Text to Video AI
Describe one shot in plain words, pick a look and a quality, and get a 5 second widescreen clip in a few minutes.
Updated
Example outputs
Made with the tool on this page. Prompt: "a girl in a white summer dress rides a bicycle down a hill toward the sea on a bright summer morning, the camera tracking beside her, the ribbon of her straw hat fluttering in the wind". Look: 2D anime. 480P, 31 credits.
Prompt: "a lighthouse keeper in a yellow oilskin coat climbs a spiral staircase with a lantern while rain lashes the windows, the camera rising with her, cold blue light outside and warm lantern light on her face". Look: Cinematic live action. 768P, 50 credits.
Prompt: "paper lanterns drift up from a dark river at night while a small boat glides under a stone bridge, the camera tilting up with the lanterns". Look: Watercolor painting. 480P, 31 credits.
Prompt: "a small round robot with one blue eye waters a single red flower in a rooftop garden at sunset, slow push in, the city skyline glowing behind it". Look: 3D animated film. The model gave the robot two eyes. 768P, 50 credits.
How it works
Describe one shot: who is in it, where they are, the one thing that happens, and how the camera moves.
Pick a look, from cinematic live action to 2D anime or stop-motion clay, and a quality: 480P (31 credits) or 768P (50 credits).
MiniMax H3 Max Turbo makes a 5 second 16:9 clip, usually in one to three minutes. Download the MP4.
Ideas and uses
| Write | What you get | Use it for |
|---|---|---|
| wide shot from high up | The whole place, people small in it | Opening a scene, showing where we are |
| medium shot at eye level | A person from the waist up, a neutral view | Most action and talking |
| close-up | A face or one object fills the frame | A reaction, a decision, a detail that matters |
| low angle, the camera near the ground looking up | The subject looks tall and strong | A hero moment, a threat |
| overhead, looking straight down | A map-like view of the scene | Tables, crowds, a fall, a fight seen from above |
| the camera tracks beside them | The camera moves along with a walking or running subject | Chases, journeys, travel |
| slow push in | The camera moves closer during the shot | A realization, tension building |
Write one shot, not a story
Text to video AI films one shot at a time. The clearest prompts read like a line from a shot list: who is in frame, where they are, the one thing that happens, where the camera is and how it moves, and what the light is doing. Everything else is a guess the model makes for you.
- Subject and action: "a girl in a red raincoat jumps over a puddle". One action, one instant.
- Place: two or three things the camera sees, not the whole world. "Wet cobblestones, a bakery window, evening fog."
- Camera: size, angle and move. "Medium shot at waist height, the camera tracks beside her."
- Light: where it comes from and its color. "Warm light from the bakery window on her left, cold blue sky behind."
Weak prompt, better prompt
Weak: "a cool cinematic video of a samurai, epic, 8k". The model has to invent the place, the action, the camera and the light, so you get a generic poster that barely moves.
Better: "a samurai in a gray coat walks through tall grass at dusk toward a lone tree, the camera low in the grass tracking beside him, the wind bending the grass, orange sun low behind the tree". Every part of that sentence tells the model something it can draw and move.
Leave out quality words such as masterpiece, 8k or award-winning. They do not add detail, and they push the result toward a still poster.
Honest limits
- Five seconds, one shot. Longer videos are several clips cut together.
- Fast, complex action such as a fight with many moves can look floaty. Show the wind-up and the result in two shots.
- Hands, small objects and text are the weakest parts of any AI video.
- The same character in many clips needs a reference picture. Text alone redraws them every time.
From a clip to a film
When one clip is not enough, Ziztu Studio plans the shots for you. Write the idea or a script in the AI Script Writer, and the Director turns each scene into shots with sizes, angles and camera moves, keeps the cast the same with character sheets, and puts clips, music and captions on one timeline.
Frequently asked questions
Is this text to video AI free?
You can try it free. A new account gets 40 free trial credits, and a 5 second clip at 480P costs 31 credits, so the trial covers one clip. Trial clips carry a small watermark. A clip at 768P costs 50 credits. Plans start at $12 a month for 1,000 credits.
How long is each video?
Each clip is 5 seconds, in 16:9 widescreen. In Ziztu Studio one shot can run up to 15 seconds, and longer videos are made from many shots on a timeline.
Does the video have sound?
Yes. The model adds sound that fits the scene, such as rain, wind, footsteps or a city at night. There is no voice-over or music; add those in an editor or in Ziztu Studio.
Can I make a vertical video?
This page makes 16:9 clips. For 9:16 clips for YouTube Shorts, TikTok and Reels, use the AI Shorts Generator, which uses the same model.
What makes a good text to video prompt?
One shot, not a story. Say who is there, where, the one thing that happens, how the camera moves and what the light is like. For example: "a lighthouse keeper climbs a spiral staircase with a lantern while rain hits the windows, the camera rising with her, cold blue light".
Can the same character appear in several clips?
Not reliably from text alone: each clip draws the character again. For a character who must look the same in every shot, make a picture first and use image to video, or use Ziztu Studio, which sends character sheets with every clip.
Can it write words on screen?
Text inside AI video is often garbled, so the prompt asks for no on-screen text. Add titles and captions afterwards in an editor, or on the Ziztu Studio timeline.
Can I use the clips commercially?
Clips made with paid credits have no watermark and every plan includes commercial use. Free trial clips carry a small watermark.
Which AI model is used?
MiniMax H3 Max Turbo text to video. Ziztu Studio also offers Kling, Seedance and Grok Imagine video models.