Why is OpenMontage suddenly blowing up?

Is OpenMontage really blowing up again right now?

Yes: OpenMontage is going through a real second growth wave right now, with its GitHub star growth roughly tripling from one week to the next.

GitFind currently tracks OpenMontage at around 54,100 stars and 6,700 forks. More interestingly, it recorded about 4,400 new stars over the latest week, compared with roughly 1,500 the week before. That jump is large enough that we are looking at a fresh acceleration rather than the leftover effect of an old launch.

OpenMontage had already gone viral earlier. Trendshift records the repository reaching number one on GitHub Trending during its first big breakout in June, after starting the spring with only a few hundred stars. Growth later cooled, as we would expect after such a large spike.

Now the curve has steepened again. That second wave is the unusual part. Plenty of open-source projects explode on GitHub Trending for three days and then settle into slow background growth. OpenMontage has managed to become hot again after already becoming one of the largest projects in its category.

Period	Approx. OpenMontage stars	What happened
Early April	~600	Still a small project
First June breakout	20,000+	Reached #1 on GitHub Trending
Shortly after	~29,000	Viral growth continued
Recent comparison point	45,109	Still gaining after the initial wave
Currently	~54,100	Growth has accelerated again

Why is OpenMontage growing so fast now?

OpenMontage is growing so fast now because the idea behind the project has become easier for developers to understand at exactly the same time that coding agents have become much more important.

When OpenMontage first appeared, “let Claude Code make a video” sounded like a clever hack. These days, developers already let Claude Code, Codex and similar agents edit files, run commands, call APIs, browse documentation and work through multi-step tasks. Asking the same agent to produce a video feels far less strange.

OpenMontage has also become much bigger since the first wave. The canonical repository currently describes 12 production pipelines, more than 100 tools and more than 700 agent-skill or production-knowledge files. Earlier versions promoted roughly 52 tools and 500-plus skills. The project kept adding real surface area while the attention temporarily cooled.

The latest GitHub environment helps too. Reusable agent skills, MCP tools and software designed specifically for coding agents are all attracting attention. OpenMontage fits neatly inside that movement because much of its “product” is essentially production knowledge that an agent can read and act on.

So the renewed growth looks like a convergence. The project improved, coding-agent behavior became more familiar, and OpenMontage suddenly makes more intuitive sense than it did during its first viral week.

What does OpenMontage actually do?

OpenMontage lets an AI coding agent handle most of the work required to turn a brief into a finished video, including research, scripting, scene planning, asset creation, narration, composition and rendering.

That makes OpenMontage quite different from the typical AI video product.

A tool such as Kling or Veo mainly answers a narrower request: generate this clip. A normal video editor helps a human arrange footage on a timeline. OpenMontage tries to manage the whole production.

The current workflow can start with something as simple as a product brief and a folder containing a logo, screenshots and screen recordings. The agent can research the topic, write the script, decide which scenes need stock footage or motion graphics, generate images or video where useful, create narration and music, assemble everything and check the result.

The system can mix several types of media inside one production. A product launch video might use the company's own dashboard recording for one scene, free stock footage for another, HTML animation for the next and a generated five-second clip where synthetic footage genuinely helps.

That flexibility is a big part of the appeal. Generative video models have become impressive, but generating every second of a video is usually expensive and often unnecessary. OpenMontage lets the agent choose the cheapest or most appropriate medium scene by scene.

Why does OpenMontage fit Claude Code and Codex so well?

OpenMontage fits Claude Code, Codex and other coding agents unusually well because developers can use the same chat-driven workflow they already use for software and point it at video production instead.

The user does not have to learn a completely new creative interface. OpenMontage runs through the agent environment and gives that agent specialized tools and instructions.

Underneath, much of the production logic lives in YAML pipeline definitions, Markdown instructions and callable tools. The agent reads those files, follows the production process and decides what to use at each step.

That architecture also makes OpenMontage easier to expand. Supporting a new video provider does not necessarily require redesigning the whole application. A new production technique can often be added through tools, instructions or a new skill.

The official OpenMontage website now describes the idea in six words: “Make video the way you now make software.” That positioning is unusually accurate. The people discovering OpenMontage on GitHub already understand what it feels like to tell an agent what they want and watch the agent operate a project until it gets there.

Video production has simply become another project for the agent to operate.

Is GitHub Trending doing most of the work for OpenMontage?

GitHub Trending is clearly amplifying OpenMontage, but it does not fully explain why the repository keeps coming back.

During the first breakout, Trending was enormous. Trendshift records OpenMontage reaching number one across GitHub, and archived snapshots from that period show it adding thousands of stars over very short periods.

That gave the project a distribution advantage it still enjoys today. A developer encountering an unfamiliar repository with 54,000 stars reacts differently from someone encountering the same idea with 300 stars. The existing number creates curiosity: why have so many people already starred this?

Trending can then reinforce the loop. More visibility brings more stars, which keeps the repository visible, which brings another round of developers.

Yet Trending alone cannot easily explain the second wave. The initial ranking disappeared. OpenMontage continued accumulating stars afterward and has lately accelerated enough to become interesting again.

GitHub Trending looks like an amplifier here. Plenty of repositories reach the Trending page and then disappear from view. OpenMontage found an idea that developers keep rediscovering after the first burst ends.

Are people actually using OpenMontage or just starring it?

We cannot prove that tens of thousands of people actively use OpenMontage, but the current data looks healthier than a repository collecting passive stars.

Fork growth is the strongest clue.

A recent Valmera survey checked OpenMontage directly against the GitHub API and found 45,109 stars alongside 5,553 forks. The live repository is now around 6,700 forks. Stars increased roughly 20% over that interval, while forks rose at almost the same rate.

GitFind also recorded around 447 additional forks during its latest weekly window. Forking a repository still does not mean somebody successfully produced a video, but it requires more intent than pressing the Star button and moving on.

Development is also unusually active. GitFind counts 122 commits over the latest 30-day period, while GitHub currently shows 448 commits across the repository's entire history. Roughly a quarter of that recorded commit history therefore happened very recently.

The data we really want is still missing. OpenMontage does not publish weekly active creators, finished renders, total production minutes or cloud-generation revenue. Without those numbers, claiming mass adoption would be premature.

But “54,000 people saw a cool demo and forgot about it” no longer fits the evidence particularly well.

Is OpenMontage already winning open-source agentic video?

OpenMontage currently has the clearest lead inside the narrow category of open-source systems where an AI agent takes responsibility for producing or editing the video.

A recent Valmera survey verified ten agentic-video repositories against GitHub's API. OpenMontage had 45,109 stars at the time. video-use was second with 19,291, followed by FireRed-OpenStoryline at 3,184 and HKUDS VideoAgent at 1,645.

Across all ten projects, we calculate roughly 71,400 stars. OpenMontage represented about 63% of that total by itself.

The category definitions are imperfect. video-use concentrates on editing footage that already exists, while OpenMontage can create the entire production. VideoAgent comes from a more research-oriented direction. Comparing stars therefore does not tell us which tool produces the best video.

It does tell us which project has captured the category's attention.

Since that 45,109-star snapshot, OpenMontage has kept growing, so its absolute lead is larger now even though several rivals have also moved up. For a category that barely had a recognizable name a year ago, owning that much developer mindshare is significant.

Project	Stars in the verified survey	Main job
OpenMontage	45,109	Produce an entire video through an agent
video-use	19,291	Edit existing footage through an agent
FireRed-OpenStoryline	3,184	Build story-driven videos and edits
VideoAgent	1,645	Research framework for video agents
Remaining six combined	2,150	Smaller specialized approaches

What does OpenMontage do differently from OpenCut, HyperFrames and video-use?

OpenMontage stands out because it tries to own the production decisions above the editing and rendering layers, while OpenCut, HyperFrames and video-use each attack a narrower part of the workflow.

OpenCut is currently much larger overall, at roughly 88,000 GitHub stars. Its core idea is an open-source CapCut alternative: a real video editor across web, desktop and mobile. The team is also building an MCP server and headless automation, so AI agents are clearly part of its future. The center of gravity remains the editor itself.

HyperFrames, currently around 43,000 stars, is closer to infrastructure. HeyGen built it so HTML, CSS, media and seekable animations can become deterministic MP4 video. Coding agents are good at creating web layouts, which makes HTML a clever way for them to create motion graphics and designed scenes.

video-use, currently above 20,000 stars, solves another concrete problem. Give a coding agent raw footage and it can remove filler words and dead space, color-grade clips, generate overlays, burn subtitles and render the result.

OpenMontage can sit above tools like these. The interesting part is deciding that scene four should use stock footage, scene five needs a generated image, scene six should become an animated composition and the final piece needs narration plus captions. OpenMontage wants the agent to make and execute those production choices.

That broader scope makes it more ambitious and also harder to make reliable.

Project	What you mainly ask for	Core job
OpenMontage	“Make this video”	Production orchestration
video-use	“Edit this footage”	Agent-driven editing
HyperFrames	“Render this designed scene”	Deterministic video rendering
OpenCut	“Let me edit this video”	Full video editor

Is OpenMontage cheap enough for people to try casually?

OpenMontage is cheap enough to experiment with almost impulsively because local production is free and users only need to pay when they choose paid models or cloud generation.

The official site currently says local work is never metered. Users can bring their own agent, API keys and providers, and some workflows can rely heavily on local tools, free footage and normal rendering.

OpenMontage gives a useful example on its homepage. Its hypothetical 60-second product video uses screen recordings, stock footage, motion graphics and captions for free, then pays for a small number of generated assets, narration, music and one generated five-second clip. The displayed estimate comes to $2.09.

We should treat that as a product example rather than an average production cost. Someone generating twenty premium video clips could spend far more.

Still, the pricing structure is excellent for distribution. Trying OpenMontage does not require another $30 monthly subscription before the user has made anything. A developer can clone the repository, use tools they already pay for and keep the expensive generative steps to a minimum.

That removes a lot of friction from the “I saw this on GitHub and want to play with it tonight” use case.

Can normal creators use OpenMontage today, or is it still for developers?

OpenMontage is still mainly a developer product today, although OpenMontage Studio is clearly being built to change that.

The open-source route expects a user to be comfortable around coding agents and a local toolchain. The documented setup includes Python, Node.js, FFmpeg and an agent such as Claude Code, Codex or OpenCode. Depending on the production, users may also need provider keys and additional local models.

That is fine for the audience finding OpenMontage through GitHub. It is a poor onboarding flow for a marketing manager who simply wants to make a launch video.

OpenMontage Studio addresses exactly that problem. The current website shows a desktop production board containing the brief, script, storyboard, assets, approvals, spend and renders while chat still controls the production.

Studio remains in private alpha, with broader beta access promised next. So we should separate the product people are excited about from the product most ordinary creators can comfortably use today.

OpenMontage has already proved the developer interface can attract enormous attention. The graphical version is the attempt to turn that idea into something a much wider audience can actually adopt.

Is OpenMontage good enough for real production yet?

OpenMontage looks good enough for real experiments and some real videos today, but we would still hesitate to build a high-stakes production workflow around it without testing the exact pipeline ourselves.

A fresh independent review by MrKeyoor gives us a useful reality check. In a clean Debian environment, the OpenMontage Python installation succeeded and the build passed. The test suite then reported 1,805 passing tests, nine failures, 26 skipped tests, three expected failures and four setup or collection errors.

Some failures came from missing system prerequisites such as FFmpeg and npx, both of which the documentation tells users to install. Others exposed rough edges around package resolution and runtime behavior.

That result is fairly encouraging for such a young project. There is a substantial automated test suite and most of it passed. Still, the test shows how far the experience remains from installing a polished desktop editor and expecting every workflow to behave identically.

GitHub currently shows 86 open issues and 172 open pull requests, along with 448 commits. The repository still has no formal GitHub release. Users are effectively following a fast-moving main branch.

The contributor picture is also very concentrated. Current trackers count only a tiny number of core contributors relative to the project's enormous audience.

So OpenMontage's maturity lags far behind its popularity. That is probably the biggest caveat in the whole story. The project has already reached the attention level of mature open-source software while the software itself is still changing at startup speed.

Can OpenMontage make money without killing what people like about it?

OpenMontage has a believable business model because it can charge people for convenient generation while leaving the open-source engine genuinely useful.

The current Studio plan is straightforward. Studio itself is supposed to be free. Local composition and rendering remain free. Users can bring their own agent, keys and providers. OpenMontage charges when customers choose to generate assets through its cloud, with the price shown before the generation happens.

That preserves most of the reason developers liked the project in the first place. OpenMontage does not need to lock local rendering behind a subscription or cripple the repository to create something customers can pay for.

The harder question is how good the economics will be.

Reselling access to image, voice and video models is unlikely to become a fantastic moat by itself. Providers are interchangeable and inference prices tend to fall. OpenMontage becomes much more interesting commercially if people value everything around the generation: production history, reusable workflows, collaboration, assets, approvals, provenance, provider routing and the ability to move quickly from brief to finished deliverable.

The Studio interface already points in that direction. Its storyboard cards, budget gates, decision logs and editable production artifacts make the workflow itself look like the product.

For now, though, we have no public revenue figure or convincing evidence of paid demand. OpenMontage has proved distribution before monetization.

Will OpenMontage keep growing this fast?

OpenMontage is unlikely to keep adding stars at the current rate continuously, but the project now has a good chance of staying unusually visible because it has already produced two separate waves of attention.

GitHub growth is lumpy. Trending exposure can send thousands of people toward a repository within days, followed by weeks of slower accumulation. Extrapolating 4,400 stars per week into a straight line would therefore be silly.

The more encouraging part is the comeback itself.

OpenMontage survived the end of its first viral period, continued growing and then found another burst of momentum while still shipping quickly. That is a much stronger pattern than a repository that hits Trending once and slowly fades away.

The surrounding market also keeps moving toward OpenMontage's idea. HyperFrames is built specifically for agent-friendly video rendering. video-use lets coding agents edit footage. OpenCut is adding MCP and headless capabilities. Major coding agents are getting better at operating tools and carrying out longer tasks.

OpenMontage does not need today's growth rate to continue forever. It needs “agents can make and edit videos” to become a durable behavior.

Right now, that looks increasingly plausible.

So why is OpenMontage suddenly blowing up?

OpenMontage is suddenly blowing up because it found a very good abstraction for AI video at the moment the rest of the developer world became ready for it: instead of asking users to operate another video tool, OpenMontage lets the coding agent they already use operate the whole production.

The recent acceleration is real. GitFind has OpenMontage adding roughly 4,400 stars in a week after around 1,500 the week before. Forks are climbing alongside stars. Development is moving quickly. The project is also substantially broader than it was during its first GitHub breakout.

GitHub Trending deserves part of the credit, especially for the explosive first wave and the latest rediscovery cycle. But distribution alone does not explain why OpenMontage keeps resurfacing. The underlying idea has become more relevant as Claude Code, Codex, MCP tools and reusable agent skills spread.

The competitive numbers make the same point from another angle. In a verified survey of ten open-source agentic-video projects, OpenMontage held roughly 63% of all stars. OpenCut is larger as a general open-source editor and HyperFrames has also become huge, but neither currently occupies exactly the same position: OpenMontage is trying to make the agent responsible for the production itself.

The biggest uncertainty has moved elsewhere. We no longer need much convincing that developers find OpenMontage interesting. What we still do not know is how many people finish videos with it repeatedly, whether Studio can bring the workflow beyond technical users and whether that attention can become a meaningful business.

The read is pretty clear: OpenMontage's breakout is real, and the second wave makes it more interesting than a normal GitHub viral hit. The project has probably found an early version of how people will make some videos with AI agents. The 54,000-star headline is impressive, but proving that this becomes a widely used production workflow is the next test.