A film-grade digital human model that generates realistic video from a single image, audio clip, and optional text prompt.
Frequently asked questions
What is OmniHuman 1.5 and what does it do?
OmniHuman 1.5 is a film-grade AI model by ByteDance that generates realistic, character-driven video from a single image, an audio clip, and an optional text prompt. It produces synchronized speech, nuanced emotions, and cinematic camera movement without requiring actors, cameras, or a production studio.
Who made OmniHuman 1.5?
OmniHuman 1.5 was developed by ByteDance, the company behind TikTok. It is available to use via Replicate under the slug bytedance/omni-human-1.5 and is also accessible through the Craisee platform for creative workflows.
What types of images does OmniHuman 1.5 support?
OmniHuman 1.5 supports a wide range of subject types including realistic human portraits, animals, stylized illustrations, pixel-art characters, and animated avatars. It is not limited to photographic portraits, making it versatile for both realistic and creative animation projects.
Does OmniHuman 1.5 support multiple characters in one scene?
Yes. OmniHuman 1.5 includes multi-character scene management, allowing creators to specify which character speaks while independently controlling background character reactions. This enables complex ensemble scenes to be generated from a single input image.
What languages does OmniHuman 1.5 support for text prompts?
OmniHuman 1.5 accepts text prompts in six languages: Chinese, English, Japanese, Korean, Spanish, and Indonesian. This multilingual support makes it accessible to a global audience of creators working in diverse linguistic contexts.
How is OmniHuman 1.5 different from earlier versions?
OmniHuman 1.5 adds text prompt integration that earlier versions lacked, enabling creators to guide character behavior and scene composition with written instructions. It also improves micro-expression rendering, free camera movement, and support for sequential character actions within a single generation.
Can OmniHuman 1.5 generate cinematic camera movement?
Yes. OmniHuman 1.5 supports unrestricted camera movement and multiple sequential character actions within a single generation. This allows shot compositions that go well beyond simple talking-head formats, enabling more dynamic and cinematic video output.
How do I use OmniHuman 1.5 on Craisee?
On Craisee, OmniHuman 1.5 is available as a ready-to-use model for image-to-video workflows. You provide a single image, an audio clip, and an optional text prompt. The model then generates a realistic, emotionally coherent video with synchronized speech and cinematic motion.
