CraftStory, a company pioneering artificial intelligence generated human-centric video, announced the release of its first image-to-video model today, which allows users to generate up to five-minute videos.

The new capability expands the company’s existing video-to-video model, dubbed Model 2.0, which launched in November 2025.

As more companies lean into video as a communication format, the image-to-video workflow is could drive use cases including marketing and advertising, business communications and educational content. Allowing teams to produce consistent “on-camera” human performances without traditional production.

Currently, most video generation models struggle to produce coherent footage beyond 10 to 30 seconds. To create longer videos, users often stitch together shorter clips into longer narratives, but additional generations can drift — producing slightly different faces, outfits, lighting or motion — creating alignment issues. Advanced AI workflows and tooling can push videos past two minutes, but longer narratives can quickly dissolve into algorithmic chaos.

CraftStory achieves its long-form capability using a proprietary parallelized diffusion pipeline that processes different segments simultaneously. The approach allows the platform to enforce coherence between clips, maintaining visual consistency across minutes of footage.

“Image-to-video is a major step toward fully script-driven video creation,” said founder and Chief Executive Victor Erukhimov (pictured), who previously sold his computer vision startup Itseez Inc. to Intel Corp. “You no longer need to record a video to get a realistic human performance.”

Erukhimov said that with Model 2.0, starting with just an image users can achieve believable human presence in long-form videos, complete with gestures and expressiveness that matches the message.

The model was trained with high-frame rate footage of real actors, capturing the dynamics of facial expression, hand motions and body language. According to CraftStory this allows for high fidelity production of human “actors” that feel fluid and lifelike, not static or robotic.

Videos can be generated in both portrait and landscape formats at 480p and 720p, with upscaling to 1080p for higher-end output. The company also introduced support for moving cameras, enabling walk-and-talk videos up to 80 seconds long with natural motion through a scene.

Users can create videos from a single image plus a script or audio track. The system generates a scene that follows the script and lip-syncs AI actors, while built-in gesture alignment aims to keep body movement natural and match the cadence and emotion of the speech.

Image: CraftStory