TL;DR
- At OpenAI’s DevDay on September 29, OpenAI introduced three changes that move video inside AI assistants closer to reality.
- Plugin extensions let apps live in the ChatGPT sidebar, open in a side panel next to the chat, and run fullscreen, making a sustained viewing experience much more practical.
- ChatGPT is surfacing plugin recommendations during relevant conversations, so the names, descriptions, and scope of your tools can influence whether your service is surfaced for a relevant request.
- Sign in with ChatGPT gives services a shared identity for account linking, a first step toward recommendations that reflect what viewers already subscribe to.
- Discovery is a practical place to start, and our open-source Bitflix MCP App shows video playback working inside the assistant today.
The V3 HEVC codec matches the previous codec’s quality at 17.6% lower bitrate on a fixed ladder, or delivers 3.9 VMAF points more at the same bitrate. With Per-Title Encoding on both sides, the saving is 15.5% and the gain 3.0 VMAF points. Against a leading public cloud encoder, the gap is larger: 23.3% fewer bits than its QVBR ladder and 36.3% fewer than its automated per-title ladder. The gains are smallest on 50p and 60p television, where one clip saves 1.6%.
If you’re the engineer who has to answer whether it’s worth requalifying, this blog is for you. We’ll go into how we tested, where the gains hold up and where they narrow, and how to run the same test on your own content.
We recently published A Vision For Video Streaming Into AI Assistants, where we set out why we believe AI assistants will become the first place people go to find something to watch, and what it would take to get video playing inside the conversation rather than sending viewers elsewhere. A few days later, at its DevDay on September 29, OpenAI introduced three changes that move that vision closer.
In this blog we’ll go into what each change does, why it matters if you publish video, and where a streaming service can realistically start building today.
Your app can now stay open
Before plugin extensions, app experiences inside ChatGPT were largely centered on the conversation itself. That works for a chart or a form, but not for something a person wants to sit back and watch.
Plugin extensions change this. A plugin can now have its own home in the ChatGPT sidebar, open as an interactive panel alongside an existing conversation, and run fullscreen. Canva, Figma, and Adobe were among the early integrations highlighted around DevDay.
For video, this is the most significant of the three changes. A player that survives beyond a single exchange, and fills the screen when you want it to, is the difference between a preview and something a person actually watches.
Plugin discovery is becoming part of the conversation
OpenAI is also making plugins easier to discover, including through recommendations surfaced during relevant conversations.
This makes the commercial question concrete. Being available inside an assistant was never the same as being the one it picks. When someone asks what to watch tonight, the assistant’s suggestion can play a major role in which service gets the viewer, and that suggestion is now a designed feature rather than something that happens by accident.
For broadcasters and streaming services watching discovery traffic move away from their own apps and search results, this creates a new route back into their content and services. It also rewards learning a new discipline early, since the names, descriptions, and scope of your tools can influence whether they are relevant to a user’s request and surfaced at the right moment.
Sign in with ChatGPT
Sign in with ChatGPT lets people log in to other products with their ChatGPT account. At launch, 16 partners, including Notion and Vercel, also let users draw on their ChatGPT plan inside their tools.
The hardest unanswered question in our vision was always entitlements: how does the assistant know what you already subscribe to? Without that, recommendations stay generic and playback stops at a paywall. A shared identity layer doesn’t answer the question on its own, because the subscription data still lives with each video service. What it does provide is a shared identity that services can use as part of an account-linking flow, which is a credible first step.
If a service exposes that entitlement information to the assistant, the assistant could recommend an unwatched show on a service the viewer already has, rather than one they would need to sign up for. That could make recommendations more relevant while increasing the value of an existing subscription to the viewer.
What this means in practice
A discovery experience is the natural place to start. It’s one of the parts of the viewer journey most exposed as discovery shifts toward AI assistants, and it’s the part these new capabilities make most immediately buildable. For a streaming service, the flow looks like this:
- A viewer asks what to watch
- Relevant titles from your catalog appear in the conversation
- They browse and preview a title
- They continue to your service to watch
That’s also where our own work goes next. We’ve open-sourced Bitflix, our MCP App with real video playback inside the assistant, and over the coming months we’ll release new examples, starting with discovery and previews and moving on to news and sports. You can find the code in the Bitflix repository on GitHub.
From there the path extends to connecting the viewer’s account, making recommendations that reflect what they already subscribe to, and eventually playing the content directly inside the assistant.
Where this goes
Apps inside assistants are becoming persistent, discoverable, and tied to who you are. Those are three important building blocks for any video service that wants to exist meaningfully inside an AI assistant, and DevDay moved all three forward at once.
If you’re a streaming service, broadcaster, or publisher thinking about how your video shows up inside AI assistants, we’d like to hear from you. Get in touch.
