AI Video Generator

An AI video generator creates a new video clip from a text prompt, a still image or reference footage. On Arumi, one of 35 video models is selected, the shot is described and the model renders it. The clip can then be downloaded, or its settings reused for the next take.

Generating video uses credits. Free credits, such as the sign-up bonus or daily check-in credits, work here at each model's lowest quality and resolution. Higher settings and paid extras such as generated sound require a credit pack, which is bought once with no renewal. The Pricing page lists the packs.

No single model is best: the models differ in motion, realism, length, resolution and sound. The comparison on this page shows what each model does well and where its limits are. A practical approach is to draft short, low-resolution takes on a cheaper model, then render the final shot on the model whose look is preferred.

Yes. Select a model that lists image to video, upload the picture as the first frame and describe the motion. Some models also accept a last frame and generate the movement between the two. Starting from an image is the most reliable way to control the look of the clip.

Up to 30 seconds in one generation, depending on the model. Most models allow the length to be chosen before generating. Some models can also extend an existing clip. For longer pieces, the shots can be generated separately and edited together.

Up to 4K, depending on the model. Higher resolutions cost more credits and take longer, so testing at a lower resolution and rendering the final version at the top setting is a common approach.

18 of the 35 video models list a sound option and can generate dialogue, effects or music together with the picture. Sound is turned on before generating. The With sound filter in the comparison shows which models these are.

From under a minute to several minutes, depending on the model, the length and the resolution. Generation continues if the page is closed, and the finished clip appears in Recent and in History.

The cost depends on the model, the length, the resolution and sound. The comparison shows each model's lowest price at today's rates, and the generator shows the price for the exact settings before Generate is clicked. A successful generation is charged exactly the quoted credits.

As between you and Arumi, the videos generated for you are yours to use, subject to our Terms of Use, the terms of the provider that runs the model, and the law, which in some countries does not grant copyright in machine-generated work. You are responsible for having the rights to anything you upload.

Uploads are stored privately and open only to you while you are signed in. We do not use prompts, uploads or results to train models of our own. To generate a video, the prompt and reference files are sent to the provider that runs the selected model.

Every request passes a content safety check before it reaches the model. If our check or the model's own safety system stops a generation, or the model fails, the card states the reason and whether any credits were charged. Our check and failures on our side are never charged.