Higgsfield plugs Gemini Omni into its supercomputer as an orchestrator for multimodal content production

Higgsfield has started positioning Google’s Gemini as the orchestration layer inside its Supercomputer product, a move that ties the startup’s multi-model creative workflow software to Google’s newest multimodal video system. Higgsfield’s Supercomputer page says the product is now “Powered by Gemini” and that an “Orchestrator” chooses “the best-fit model for every step.” Google, for its part, said at I/O 2026 that Gemini Omni can “create anything from any input, starting with video.” The pairing matters because Higgsfield is selling Supercomputer as a single chat-based system for planning, generating and delivering creative work across multiple tools and models. The company’s product page says users can connect Slack, Drive, Notion, Gmail and Figma, import memory from Claude, Codex and ChatGPT, and use scheduled tasks and reusable “skills” to automate recurring workflows. ### What exactly did Higgsfield change? (higgsfield.ai) Higgsfield’s website on May 21 described Supercomputer as “Automate workflows, run agents, skills & more” and added the line “Powered by Gemini.” The same page says the “Orchestrator now picks the best-fit model for every step,” alongside claims of lower cost and faster answers. The product is not presented as a single-model system. Higgsfield says Supercomputer can route across “every frontier model under one roof,” listing Claude Opus 4.7, Claude Opus 4.6, Sonnet 4.6, GPT-5.5 Pro and Gemini 3.1 Pro. (higgsfield.ai) ### What is Gemini Omni, and why is it relevant here? Google announced Gemini Omni at I/O 2026 on May 19. In Google’s keynote materials, the company said Gemini Omni is capable of generating “samples in any output modality from any input,” beginning with video outputs. (higgsfield.ai) Google DeepMind’s model page says Gemini Omni is where Gemini’s reasoning meets media generation and editing. Google’s model card for Gemini Omni Flash says the system is designed to create and edit from mixed inputs and describes it as a step toward models that can “create and edit anything from any input—starting with video.” Google also said Omni Flash is rolling out in Flow and other Google AI surfaces. (higgsfield.ai) ### How does this fit into Higgsfield’s broader product stack? Higgsfield already sells itself as a multi-model video and image platform. (blog.google) Its AI video page says users can access models including Kling 3.0, Veo 3.1 and Sora 2 in one workspace, while its image product page says Nano Banana is powered by the “Gemini Flash reasoning engine.” The company’s homepage also markets adjacent products for marketing, cinema and canvas-based workflow building. (deepmind.google) Supercomputer is described there as an agent with “skills, memory, and 24/7 automations,” rather than as a standalone video model. ### What does Higgsfield say the orchestrator does? Higgsfield says the orchestrator reduces manual model selection by routing each step to the model it considers most suitable. The Supercomputer page links that routing to lower cost, faster responses and continuity across projects through saved files, memory and reusable workflow commands. (higgsfield.ai) The company also says users can ask the agent to research, generate documents, publish HTML briefs, make “5-10 minute” films and produce “100 UGC and ad variants” from one prompt. (higgsfield.ai) Those are product claims on Higgsfield’s own site; the company does not spell out, on that page, which tasks are handled directly by Gemini Omni versus other integrated models. ### What happens next? Google said on May 19 that Gemini Omni Flash was beginning to roll out in its own products, including Flow, and that broader capabilities would expand over time. (higgsfield.ai) Higgsfield’s next public test will be whether Supercomputer demonstrations and customer output show the Gemini-based orchestration working across the company’s listed connectors, skills and model-routing system. (blog.google)