Video: from a description to a clip

Three ways to make a clip

• From text. You describe the scene in words and the model shoots it from scratch.
• From an image. You upload a frame — your photo or an image from the studio — and the model brings it to life. The frame becomes the start of the clip.
• From references. Some models accept several images, and the Seedance models also take video and audio clips. This keeps a character, a product or a rhythm consistent.

The list of video models with length, resolutions and price is in the catalogue.

First and last frame

Some models, such as Wan and MiniMax, take both a first and a last frame and work out how to get from one to the other. The frame fields appear in the studio only for models that support them.

• Make both frames in the same style, light and aspect ratio, or the transition will be jumpy.
• Describe what happens between the frames: the model will not guess the story from two pictures.
• First and last frames cannot be combined with any other references (images, video or audio) — pick one approach.

Length, resolution, aspect ratio and sound

• Length is set in the settings: from 2 to 30 seconds depending on the model; the exact range is on the model page. Longer clips cost more.
• Resolution — from 360p to 1080p, higher on some models. 480p or 720p is enough for a draft; use 1080p for publishing.
• Aspect ratio — 16:9 for screens and YouTube, 9:16 for Stories, Reels and Shorts, 1:1 for feeds.
• Sound. Some models can add sound — speech, effects, music; the switch appears only for those models. With Kling and PixVerse a clip with sound costs more.
• Grok Video mode: fun — a playful take, normal — balanced motion, spicy — sharper and more expressive. Spicy is unavailable with references.

How to describe motion and the camera

In video, the prompt covers not only what is in the frame but also what happens to it. Describe three things: the action, the camera movement and the pace.

• Action: “a cat jumps onto the windowsill and turns around”, not “a cat on a windowsill”.
• Camera: “slow push-in”, “the camera circles around”, “a dolly shot along the stalls”, “top view, the camera descends”.
• Pace and mood: “smoothly”, “slow motion”, “quick changes of angle”.
• One scene, one main movement. If you need several events, make several short clips.

A flower market

A low, smooth dolly shot along the stalls of a flower market in the morning: buckets of tulips, peonies and sunflowers, pollen swirling in the sunbeams, a florist in the distance with her back to the camera, shallow depth of field

What it costs

The price of a clip depends on the model, the length and the resolution, and for models with sound on the sound too. The studio recalculates it with every change in the settings and shows it before you run.

Tokens are charged only for a finished clip: if a generation fails, they stay on your balance. Prices for every model are in the video catalogue.

If it did not work

• The clip is too static — add an action and a camera movement to the prompt.
• The character changes from frame to frame — upload its image as the first frame or a reference.
• The prompt was rejected — rephrase it neutrally and without forbidden topics.
• A clip takes longer than an image — the result appears in the result block and in History, so you do not need to keep the page open.

How to make a clip in the studio

1. Pick a video model. Choose a video model in the Model field. It decides the length, resolutions, sound and the ways to set frames.
2. Add a frame or references. For a clip from an image, add it to the References block; for models with first and last frames, use the First frame and Last frame fields. Skip this step for a clip from text.
3. Describe the motion. Write what happens in the scene, how the camera moves and at what pace.
4. Set up the clip. Choose the length, resolution, aspect ratio and, if the model supports it, sound. The price updates right away.
5. Run. Press Run. The finished clip appears in the result block and in History.