy arc in 30 seconds, from the dragon's fire blast to a gentle forgiveness beat. Character sheets for both leads, a setting list and a separate audio line for fire, steam and dragon sounds keep the look and sound consistent.
Seedance 2.5 · Text-to-video · Source: @Alina_with_Ai on X · 1,043 likes on 2026-10-27
fantasy dragon short-film multi-shot audio
Cinematic 16:9 fantasy short film, 30 seconds, 30fps, photorealistic, warm golden-hour light, rich film color grading, shallow depth of field, smooth camera, natural motion blur, heartwarming comedy tone.
Characters:
1) A young woman with messy wavy blonde hair, thin round glasses, a linen cream blouse with rolled sleeves and a dark brown pinafore apron dress. Expressive, funny, warm face.
2) Her pet baby dragon: a huge, friendly, horse-sized dragon with white-cream scales covered in pink-magenta patterns, a crest of soft pink spikes, big amber eyes with vertical pupils, a gentle snout. It acts like a naughty, affectionate puppy.
Setting: a rustic medieval cottage courtyard and kitchen: a stone oven, a wooden table, hanging lavender and herbs, stacked firewood, a low stone wall, green hills in the background. Later, the warm inside of the cottage with a wooden table, a candle, pots on shelves, and sunbeams through a small window.
0-3s: Wide shot, courtyard. The woman turns a big pink raw pork roast on a metal spit over the table. The dragon's head peeks over the stone wall on the right, watching curiously. She talks to the dragon, laughing, asking it for just "a little fire".
3-5s: Extreme close-up of the dragon's scaly snout and nostrils, then a close-up of her smiling face, laughing and confident.
5-7.5s: The dragon opens its mouth and unleashes a huge blast of fire at the roast, flames and sparks everywhere. She throws her hands up in panic.
7.5-10s: Extreme close-ups of her shocked face, wide eyes, hair blown by the heat, hands raised, shouting "No, no, no!". Cut to the roast, completely burnt black and surrounded by flames.
10-12s: She grabs a wooden bucket and throws water on the fire, a giant cloud of steam rises. The dragon laughs with its mouth open, tongue out.
12-15s: Steam clears. She stands with her arms out, scolding the dragon like a child. The dragon lowers its head with guilty puppy eyes, and she gently touches its snout, forgiving it.
15-18s: Cut to the cozy kitchen. She wipes her forehead, then cuts a small piece of fresh meat and puts it on a tiny skewer, calmly explaining to the dragon how little fire is needed. The dragon leans in beside her.
18-21s: Close-up: she holds the skewered meat cube. The dragon carefully breathes a tiny, gentle flame, and the meat sizzles with small sparks. She smiles proudly.
21-23s: Extreme close-up: the perfectly seared meat cube on the skewer, in front of the dragon's amber eye, a reflection of tiny fire in its vertical pupil.
23-26s: She sits at the table with a plate of cooked meat, cuts a piece, tastes it with a delighted face and gives a thumbs-up. She feeds a piece to the dragon, who opens its mouth happily.
26-30s: Wide, warm shot of the cottage interior, golden sunbeam through the window, a candle on the table. She eats while the dragon rests its big head on the table next to her, its eyes slowly closing in contentment, its body curled behind. Peaceful, heartwarming final frame, hold.
Style: Pixar-quality emotional storytelling with photorealistic detail, warm golden and orange tones, soft volumetric light, steam and sparks, cinematic grading.
Audio: gentle fantasy folk music (soft strings and flute), crackling fire and a whoosh on the fire blasts, steam hiss, a deep playful dragon rumble and purr, her laughter and short dialogue, light kitchen sounds.
Anime and animation
Phone camera sketch of friends and a bedsheet ghost
Separate reference images lock each girl's hair and outfit, while the style reference supplies only ink and palette so its pose is not copied. Timed shots with explicit cut and whip pan points keep the handheld phone camera motion readable across eight beats.
Seedance 2.5 · Reference-to-video · Source: @aimikoda on X · 19,513 likes on 2026-11-03
anime comedy hand-drawn multi-shot
References:
@[ref1]: Yuki. Shoulder-length silver-white hair with long bangs, olive-green eyes, small round silver earrings, loose pale camisole, flat deadpan face.
@[ref2]: Sumi. Short black bob with uneven bangs, brown eyes, small gold earrings, cream camisole under an oversized off-white shirt, loud and easily flustered.
@[ref3]: the style reference. Hand-drawn look, line quality and palette only; not its pose or layout.
Visual style: Rough hand-drawn anime sketch, loose black ink lines, flat watercolor washes, muted grey, sage and beige palette, soft indoor daylight. Shot as a phone front camera video: handheld, restless, constantly bobbing and tilting with whoever holds it, quick crash zooms and whip pans.
Avoid: No subtitles. No background music. No slow motion. No other people. Yuki and Sumi never merge or swap faces.
Two friends goof around nonstop with a phone camera through a small apartment, each new spot in front of a different strange painting. Every change of place comes through a whip pan or a cut from the phone itself; nobody stands still. A visible gap always separates their faces. Sound: footsteps, phone rustle, shrieks, laughter, one line. Shot list: 8 shots, about 30 seconds.
Scene 1: Continuous, in front of an orange painting with two huge round owl eyes.
Shot 1 (0-4s): Close two-shot, high angle; the camera bobs quickly with Sumi, who holds the phone with her raised right arm from the top right corner. Yuki springs up into frame from below on screen left and flashes a V sign with both hands, face completely deadpan. Sumi sticks out her tongue and shakes her head side to side, her black bob flying. Cut as Yuki snatches the phone from Sumi's hand; the camera whips left in a blur.
Scene 2: Continuous, in front of a blue-grey blob painting with one black hole for an eye.
Shot 2 (4-8s): Medium shot, eye level; the whip pan lands here, and the camera, now in Yuki's right hand, quickly pans right to follow Sumi. Yuki stays at the bottom left edge of frame, deadpan. Sumi skids across the floor from screen left to screen right in her socks, arms flailing. Cut as Sumi spins and lunges straight at the lens.
Shot 3 (8-11s): Close-up, eye level; the camera, still in Yuki's hand, does a fast crash zoom onto Sumi. Sumi's face is pressed almost against the lens, mouth wide open. Behind her, Yuki does a stiff robot dance with her free arm, still deadpan. Sumi bumps the phone; it tumbles out of Yuki's hand. Cut the instant the phone hits the floor.
Scene 3: Later, in front of a dark green painting of a faded face.
Shot 4 (11-15s): Extreme low angle from the floor; the phone already lies face up, locked-off camera looking at the ceiling. Both girls lean in over the lens from opposite sides, Yuki on screen left, Sumi on screen right. Yuki flips her silver hair forward over her whole face; one big eye peers through a gap. Sumi shrieks and jumps back out of frame, then her hand dives in and grabs the phone. Cut as the phone lifts off the floor.
Scene 4: Continuous, a narrow hallway with a grey wall and a brown smeared painting.
Shot 5 (15-19s): Medium close-up, Sumi holding the phone out in front of her and facing it while she runs backward down the hallway, still running flat out, laughing. Far behind her, at center, Yuki glides after her with her hair still over her face, arms limp, closing the distance. Cut as Sumi swings around a corner; the camera whips right.
Scene 5: Continuous, in front of a beige wall with a round orange painting.
Shot 6 (19-22s): Medium close-up, eye level, handheld; the camera heaves with Sumi's breathing. Sumi, holding the phone in her right hand, presses her back to the wall, panting, eyes darting left and right. Yuki is nowhere in frame. Cut as a white cloth drops in from the top of frame.
Shot 7 (22-25s): Insert, close to the lens; the camera jolts once. A single white bedsheet with two black eye holes lowers from above in front of the lens, held by Yuki's hands.
Yuki's line (flat, deadpan): "Boo."
Shot 8 (25-30s): Wide two-shot, high angle; the camera quickly pulls back, then spins with them. Sumi, phone in her right hand, yanks the sheet with her left; the sheet tangles around both girls and they spin together, laughing, Yuki's hair now neat. Both thrust V signs at the lens. Sumi's index finger is still reaching for the lens when the clip cuts to black as the fingertip covers it.
Katsudon cooked in fast-cut anime style
Opens with an explicit ban on imitating any existing anime, then walks through the recipe step by step so every cut has a job.
Seedance 2.5 · Text-to-video · Source: @Goodmanprotocol on X · 4,935 likes on 2026-10-27
Create a 30-second fast-paced cinematic Japanese anime cooking video showing the preparation of authentic katsudon, entirely from the text description below.
IMPORTANT: Do not display, recreate, trace, reference, or imitate any storyboard, sketch, reference image, panel, annotation, or source material. Generate only original anime-style animation.
STYLE
High-quality Japanese anime film style, cinematic summer lighting, ultra-detailed food textures, realistic cooking physics, visible steam and moisture, shallow depth of field, macro close-ups, smooth camera movement, warm Japanese kitchen atmosphere. Fast rhythmic editing with natural match cuts based on movement, shape, texture, and composition.
TIMELINE
0–2.5s — Prepare Pork
Close-up of hands pounding a thick pork loin with a wooden mallet on a cutting board. The meat gradually flattens and its fibers loosen. Sprinkle salt and pepper evenly. Crisp rhythmic impacts.
2.5–5s — Bread the Pork
Quick match cuts: press the pork into flour, dip into beaten egg, then firmly coat with coarse panko breadcrumbs. Show realistic egg dripping and breadcrumbs adhering to the surface.
5–7.5s — Fry
Lower the breaded pork into hot golden oil. Intense bubbling surrounds the cutlet as the crust turns golden brown and crispy. Warm cinematic lighting and detailed oil movement.
7.5–9.5s — Slice
Place the fried tonkatsu on a wooden board. A sharp knife cuts it into even strips. The crispy crust cracks naturally, revealing juicy white pork with gentle steam escaping.
9.5–12s — Simmer Dashi & Onion
Thinly sliced onions gently simmer in amber dashi inside a shallow pan. Small bubbles rise while the onions become translucent. Chopsticks gently move them through the broth.
12–14s — Add Tonkatsu
Place the sliced tonkatsu carefully over the simmering onions using chopsticks. The broth begins soaking into the crispy edges while steam rises.
14–16.5s — Pour Egg
Slowly pour beaten golden egg over the tonkatsu and onions. The egg spreads naturally through the gaps and begins setting around the hot edges.
16.5–19s — Set the Egg
Close-up of the egg gently coagulating over low heat. Edges become soft and golden while the center remains glossy, slightly runny, and trembling. No stirring.
19–21s — Prepare Rice
A white ceramic donburi bowl receives a generous mound of steaming Japanese short-grain rice. The rice looks fluffy and slightly glossy.
21–24s — Assemble
Carefully slide the egg-and-tonkatsu mixture from the pan onto the rice. The egg settles naturally in a soft wave while dashi lightly absorbs into the rice.
24–26s — Final Close-Up
The finished katsudon rests on the counter. Glossy egg slowly settles over the tonkatsu and rice, gently trembling. Slow cinematic push-in.
26–30s — Hero Reveal
Present the finished katsudon in a traditional blue-and-white ceramic bowl on a wooden surface. Golden egg covers crispy tonkatsu over glossy rice, topped with fresh mitsuba. Steam rises as the camera slowly arcs around the bowl for a beautiful anime-food-film ending.
AUDIO
Bright 1980s-inspired Japanese city-pop instrumental, 110–120 BPM, with subtle koto and light chime percussion. Synchronize realistic ASMR cooking sounds: mallet impacts, knife slicing, frying sizzle, bubbling dashi, chopsticks, egg pouring, steam, and a soft ceramic clink.
End with one delicate wind-chime tone during the final reveal.
NEGATIVE
No storyboard, reference image, sketch, panels, borders, numbers, arrows, annotations, subtitles, captions, UI, logos, or text overlays. No unrelated ingredients or dishes. Katsudon only. Keep the food, hands, utensils, lighting, and environme