Generate and edit video from any input, text, image, video, or audio, through Runware, the lowest-cost API on the market. Commercial use permitted from day one.
Gemini Omni Flash, Google DeepMind’s new multimodal video model, is now available through Runware’s API. No waitlist, commercial use from day one, and a single integration that runs it alongside every other model Runware serves.
Nano Banana quality, now for video
Gemini Omni Flash is the first model in Google DeepMind’s Omni family, which Google calls its most advanced multimodal video generation and editing model to date. Unlike standalone text-to-video models, Gemini Omni Flash accepts text, image, and video as inputs and produces video grounded in Gemini’s knowledge of the real world, including physics, geography, culture, and language.
Its defining capability is conversational, multi-turn editing. Every edit builds on the last. Characters stay consistent across scenes, and locations carry over. A production team can iterate on a video the way they edit a document, one natural language instruction at a time, without regenerating from scratch.
Marketing Technology News:Â MarTech Interview with Miguel Lopes, CPO @ TrafficGuard
For teams that have stitched together separate generation and editing pipelines, that removes a major step.
Available today on Runware
As of today, Gemini Omni Flash is available through Runware’s API. Developers and enterprise teams can access it immediately alongside Runware’s full library of image, video, audio, 3D, and LLM models through a single integration.
Runware runs on its proprietary Sonic Inference Engine, with usage-based pricing, high concurrency, no GPU provisioning, and no long-term contracts. For teams running production-scale workflows, Gemini Omni Flash on Runware means the full generative media stack behind one API.
Marketing Technology News:Â Is the Traditional CDP Already Out of Date?
What Gemini Omni Flash supports on Runware
– Text-to-video: Generate video from a prompt, grounded in real-world accuracy
– Image-to-video: Animate a still image or reference into a full scene
– Reference-to-video: Drive output style, character, and setting from existing assets
– Multi-turn video editing: Iterate conversationally, swapping locations, updating characters, or changing dialogue, with each edit preserving what came before
– Native audio sync: On-screen text and sound generated in sync with the visual
– Location and world awareness: Regionally accurate settings, signage, and language, without manual prompting
– Character and object swaps: Replace elements by asking, with motion and dialogue preserved
– Sketch-to-video: Turn a drawing into footage, using the doodle as a guide for how elements move
– Motion and style transfer: Apply the motion or visual style of one clip or image onto another
– Production at scale: Turn a small asset set into many production-grade outputs quickly
Built for enterprises
Enterprise teams get more than API access: dedicated requests-per-minute (RPM) allocation, volume-based pricing, 24/7 engineering support, and a single API across Runware’s full model library. Runware is SOC 2 and ISO 27001 certified and GDPR compliant, with SSO and SAML for enterprise authentication. Contact Runware to discuss capacity and onboarding.
