Runway opens ChatGPT integration for video and image generation

• runwayml's launch post says you can now generate and edit Runway videos and images inside ChatGPT, with outputs staying in the same chat flow instead of sending you back to a separate app.
• Runway's official MCP announcement and setup guide show that ChatGPT is one of the supported surfaces for Runway's media tools, alongside Claude, Cursor, and Replit.
• The current Runway model menu inside agent workflows includes Seedance 2.0, GPT Image 2, Kling, Nano Banana Pro, and Gen-4.5, so the ChatGPT hook is also a distribution layer for third-party models, not just Runway's own stack.
• In a side by side test, CuriousRefuge's comparison said Seedance 2.0 Omni looked stronger on realism, physics, and motion, but Aleph 2.0's frame-based edit workflow gave creators more direct control over what changes.

You can browse Runway's MCP launch post, open the ChatGPT setup steps, and check the model list on Runway's MCP page. For the more interesting creative angle, Runway's Aleph 2.0 page and CuriousRefuge's test clip both point to the same thing: editing one frame, then regenerating the rest of the shot, is a different workflow from prompt-and-pray video iteration.

The June 15 push was simple: Runway is now callable from ChatGPT for video and image generation, and runwayml's wording emphasized that the pitch is staying in one thread, with no tab switching.

Under the hood, this sits on top of Runway MCP, which Runway introduced in May as an MCP server for agents. Runway's own connection guide says ChatGPT users add Runway from More → Apps, then connect a Runway account.

Runway is treating ChatGPT less like a single-model endpoint and more like a front door to its media stack. The MCP product page says agent users can access:

• Seedance 2.0
• GPT Image 2
• Kling
• Nano Banana Pro
• Gen-4.5
• Veo and other models, depending on plan, per the help guide

That matters because the ChatGPT integration is not limited to native Runway generation. It also exposes the aggregator role Runway has been building, where one chat surface can route requests to several video and image systems.

The sharper creative finding came from CuriousRefuge's comparison, which tested Aleph 2.0 against Seedance 2.0 Omni. Their verdict was that Seedance still looked better overall, especially on realism, physics, and believable motion.

But the workflow difference is the part worth saving. According to that comparison, Aleph 2.0 lets you change a specific frame, upload that as reference, and generate a new version of the video from it. Runway's Aleph 2.0 product page describes the same mechanic as editing one frame and having the rest of the clip update to match, while preserving the parts you did not ask to change.

Runway's Edit Studio writeup adds two concrete limits: multi-shot edits can run up to 30 seconds, and the tool is aimed at swaps, removals, insertions, and shot-level transformations on existing footage.

Runway's help doc says ChatGPT access does not require an API key. Users sign in with a regular Runway account, and generations consume the same Runway credits used in the main app.

The pricing wrinkle is model dependent. Runway's API pricing page lists aleph2 at 28 credits per second with a 56-credit minimum, while seedance2 runs at 36 credits per second for 480p or 720p and 40 credits per second at 1080p. So the new ChatGPT surface is convenient, but the actual cost profile still depends on which model the conversation calls.