Omni Flash

Omni Flash reframes video editing as a conversation. Rather than assembling a timeline with masks and layers, you describe what should change in plain language and let the model apply it, then describe the next change, and the next. Each turn builds on the previous one, which is what makes the multi-turn approach useful: characters, props, lighting, and motion tend to stay recognizable as you iterate, instead of the clip drifting into something unrecognizable after the third revision.

Beyond prompt-only editing, the tool accepts a wide spread of references. You can guide an edit with an image, a text brief, a video clip, an audio cue, a sketch, or a motion and style reference, and combine several of those ingredients into one output. Reference images are especially relevant for character and object swaps — replacing a subject while matching the original motion, timing, and dialogue. The platform also leans on world knowledge for believability, citing physics, fluid dynamics, momentum, history, science, and narrative logic as inputs to how scenes move.

In practice that makes it a fit for short-form creators, ad and brand teams, storyboard artists, and motion designers who need precise, iterative edits rather than one-shot generation. The workflow is upload or describe, refine in conversation, then export — with generation history and downloads available after you sign in, and a set of free starter credits for new accounts.