Inspiration
Studio product
PROMPT-ONLY CREATION · NATIVE AUDIO
MiniMax H3 Text to Video
Create a new scene from a written prompt—no image required. Guide the action, camera, lighting, and sound, then generate a 5- to 15-second video at 768p or up to 2K.
Model
Choose Standard for up to 2K output. Typical generation: 2–5 minutes.
768p · 3 credits/sec · No image requiredExample videos
Get inspired
Inspiration
Architectural light
Inspiration
Fashion motion
Inspiration
Coastal aerial
First time here? Learn how to use MiniMax H3 in seven simple steps.
From idea to shot
Create a Scene from a Written Prompt
Use MiniMax H3 Text to Video when your starting material is an idea, a script beat, or a creative brief. Describe what a viewer should see and hear, then generate a short audiovisual scene without preparing an image.
This workflow suits early product concepts, atmospheric b-roll, and vertical social clips. It gives you room to explore a look before choosing a specific visual reference. For model capabilities and broader examples, visit the MiniMax H3 overview.
Quick start
How to Use MiniMax H3 Text to Video
Start with one shot, check the settings, and use the first result to decide what to refine.
- 01
Write One Shot
Build the scene in words: who or what is visible, where it sits in the frame, and what changes. For example, “a ceramic cup centered on a café counter, with steam rising” gives the generator a subject, a location, and a small action without a source image.
- 02
Direct the Camera and Audio
Separate subject movement from camera movement. “The taxi drives left to right while the camera tracks beside it” describes their relationship. Then add an audible event, such as tire noise on wet asphalt, instead of a generic request for cinematic sound.
- 03
Set the Format and Check Credits
Choose H3 Standard when you need up to 2K output. Use Max for fast 480p or 768p drafts, or Turbo for lower-credit iterations. Select an aspect ratio that leaves room for the described action, then review the live cost. Model guidance above Prompt updates with your selection.
- 04
Generate and Review
Click Generate and sign in if prompted. Standard H3 usually takes 2–5 minutes, depending on the shot and queue. Watch the complete result with audio, check the final frame, and change one instruction for the next version. You can return to My Creations later.
Need a walkthrough of the account and workspace? Read the step-by-step MiniMax H3 tutorial.
Starter prompts
MiniMax H3 Text to Video Prompt Examples
Use these as starting points, then adjust the subject, framing, and sound for your shot.
Ceramic Cup · Product Concept
A handmade ivory ceramic cup sits on a pale stone counter beside a softly lit café window. Steam curls gently above the cup. The camera makes a slow push-in, keeping the cup centered and its handle shape consistent. Warm morning light, natural ceramic texture, and a softly blurred background. Quiet café room tone and a faint espresso machine hum. End on a steady close-up. No text or extra objects.
Text to Video · H3 Standard · 16:9 · 768p · 5s
Rainy Street · Cinematic B-Roll
A yellow taxi drives slowly through a rain-soaked city street at night. The camera tracks beside it at a low angle at the same speed. Amber shop lights reflect across the wet pavement. Keep the taxi shape and wheel motion realistic and consistent. Light rain, restrained contrast, and subtle mist. Tire noise on wet asphalt, a quiet engine, and distant city ambience. End with the taxi still fully in frame.
Text to Video · H3 Standard · 16:9 · 768p · 5s
Dessert Reveal · Vertical Social Clip
A pastry chef's hands place a strawberry tart on a white plate on a marble counter. The camera stays locked in a close-up with vertical framing. A light dusting of powdered sugar falls over the tart, then settles. Soft window light, realistic food texture, and a clean blurred background. A gentle plate tap and quiet kitchen ambience. End on the finished tart with space above it for a caption added later. No generated text.
Text to Video · H3 Standard · 9:16 · 768p · 5s
Video example
A Prompt and Its Video Result
This night-driving example is also featured in our showcase. Compare the written camera direction with the moving shot.
Watch how the camera follows the car, how reflections move across the road, and whether the wheels and body stay consistent. Listen to the audio as well as reviewing the image.
Neon Night Drive — Cinematic Motion
A black sports car drives through a rain-soaked downtown street at night while the camera tracks smoothly beside it at a low angle. Red and blue neon signs reflect across the wet pavement and glossy bodywork. Keep the car geometry and wheel motion realistic and consistent. Cinematic contrast, light rain, subtle atmospheric haze, engine sound, wet-road tire noise, and distant city ambience.
Text to Video · H3 Standard · 16:9 · 768p · 5s
Output settings
Match the Shot to Its Destination
Choose the delivery format before generating so the scene has room for the action you describe.
Try 768p, Then Generate at 2K
Use a short 768p draft to assess the action and framing. When the prompt is ready, select 2K and generate a new video. This creates a separate result and uses additional credits; it does not upscale the earlier draft. Refine unclear instructions before increasing resolution.
768pUp to 2K
Compose for the Aspect Ratio
Use 16:9 for web and landscape video or 9:16 for mobile feeds. In vertical prompts, describe a centered subject and keep important movement inside the frame. A wider format can give an environment more space.
21:916:94:31:13:49:16
Give the Action Enough Time
Five seconds can suit a single reveal or camera move. Use a longer clip when the action needs time to unfold, and describe a clear ending. Avoid adding more unrelated events simply because you select 15 seconds.
5–15 secondsNative audio
Prompt fixes
Fix Ambiguity When You Have No Source Image
Use concrete relationships and observable actions to narrow the next version.
The Scene Adds Unwanted Objects
Replace “a beautiful café scene” with a defined composition: “one ivory cup on a bare stone counter, with a blurred window behind it.” State which objects belong in the frame before adding style words. Avoid listing unrelated props that compete with the subject.
The Action Has Too Many Stages
“A chef prepares, plates, and serves dessert” asks for several events in one short clip. Narrow it to “the chef places a finished tart on a plate.” If the process matters, generate separate shots and assemble them in an editor.
The Camera Loses the Subject
Define the relationship: “track beside the taxi at the same speed, keeping the entire vehicle in frame.” For a product, try “a static close-up with the cup centered.” Add a final hold when you need a clean endpoint for editing.
For a broader writing framework, use the MiniMax H3 Prompt Guide. Find completed drafts in My Creations.
Questions
MiniMax H3 Text to Video FAQ
Answers about prompt-only creation, video settings, audio direction, and refining a draft.
Can I create a video without uploading an image?
Yes. MiniMax H3 Text to Video starts with a written prompt. Images, keyframes, and reference files belong to other workflows; you do not need them to generate a new scene from an idea.
How specific should I be when there is no source image?
Name concrete visual details that would otherwise be open to interpretation: the subject’s material or clothing, its position, the background, and the action. “An ivory cup centered on pale stone” is more specific than “a stylish product shot.” Standard H3 accepts up to 7,000 characters; clarity matters more than filling the counter.
How do I add dialogue, music, or sound effects?
Write the audio direction in the same prompt. For dialogue, name the speaker and include a short quoted line. For effects, connect the sound to an event, such as a door closing. H3 can create synchronized stereo audio; listen to the entire result to check clarity and timing.
Which model should I choose for a prompt-only draft?
Choose Standard for 768p or up to 2K. H3 Max and H3 Max Turbo offer 480p or 768p and faster draft iteration; Turbo uses fewer credits at the same resolution. All three accept text-only input and 5- to 15-second output. A faster or higher-resolution setting does not guarantee better adherence to a complex prompt.
Can I generate vertical video for Reels or Shorts?
Yes. Select 9:16 before generating and describe a composition that keeps the subject visible in a narrow frame. You can also choose 21:9, 16:9, 4:3, 1:1, or 3:4. Add exact captions in your video editor when reliable typography matters.
How much does a prompt-only video cost?
A standard five-second H3 generation estimates 15 credits at 768p or 25 credits at 2K. Longer clips and different models can change the cost. Review the live estimate and see current credit pricing before starting.
When should I switch from text to an image or references?
Use Image to Video when a specific product, character, or composition must anchor the shot. Choose Reference to Video when source media should guide appearance, movement, or audio. A text description can guide a look, but faces, product proportions, and small labels may vary between attempts.
Can I turn a multi-shot script into one generation?
For more control, split the script into individual shots. Give each generation one main action and a clear beginning and ending, then assemble the clips in an editor. Repeat key appearance and audio details across prompts, and use an image reference when visual continuity matters.
Start with One Clear Shot
Describe what happens, how the camera sees it, and what the viewer hears. Review your settings and credit estimate, then generate when you are ready.
Create Video from Text →Sign in to generate · No image required