Skip to content

Generating Videos

Video generation works best when your prompt describes motion, not only appearance. Treat the prompt as a short direction for a clip: what moves, how the camera moves, how the scene changes, and what the mood should feel like.

Select Video Studio from the sidebar.

Use the project picker near the top of the canvas before rendering, or change it after the video is created.

Choose a model that fits your goal. Some models favor cinematic quality, some are faster, some support audio, and some require or support start frames.

If you want image-to-video, add a Start frame. If the model supports first-last-frame animation, add an End frame too.

Describe the shot in the prompt field. Include subject movement, camera movement, timing, lighting, and style.

Choose the duration, resolution, aspect ratio, audio, seed, or negative prompt options that are available for the selected model.

Select Render video. Direct Studio renders wait for the provider result, then show the completed clip on the canvas.

Strong video prompts usually include:

  • Subject: the main person, object, product, or scene.
  • Action: what changes over time.
  • Camera motion: pan, dolly, orbit, handheld, push-in, pull-back, locked-off.
  • Timing: slow, quick, subtle, continuous, dramatic reveal.
  • Look: lighting, lens style, color palette, realism, animation style.
  • Audio direction: only when using a model that supports or responds to audio.

Example:

A slow orbit around a matte black espresso machine on a stone counter, steam rising from the cup, soft morning light through a kitchen window, cinematic product commercial, subtle ambient cafe sound.

While the video is rendering, the canvas shows a loading state. Video renders can take a few minutes. Keep the Studio tab open until the render finishes or you choose Stop.

If generation succeeds, the finished clip appears in the canvas and is saved to the Studio Gallery under Videos.

Add a prompt and make sure the selected model has valid settings. Image-to-video-only models also need a start frame.

Upload a start frame or choose a model with text-to-video support.

Add stronger motion language. Name the camera move and subject action instead of describing only the still scene.

Direct Studio generation waits on the page. If you need background rendering from chat instead, use Chat-Assisted Video.