Changelog

This release is all about making Program feel more capable, more predictable, and nicer to use across the full creation loop.

We added support for new generation models, tightened up approval flows for media generation, and spent time on the desktop app, especially on Windows. The result should feel simple: more creative range when you ask Program to make something, and fewer interruptions when you are trying to get from an idea to a finished video.

New models

Program now supports a fresh set of models across chat and video generation.

• Seedance 2.0 is now available for video generation through AI Gateway, giving Program access to ByteDance's latest video model.
• GPT-5.5 is now available for agent and editing workflows.
• Claude Opus 4.7 is now available for higher-capability reasoning and creative editing tasks.

These models expand the range of projects Program can handle, from fast edits and scene changes to richer generated media.

Better generation approvals

We improved the way Program handles permission-sensitive generation tools.

When an agent wants to generate media, such as audio, speech, images, or video, Program can now put the user more clearly in the loop before the tool runs. Approval prompts are displayed closer to the tool activity they relate to, with a cleaner visual treatment and less lingering UI after the decision is made.

This gives you better control over when generation happens, while still allowing faster workflows when you choose to use automatic approval modes.

Windows desktop improvements

The desktop app should now behave much better on Windows.

We improved local database startup, desktop auth redirects, chat streaming, local worker startup, and media playback behavior. We also fixed several Windows-specific rough edges around generated audio and local asset handling, so projects with generated media should preview more reliably.

Desktop playback and assets

Playback is smoother when opening a project and pressing play for the first time. Generated assets are also easier to find: audio, image, and video outputs tied to the current project now show up in the generated assets area instead of disappearing into the background.

General fixes

This release also includes a round of general stability, UI, and release pipeline fixes across the web app, desktop app, and CLI.

As always, Program is still moving quickly. Thank you for testing the edges, reporting what feels off, and helping us make the creative loop faster and more dependable.

We've been working on Program for a while now, and today we're making it available to everyone.

The idea behind Program is simple. Video creation has always been manual. You open an editor, drag clips around, keyframe animations, export, re-export, tweak, re-export again. It works, but it's slow, and it doesn't fit into the way a lot of us actually build things now.

If you're someone who works with AI coding agents every day, who writes software from the terminal, who thinks in components and compositions rather than timelines and keyframes, we built Program for you.

What is Program?

Program is an AI video composition platform. You can create and edit videos through the web app, the desktop app, or the Program CLI. Every scene in a video is a React component. Describe what you want, and your agent writes the code, builds the composition, and updates it in real time.

The desktop app

We also built a desktop app. It runs a local backend on your machine, so you can work offline and render without uploading anything to the cloud. Preview your compositions as they're being built, make edits, and export when you're ready. Everything stays on your hardware.

You bring your own API keys for generation. We don't proxy anything through our servers and we don't take a cut.

Download the desktop app

Use Program with your favorite coding agents

It works with the tools you're already using. Claude Code, Codex, OpenCode, OpenClaw. Give your agent the context, tell it what you want, and it handles the rest. Authentication, project setup, scene creation, voiceover. All of it.

We wrote guides for each of these if you want to get started:

• Generate videos with Claude Code
• Generate videos with Codex
• Generate videos with OpenCode
• Generate videos with OpenClaw

Why we built this

It's easier than ever to build things right now. The hard part isn't building anymore, it's getting what you've built in front of people. You can ship a great product and nobody will ever know it exists if you don't show it to them.

That's the problem we kept running into. We were building constantly but not sharing enough, because making content to actually show off what we'd built felt like a whole separate job. Video is the best way to do that, but the tools were slow and the process was manual. So we built Program to close that gap. Build something, then use the same agents and the same workflow to create the video that shares it with the world.

We're only getting started

We'll post updates here as we ship. If you want to try Program today, sign up for free or download the desktop app.

Thanks for being here early. We're excited to see what you make.