Hailuo 2.3 is MiniMax’s text-to-video and image-to-video model for character motion, subtle expressions, and stylized animation. Describe a scene from scratch, or add a starting image to guide the opening. Choose a 6- or 10-second clip.
Explore text-to-video and image-to-video results with their complete prompts. Reference images are shown for the examples that used them.
Text-to-video · Text prompt only
Text-to-Video: A Quiet Change of Expression
Created from text without a reference image: a woman looks up from a note and gradually smiles as the camera moves closer. Watch the change in her gaze and expression.
A young adult woman stands beside a cafe window in soft morning light, wearing a plain blue sweater. Medium close-up, realistic live-action style. She looks down at a small paper note, raises her eyes toward the camera, and gradually gives a warm, restrained smile. Her shoulders relax slightly. The camera makes a slow gentle push-in. One continuous shot, natural facial movement, stable clothing and background. No on-screen text, subtitles, logos, or watermark.
A live-action woman and cartoon counterpart start in matching outfits, each holding a drink. The prompt directs reactions and walking; the clip shows the cartoon character turning and the pair moving out of the frame.
The main character winks at the camera, smiles, clinks glasses, while the cartoon character stands in the opposite direction to the main character, pretending to be angry. The scene transitions, and they engage in a happy interaction, walking together towards the left side of the frame.
Starting from a fox wearing a scarf, the clip combines falling snow with a head turn, a sniffing motion, and a lowered head. Watch how small character movements work alongside the moving environment.
Snowflakes gently fall around the fox, it sniffs the air, then nuzzles deeper into the scarf with a content expression. Cinematic realistic style, gentle natural movement.
The starting image combines a person, toys, and a retro television. The camera moves toward the screen as colorful mechanical figures and a neon city appear, shifting the scene from a portrait into surreal animation.
A retro television displaying an alternate reality. The camera zooms directly into the TV screen. Suddenly all the toy monsters come alive and begin dancing chaotically. The Ultraman figure grows into a giant mechanical guardian. The screen glitches violently as multiple versions of the character emerge from different dimensions. Toys leap into the air and orbit around the floating television. The room transforms into a neon cyber-toy city. Reality folds and repeats recursively, creating an infinite screen-within-screen effect. Hyper dynamic camera movement, playful chaos, retro-futuristic aesthetics, cinematic toy universe, surreal storytelling, highly detailed animation.camera movement highly dynamic, cinematic motion blur, volumetric lighting, depth of field, realistic physics, ultra detailed, luxury fashion commercial, dramatic visual storytelling, 4K, 8K, blockbuster VFX, seamless animation
Hailuo 2.3, officially MiniMax-Hailuo-2.3, supports text-to-video and image-to-video generation. MiniMax highlights body motion, character micro-expressions, stylization, and object movement. Describe a character performance, illustrated animation, or creative scene from scratch, or guide its opening subject and composition with an image.
Character Motion and Subtle Expressions
Build a character clip around a turn, walk, head movement, or expression. The live-action/cartoon example shows coordinated movement, while the fox uses smaller head motions. Review faces, limbs, and clothing continuity in the result.
Stylized Animation and Object Motion
MiniMax lists anime, illustration, ink wash, and game CG among the supported styles. Describe the look in text or guide it with a starting image, then state what should move. The TV and energy-pulse examples show surreal effects and character transformations.
WORKFLOWS
Create a Scene from Text or Bring an Image to Life
Hailuo 2.3 supports text-to-video and image-to-video. Start with a scene description, or upload an image as the opening frame.
Start with Text
Create a new scene by describing its subject, setting, style, and action. Add camera movement to guide the shot.
Start with an Image
Upload one image as the first frame, then describe how the subject, objects, or camera should move. Use this mode to animate an existing photo or illustration.
PROMPT GUIDE
How to Direct Hailuo 2.3 with a Prompt
Read the complete prompts alongside their results: a text-only watercolor scene, outfit changes, an effect-triggered transformation, and a cooking action sequence.
This prompt describes a watercolor forest stream, a red paper boat, its route past a mossy stone, and a tracking camera. The result shows the boat drifting with ripples across textured watercolor scenery.
A small red paper boat floats along a shallow forest stream at sunrise. Hand-painted watercolor animation with visible paper texture, soft green foliage, and warm golden light. The boat drifts past a moss-covered stone, turns gently with the current, and continues toward the lower right. Small ripples spread across the water while leaves sway lightly in the background. The camera tracks slowly alongside the boat in one continuous shot. No people, lettering, logos, or watermark.
Name the Outfit Changes and the Stable Subject
The prompt names a standing model, then lists suit, streetwear, couture, and sportswear as transformation targets. The video cycles through outfits. When adapting it, describe the person’s motion separately from the clothing changes and check both subject and garment details.
same fashion model standing still, outfit rapidly transforming between luxury suit, streetwear, couture dress, sportswear, seamless fashion transition, premium editorial photography, clean studio background, luxury fashion campaign, ultra realistic, vogue style, dramatic fashion transformation
Link the Energy Pulse to the Transformation
The expanding pulse triggers luminous clothing details and cooler lighting in the prompt. The result shows a glowing ring and a character transformation. Specify the trigger, the details to change, and the features to retain as separate parts of the description.
A circular energy pulse expands from the center of the frame. Where the pulse touches, the character's appearance upgrades to sci-fi style — clothing gains luminous circuit lines, lighting becomes cooler and sharper. The pulse fades outward leaving the transformed portrait. No facial changes, only texture and lighting shift.
Describe the Cooking Actions in Order
The full prompt divides preparation, adding ingredients, and stir-frying into time ranges. The video shows an anthropomorphic dog chef moving its hands and tossing the wok. Treat the timestamps as directions for the intended sequence and check the delivered order and pacing.
0-2 seconds (Opening): Close-up quick push-in, wearing Chinese-style clothing standing in front of the stove, sinking qi to the dantian pressing the table with one hand, knuckles exerting force to shake the diced chicken, peanuts, and dried chili peppers in the bowl to fly into the air simultaneously, ingredients suspended in mid-air.
2-4 seconds (Starting stance): Spinning body to initiate the stance, right hand like a sword finger quickly pointing at the diced chicken in the air, finger wind accurately slicing through the meat fibers, diced chicken instantly cut into uniform small pieces, falling into the hot oil iron wok in an arc in mid-air.
4-6 seconds (Palm wind stir-frying): Sinking waist into a horse stance, both palms exerting energy quickly towards the iron wok from a distance, palm wind enveloping the diced chicken to rotate at high speed in the wok, stir-frying until it changes color, the meat is tender and not dry, evenly coated with glossy oil.
6-8 seconds (Flying cutting ingredients): Side kick, toes lightly picking up cucumbers and carrots on the chopping board, fruits and vegetables flying in the air, raising hand with a fast knife slashing like lightning, instantly cut into cubes, accurately falling into the wok to join the diced chicken.
8-10 seconds (Stir-frying kung fu): Holding the wok handle with one hand, exerting force with the wrist to perform the ultimate wok-tossing skill, the iron wok flying up and down, diced chicken, peanuts, and chili peppers intertwining and flying in the air, not spilling a single drop of broth, the movement is as chic as sword dancing, full of wok hei.The final large-arc spinning wok, drawing an afterimage in the air, ingredients steadily fall into the wok, accurately plating the bright red, fragrant and crispy Kung Pao Chicken into a white porcelain plate, the finished dish steaming, the camera freezes on a delicious close-up with a blurred background of a fist-clasping figure, the picture should be smooth, following the plot, must have close-ups.
HOW TO USE
Create Your First Hailuo 2.3 Video
Enter a video description, optionally add a starting image, and choose the duration and quantity. Review the points before generating, then open the result workspace to view the video.
Describe the Scene and Add an Image If Needed
For text-to-video, describe the subject, scene, action, and camera. For image-to-video, add a starting frame and describe the subsequent action, expression, object movement, or camera direction.
Choose the Duration and Review Points
Choose 6 or 10 seconds and the quantity, then review the current points estimate and your account’s model access. This form has no resolution selector; check the downloaded original for its actual dimensions and duration.
Open and Check the Result
Open the generated result and watch the entire video. Check faces, hands, limbs, clothing, object shapes, and camera movement, especially during transformations or a sequence of actions.
Choose standard Hailuo 2.3 to build a scene from text or use a starting image for character motion, subtle expressions, and stylized animation. Review Hailuo 02 for opening-and-ending-frame control, or Hailuo 2.3 Fast for the family’s efficiency-oriented image-to-video variant.
Dimension
Hailuo 02
Hailuo 2.3
Hailuo 2.3 Fast
Model role
Physical movement and instruction-led scenes
Character expressions, physical actions, and stylization
Efficiency-oriented image-to-video variant
WeShop input workflow
Animate an Opening Image; Guide the Opening and Ending Frames
Text to video; image to video
Animate an Opening Image
Current WeShop options
Duration: 6s, 10s
Duration: 6s, 10s
Duration: 6s, 10s
Model role
Hailuo 02
Physical movement and instruction-led scenes
Hailuo 2.3
Character expressions, physical actions, and stylization
Hailuo 2.3 Fast
Efficiency-oriented image-to-video variant
WeShop input workflow
Hailuo 02
Animate an Opening Image; Guide the Opening and Ending Frames
Hailuo 2.3
Text to video; image to video
Hailuo 2.3 Fast
Animate an Opening Image
Current WeShop options
Hailuo 02
Duration: 6s, 10s
Hailuo 2.3
Duration: 6s, 10s
Hailuo 2.3 Fast
Duration: 6s, 10s
Choose Hailuo 02
Choose Hailuo 02 when you want to supply both an opening and ending image to guide the clip’s endpoints.
Choose standard 2.3 for character performance, subtle expressions, or stylized animation. Start with a text description, or add a starting image to guide the opening composition. Choose 6 or 10 seconds.
Choose Fast to try the family’s efficiency-oriented image-to-video variant. Compare it with standard 2.3 using the same starting image and action description, and check the current points estimate.
Answers about this version, its inputs, and how to get started.
What is Hailuo 2.3?
Hailuo 2.3, officially MiniMax-Hailuo-2.3, is MiniMax’s text-to-video and image-to-video model, with a focus on character motion, micro-expressions, stylization, and object movement.
What is Hailuo 2.3 useful for?
Use it for character performances, illustrated animation, and creative effects from text, or animate an existing image. The live-action/cartoon, fox, and retro-TV examples on this page are image-to-video cases with their starting frames and full prompts.
Can Hailuo 2.3 generate a video without an image?
Yes. Text-to-video generates directly from a written description, so an image is not required. Describe the subject, setting, action, and camera movement. Add one starting image when you want image-to-video generation guided by an existing subject or composition.
Which duration and resolution settings can I choose?
This form offers 6- and 10-second durations and a quantity setting, with no resolution selector. Website previews may differ from the original output dimensions; download the original to check its width, height, and duration. Writing 4K or 8K in a prompt does not change the output options offered by the form.
How should I write a Hailuo 2.3 prompt?
When starting from text, describe the subject, setting, action, and camera movement. When using a starting image, focus on what should happen next. The guide preserves full original English prompts for outfit changes, an energy-pulse transformation, and a cooking sequence; all three are paired image-to-video cases.
How do I choose between Hailuo 02, 2.3, and 2.3 Fast?
Choose standard Hailuo 2.3 to build a scene from text or use a starting image for character motion, subtle expressions, and stylized animation. Review Hailuo 02 for opening-and-ending-frame control, or Hailuo 2.3 Fast for the family’s efficiency-oriented image-to-video variant.
Is Hailuo 2.3 free, and how is usage charged?
Check WeShop Pricing and your account for model access and allowance. Before generating, review the points estimate shown for your selected duration and quantity.
What should I check in the generated video?
Watch the full result for facial expressions, hands and limbs, clothing continuity, object shapes, and camera movement. Complex transformations, interactions, or timecoded actions may need a revised description and another generation.
Can I control both the opening and ending image here?
Text-to-video works without an image; image-to-video uses one starting frame. For a workflow with both opening and ending images, review the first-and-last-frame option on the Hailuo 02 page.