Firma Runway przestała być firmą modelującą sztuczną inteligencję i stała się warstwą infrastruktury dla mediów generatywnych. W czwartek firma Runway przedstawiła Media Router za pośrednictwem swojej platformy programistycznej Runway Dev, która została uruchomiona na początku tego miesiąca. Platforma ta zapewnia dostęp API do wielu modeli obrazów, wideo i audio innych firm, a także do własnej oferty Runway.
Media Router automatycznie wybiera optymalny model generacji w oparciu o priorytety programistów dotyczące jakości, szybkości i kosztów. W przeciwieństwie do routerów powszechnych w dużych modelach językowych, Runway twierdzi, że jest to pierwszy router zaprojektowany specjalnie dla mediów generatywnych. “The routing really fits into that overall promise of being the easiest one-stop shop for developers to integrate with any type of generative media model,” Anthony Maggio, Runway’s chief product officer, said.
This launch signifies another step in Runway’s evolution from an AI video startup to an essential service for companies utilizing generative media. Through Runway Dev, customers such as Adobe, Cloudflare, ElevenLabs, Expedia, Shutterstock, and Quora can integrate media generation directly into their products using Runway’s API instead of directing users to Runway’s application or website.
The introduction of the Media Router comes amid an increasing number of generative media models, complicating the evaluation of new options for developers. Runway Dev allows developers to access newly released media models, with Maggio noting that many developers are not fully aware of the capabilities or differences between various models based on their outputs across video, image, and audio.
Chinese generative media models have surged in popularity, but some businesses may preferentially turn to American models due to concerns over political issues and potential sanctions. Preferences set by developers primarily focus on token pricing and quality, especially as token costs have become a significant concern in 2026. The Media Router’s launch follows Runway’s transition from unlimited subscription plans to a token-based pricing model, which received criticism from certain users.
Według Maggio określenie, które modele zapewniają najlepszą jakość konkretnych zadań, pozostaje wyzwaniem w mediach generatywnych. The router’s intelligence layer leverages insights from Runway’s in-house creative team, examining model performance in aspects such as video motion handling, image composition, and voice lip-syncing.
Runway previously established this intelligence layer for its conversational AI product, enhancing its capability to turn text prompts into fully edited multi-shot videos. Media Router wykorzystuje technologię routingu opracowaną dla wewnętrznych produktów Runway i dostosowuje ją do potrzeb zewnętrznych programistów.
Despite past successes, like the excellent performance of its last AI video model, Gen 4.5, which topped leaderboards against competitors, there’s been a significant gap in releasing new dedicated frontier video models since December. Runway’s text-to-video and image-to-video models no longer command leading positions, with competitors like Google, ByteDance, and Alibaba ranking higher.
The Media Router allows for dynamic selection of models as optimal choices shift over time, maintaining Runway’s relevance in the competitive landscape. Co-founder and co-CEO Anastasis Germanidis stated that the startup needed to build a full stack, encompassing a developer platform, creative tool suite, and inference layer. As interest from companies for Runway to support its entire technology stack grows, he emphasized the growing importance of orchestration in enabling businesses to create comprehensive campaigns and scene generations using its models.
