Dean Nguyễn

AI Prompt Engineer at Lumio Studio
the drawer of dreams
AI Filmmaker

[Dean Nguyễn @hainth1402]

Interrupting her bath was their first mistake. 🛁⚔️

Saw this trend on X and had to try my own reprompt.

#AIVideo #AIFilmmaking

[Dean Nguyễn @hainth1402]

They thought they had her cornered in the courtyard. Chun-Li had other plans.

Real martial arts choreography meets glowing blue Kikoken energy VFX. This is how Street Fighter live-action should feel! ⚡🔥

#ChunLi #StreetFighter #LiveActionVFX #MartialArtsCinema #SpinningBirdKick #CapCut #seedance25

[Dean Nguyễn @hainth1402]

He brought fire. She brought red boots and absolutely no concern. 😏🔥

TRY ME — an AI action short by Dean Nguyen.

#AIFilm #martialarts

[Dean Nguyễn @hainth1402]

A katana. An iron club. A bridge that won’t survive.

AYAME ⚔️
An AI-crafted anime fight built around speed, weight, and counters that connect.
Ayame vs. Akane.
Created by Dean Nguyen.

#AnimeAction #AIFilmmaking

[Dean Nguyễn @hainth1402]

I turned the Tesla Roadster into an absolute menace. 🚀

Flying over traffic. Driving on walls. Breaking every law of physics.

All made with AI.

@Tesla @elonmusk — how many of these features can we get? 👀

#TeslaRoadster #AIFilm #Tesla

[Dean Nguyễn @hainth1402]

TALES OF DREAD — EP6: THE SHROUD

A little clay nightmare for your evening. 🕯️

Made with WAN 3.0.
Sponsored by RUNWAY.
#TalesOfDread #WAN30 #Runway

[Dean Nguyễn @hainth1402]

TALES OF DREAD - EP5: THE LOTUS SPIRIT 🪷

He finally saw the world. But never her face.

A bittersweet Chinese folklore romance in 2D animation style. Made with Seedance 2.5.

#TalesOfDread #Seedance25 #AIAnimation

[Dean Nguyễn @hainth1402]

Your night shift starts here. 📹

TALES OF DREAD — EP4: CAM 13
Made with Seedance 2.5.
#TalesOfDread #CAM13 #Seedance25

[Dean Nguyễn @hainth1402]

TALES OF DREAD — EP3: THE COUNT 🦇

Something has awakened.
Watch with headphones.
Made with Seedance 2.5.

#TalesOfDread #AIShortFilm #Seedance25

[Dean Nguyễn @hainth1402]

TALES OF DREAD - EP2 : THE RED DRESS

She doesn’t belong here. 👾🔴
A short horror experiment blending pixel art and live action.
Made with Seedance 2.5.

#TalesOfDread #AI filmmaking #Seedance

[Dean Nguyễn @hainth1402]

A quiet inn. An unsettling guest. 🕯️
TALES OF DREAD — EP01: THE GUEST
Sound on. Watch till the end.
Made with Seedance 2.5 · By Deannguyen
#Seedance25 #AIHorror #AIShortFilm

[Dean Nguyễn @hainth1402]

I was testing Minimax H3. She was shooting her comeback video. 🪩

Silver hair, red shades, and a neon world built around her.

My latest AI music video experiment—what’s the first shot that gives the AI away?

#MinimaxH3 #AIFilmmaking

[Dean Nguyễn @hainth1402]

https://t.co/J1WtwznJ41

GGdocs link

[Dean Nguyễn @hainth1402]

She didn’t stop fighting.
She simply kept moving forward. ⚔️🔥
This is what Seedance 2.5 looks like when live-action realism, camera movement and combat physics come together.

Prompt in the replies.
#Seedance25 #AIFilm #CinematicAI

[Dean Nguyễn @hainth1402]

Same camera. Same choreography. Same edit. New cast and location. 🎬

I built a Seedance 2.5 master prompt to recreate a reference video shot by shot while replacing every character and the setting with image references.

Full prompt in the replies 👇

[Dean Nguyễn @hainth1402]

SEEDANCE 2.5 — 1:1 VIDEO CLONE WITH CHARACTER AND LOCATION REPLACEMENT

INPUTS
@Video 1 = the source video to reproduce.
@Image 1 = replacement main character.
@Image 2 = replacement supporting cast. Identify each person separately; use the supplied role map if available.
@Image 3 = replacement location, if supplied.
ROLE MAP: [Optional: original performer/position → person in @Image 2]
CRITICAL SHOTS: [Optional: timecode → performer, visible clothing/shoes and location surface]

SINGLE OBJECTIVE
Produce a time-synchronized, shot-for-shot reconstruction of @Video 1. At every output timestamp t, match the source video at timestamp t. Change ONLY (1) the identities and complete appearances of the people according to @Image 1 and @Image 2, and (2) the location according to @Image 3 if supplied. Everything else follows @Video 1. Do not reinterpret, improve, restage or add to the source video.

EXACT VIDEO CLONE
Preserve the source duration, shot count, cut timestamps, shot order, action timing, pauses, body poses, movement paths, speed, left/right direction, hand use, prop and weapon paths, interactions, impacts, falls, particles and final pose.

For EACH shot, match the source camera’s position, height, angle, distance, shot size, framing, subject placement, horizon, movement path, direction, speed, acceleration and stopping point. Preserve its apparent lens perspective, focus, depth of field, foreground occlusion, parallax, motion blur and transition into the next shot. A low-angle shot remains low angle; an overhead remains overhead; a close-up remains the same close-up. No alternate coverage, new camera move or changed edit rhythm.

Preserve the source’s lighting direction, exposure, contrast, colors, haze, photographic texture and visual intensity. Do not apply a new “cinematic” style, grading, slow motion, lens flare or camera shake. Its existing cinematography is the exact target.

COMPLETE HUMAN REPLACEMENT
Treat every person in @Video 1 as a MOTION TRACK, never as a visual identity.

The original lead track is performed only by @Image 1’s person. Copy that person’s exact face, facial structure, skin, hair, build, full outfit, accessories, hands and shoes into every frame of the track.

Assign every original supporting-person track to one specific person from @Image 2. Keep the assignment fixed throughout the video. Reproduce that referenced person’s exact face, hair, build, full outfit, accessories, clothing hem and shoes in every appearance. Replace ALL original supporting people, including distant figures, blurred figures, figures seen from behind, partial bodies at frame edges and fallen bodies. No original person or piece of original clothing may remain.

Identity remains locked when the face is invisible. In foot shots, back views, overhead shots and close-ups of hands, the visible shoes, legs, hems, sleeves, hair and hands must belong to the person assigned to that motion track. Never let the original video’s costume, hairstyle or footwear reappear during fast movement or at a cut.

LOCATION REPLACEMENT
If @Image 3 is supplied, replace the source location in EVERY shot with its exact architecture, ground, materials and environmental details. Keep each shot’s original camera framing and spatial depth while fitting the referenced location into it. All feet, shadows, impacts, falls and particles must interact with the new surfaces. No fragment of the original location remains. If @Image 3 is absent, keep the source location.

PER-CUT CHECK
At every cut, match the source frame first: camera, composition, performer positions and action phase. Then apply the assigned people and location. Maintain the same person on the same motion track across the cut. Do not reset their identity, wardrobe, position or prop continuity.

END
Reproduce @Video 1’s final action, camera position and last-frame composition at the same timestamp. Only the referenced people and, if provided, the referenced location are different.

AUDIO
Keep @Video 1’s audio and synchronization unless replacement audio is explicitly supplied. No new dialogue, captions, text overlays, titles or watermarks.

ABSOLUTE EXCLUSIONS
No original-video faces, bodies, hair, costumes or shoes. No original background when @Image 3 is supplied. No hybrid identities, invented cast, extra shots, missing shots, left/right reversal, altered camera angle, different choreography or new ending.

[五芒星 @T_st_love_love]

AI FUSION FESで現在開催中のイラコン、フォトリアルも増えてきて嬉しいです♡

参加者の皆様へ感謝のキッスムービーやで。

イラコン締め切りは9月30日23:59までとなっております！
まだ入力していない人もこれから参加する人もGoogleフォームへの入力もよろしくねん。

[Tom𝕎 @Tomw852]

'' Arcade Machine ''
- 15s video made with #MiniMaxH3

Running locally on a 5070 12GB in #ComfyUI
(Full prompt in the replies ⤵️)

#AIvideo

[Dean Nguyễn @hainth1402]

nahhh Genjutsu actually cooked on this one 😭