AI Video Tools Roundup, June 2026: 61,000 Stars, Five Jobs, Zero Overlap

HeyGen Hyperframes added 2,036 GitHub stars between May 30 and June 5, 2026. That's 339 stars per day — the kind of velocity that means this project is actively surfacing on developer feeds, "awesome" lists, and shared links, not just accumulating slow organic growth. It's already at 24,577 stars, 13,731 ahead of the next-nearest AI video project in our dataset. Something is accelerating around this tool, and it's happening right now.

That number is the sharpest signal in today's research run, but it's not the only story worth telling. The five most-starred AI video projects on GitHub have accumulated 61,373 combined stars — and they do not compete with each other. Not at all. Each one solves a different problem for a different type of user in a different part of the video production stack. After publishing a pillar overview, three head-to-head comparisons, and now this roundup in the AI-for-video hub, the single clearest insight I've developed is this: most developers pick the wrong tool not because they evaluated incorrectly, but because they never resolved what job they were trying to hire a tool to do.

This roundup is a routing guide. Tell me your use case; I'll tell you which tool to use.

How we researched this

Our research pipeline queried the GitHub AI-for-video ecosystem on June 5, 2026, surfacing the five most-starred repositories in the space. Community discussion threads on Reddit and Hacker News returned empty results, consistent with every prior research run on this topic — these tools resonate in pull requests and Discord servers, not public forums. Official pricing pages for Runway, Pika, HeyGen, and Synthesia were fetched but returned no parseable data.

This article draws from GitHub repository metadata — star counts, descriptions, commit dates — and the cumulative analytical work from prior pieces in this hub: the pillar published May 30, the Hyperframes vs. ShortGPT comparison from May 15, the LiveKit Agents vs. TEN Framework analysis from May 8, and the Hyperframes vs. LiveKit Agents architectural piece from May 22. All star counts in this article are June 5, 2026 figures.

The five tools, mapped to their jobs

HeyGen Hyperframes — Video as a programmatic output (24,577 stars)

GitHub: heygen-com/hyperframes · Updated: June 5, 2026

The description is three sentences, and every word earns its place: "Write HTML. Render video. Built for agents."

The concept is precise: you define a video layout using HTML and CSS — the same markup language every web developer knows — and Hyperframes renders those templates into actual video frames. An AI agent, an automated script, or any programmatic system can populate those templates with dynamic content and trigger the render. The output is a video file. There is no human in the loop. There is no timeline to scrub, no prompt to engineer for each scene.

This architecture is what separates Hyperframes from every other tool in this roundup. It's not trying to help a human make a video faster. It's designed to remove the human from the video production loop entirely, replacing them with structured data and agent logic. Automated product demo videos built from a database. AI news segments that refresh on a schedule. Agent-generated video reports populated with live data. Personalized video emails generated at scale. If video is something your system should produce — not something a person should create — Hyperframes is the only open-source tool in this dataset designed for that job.

The velocity story matters here: it was at 22,541 stars on May 30. It's at 24,577 today. Watching a project add 2,036 stars in six days tells you the developer community is actively discovering and recommending this tool right now. That's not a lagging indicator; it's a leading one.

Pricing: Hyperframes is a HeyGen product. Rendering costs against HeyGen's API, but specific pricing tiers were not parseable from our pipeline. Treat this as "usage-based, verify directly."

Best for: Backend developers and AI engineers building pipelines where video is an automated output artifact.
Not for: Anyone who doesn't write code, or any use case where a human is editing the video.

---

LiveKit Agents — An AI inside a live call (10,846 stars)

GitHub: livekit/agents · Updated: June 5, 2026

This is the clearest category confusion I see in the AI video space: developers assuming Hyperframes and LiveKit Agents are alternatives. They are not. Hyperframes generates video files asynchronously. LiveKit Agents participates in live video conversations in real time.

The use cases are distinct: an AI interview coach that listens to your answer and responds to what you actually said. A virtual customer support agent that joins a help call and can see your screen. A language tutor that watches your facial expressions and adjusts difficulty in real time. An AI copilot that joins your team's video standup. None of these applications produce a video file as output. They require an AI that exists inside a live, bidirectional conversation.

LiveKit Agents provides the infrastructure for that. It's built on top of LiveKit's core WebRTC stack — a project that's been in production since 2021 and underpins real-time video conferencing applications with serious deployment history. That lineage matters at the edges: when something breaks in a live stream at 2am, the inheritance of production-hardened WebRTC infrastructure gives you documentation, community precedent, and debugging surface area that a newer framework simply can't offer. The agents framework itself handles the full audio pipeline — speech-to-text, LLM integration, text-to-speech, voice activity detection for turn-taking — as a composable, plugin-based architecture.

At 10,846 stars and active development through June 5, LiveKit Agents is the most battle-tested open-source path in this dataset for live AI video applications.

Pricing: Core framework is Apache 2.0 open source. LiveKit offers a managed cloud service with usage-based pricing; specific tiers unavailable from our pipeline.

Best for: Applications where a user interacts with an AI in real time over video or voice.
Not for: Generating video files, building async content pipelines, or anything that doesn't involve live bidirectional audio or video.

---

TEN Framework — The AI-first contender (10,649 stars)

GitHub: TEN-framework/ten-framework · Updated: June 5, 2026

TEN Framework occupies nearly identical territory to LiveKit Agents. At 10,649 stars — 197 behind LiveKit's 10,846 — the gap is statistical noise. Both are actively maintained, both target developers building live AI voice and video applications, and both are Apache 2.0 licensed. The surface-level metrics make them interchangeable. They're not.

The architectural distinction comes down to lineage. LiveKit Agents was built by a team with years of WebRTC infrastructure production experience; the AI layer was designed to sit on top of that foundation. TEN Framework was designed AI-first, without inheriting a legacy infrastructure stack. For most teams, this difference is invisible during development. It shows up when you're debugging an ICE failure in a browser that's behind an unusual NAT, or when a specific audio codec combination behaves unexpectedly in a particular network condition. At those moments, LiveKit's infrastructure community and documentation history provides resources TEN currently can't match.

That said, TEN's AI-native architecture can feel cleaner for teams starting from scratch with no WebRTC history. If you have no existing LiveKit infrastructure and you're greenfielding a conversational AI product, TEN deserves a side-by-side evaluation before you commit to either framework. The 197-star gap between them is genuinely not a deciding factor.

Pricing: Open source. No commercial tier data available from our pipeline.

Best for: Greenfield conversational AI products where an AI-native architecture matters more than inherited infrastructure depth.
Not for: Teams already on LiveKit, or any use case involving async video file generation.

---

backgroundremover — One job, done well (7,910 stars)

GitHub: nadermx/backgroundremover · Updated: June 5, 2026

The three tools above are frameworks — substantial development infrastructure that takes real effort to deploy before producing any useful output. backgroundremover is not a framework. It's a utility. It does one thing: removes backgrounds from images and video using AI, via a command-line interface, for free, locally, without an account.

The GitHub description: "Background Remover lets you Remove Background from images and video using AI with a simple command line interface that is free and open source." That's the entire value proposition and it's accurate.

7,910 stars for a CLI utility is a healthy indicator of genuine adoption. You don't accumulate that kind of following by promising something aspirational — you earn it by solving a real problem that real people had, repeatedly. The stars here represent video editors who ran the thing and it worked, not developers who bookmarked it because the concept sounded interesting.

The constraint is the CLI interface. If you don't use a terminal, this tool doesn't exist for you. But for video editors running post-production scripts, developers automating a content pipeline that needs background separation, or anyone who wants to avoid paying a SaaS subscription for a task that runs fine on local compute, backgroundremover earns its place in the stack.

Pricing: Free and open source. You provide the compute.

Best for: Video editors and developers who need free, local, scriptable background removal as part of a larger pipeline.
Not for: Non-technical creators expecting a GUI, or workflows where cloud processing is preferred over local compute.

---

ShortGPT — Short-form automation, experimental (7,391 stars)

GitHub: RayVentura/ShortGPT · Updated: June 5, 2026

ShortGPT targets the largest potential audience in this dataset — content creators producing YouTube Shorts and TikTok videos at volume — and carries the most important caveat in its own description: "Experimental AI framework for youtube shorts / tiktok channel automation."

"Experimental" in a GitHub description is not boilerplate. The maintainers are explicitly telling you this isn't production-ready. For a non-technical creator who wants to press a button and get a polished video, this warning should send you toward the commercial tools instead. For a developer-creator who produces short-form content at volume, is comfortable with Python, and is willing to debug a rough edge — ShortGPT is the only open-source tool in this dataset attempting end-to-end short-form automation: script generation via LLM, narration via TTS, B-roll sourcing, and final assembly into an uploadable file.

The cost model is worth flagging explicitly: the ShortGPT library itself is open source, but operating it requires API keys for the underlying services — an LLM (typically OpenAI) and a TTS provider (typically ElevenLabs or similar). Those costs are variable and yours to manage. At meaningful production volume, they add up.

The 7,391 stars represent real demand for what ShortGPT promises. Short-form content at volume is genuinely labor-intensive, and a working pipeline would save active creators hours every week. The community is clearly interested. Whether the tooling has caught up to that interest yet is the open question the "Experimental" label leaves unanswered.

Pricing: Open source. API costs (LLM + TTS) are user-managed and volume-dependent.

Best for: Developer-creators who produce short-form video at volume and are comfortable with a Python toolchain.
Not for: Non-technical creators expecting reliability, or anyone building a commercial content operation on an experimental stack.

---

Comparison table

Community sentiment data (Reddit, Hacker News) unavailable from research pipeline at publication.

---

Use-case routing guide

The question "which AI video tool should I use?" only resolves once you've answered the prior question: what job does your use case actually require? Here is how I'd route the most common scenarios.

You're building an AI agent that produces video as an output artifact → Hyperframes. Nothing else comes close. 24,577 stars, explicit agent-first positioning, active maintenance, and the HTML-template architecture that maps naturally to agent-generated content. This is the category-defining tool for programmatic video generation.

You're building a product where users interact with an AI live over video or voice → LiveKit Agents, unless you have a strong reason to go with TEN Framework. The WebRTC infrastructure lineage gives you production depth you won't appreciate until you need it. If you're starting fresh with no LiveKit investment, do a genuine side-by-side with TEN before committing — the 197-star gap is irrelevant, but the architectural fit to your specific application isn't.

You need to remove backgrounds from video in an automated pipeline → backgroundremover. It's free, local, requires no API keys, and 7,910 stars of evidence that it works. If you have a terminal, you can use this today.

You want to automate YouTube Shorts or TikTok production → ShortGPT, but only if you're a developer who can accept "experimental" as a real warning, not a modest disclaimer. Read the open issues before committing. If you need reliability today, the commercial tools are the honest answer even though we couldn't pull their pricing.

The use case doesn't fit any of these cleanly → The most useful diagnostic question is: does your use case require generating a video file, or does it require an AI to be present inside a live session? File generation → Hyperframes or ShortGPT. Live session → LiveKit Agents or TEN Framework. Post-processing an existing video → backgroundremover. The category confusion I've seen most often is conflating "AI video tool" as a single category — these tools live in different parts of the stack and the routing question is architectural, not preferential.

---

What we'd use and why

For building AI agent pipelines that produce video: Hyperframes, without hesitation. The 2,036-star gain in six days is the strongest near-term momentum signal in our dataset, and the explicit "Built for agents" positioning means it's being optimized for exactly the use case where nothing else in this roundup applies. If programmatic video generation is your job, this is where the open-source momentum is concentrated.

For live AI in video calls: LiveKit Agents over TEN Framework for teams with any existing WebRTC investment, because infrastructure maturity surfaces at the worst moments and you want the community depth behind you. For greenfield teams, TEN deserves a genuine evaluation — the AI-native architecture can be cleaner for projects that don't need to integrate with existing WebRTC infrastructure.

For anyone who doesn't write code: backgroundremover is the only tool here that's immediately useful to a working video editor — and only if they use a terminal. Everything else requires Python environment management, API key acquisition, and tolerance for developer-grade documentation. For plug-and-play AI video on the consumer side, the commercial tools (Runway, Pika, HeyGen's hosted platform) are the honest answer, even though we couldn't verify their current pricing from our pipeline.

---

Limitations of this analysis

Three gaps matter here. First: the persistent absence of community sentiment data. Reddit and Hacker News searches have returned empty across every research run in this hub. There are no practitioner quotes in this article — no "this broke for me when I tried X," no "ShortGPT worked until I hit Y," no "Hyperframes changed how I think about Z." For developer tooling, that ground-level signal is often more predictive than star counts. When it becomes available, this analysis would shift meaningfully.

Second: output quality is unverifiable from repository metadata. GitHub stars measure developer adoption and maintenance health — not whether Hyperframes produces visually compelling video, or whether ShortGPT's assembled output meets broadcast quality standards. For tools that produce video, the only honest evaluation is running them against your specific use case.

Third: commercial pricing for every major tool in this category is unverified. HeyGen's Hyperframes API pricing, LiveKit Cloud tiers, Runway, Pika, Synthesia — our pipeline returned no parseable data for any of them. Any cost modeling requires going directly to each vendor.

---

Bottom line

Sixty-one thousand combined GitHub stars across five tools that don't compete — that's the structure of the open-source AI video ecosystem as of June 5, 2026. The dominant momentum story is HeyGen Hyperframes, which added 2,036 stars in the past six days and is now the most-starred AI video project on GitHub by a 13,731-star margin. The category is not crowded; it's segmented. Picking the right tool requires knowing whether you're generating video files, participating in live video sessions, post-processing existing footage, or automating short-form content — and the answers to those questions route you to different tools with no real overlap between them.

If programmatic, agent-driven video generation is your job, Hyperframes is the obvious answer. If it isn't, the routing guide above should get you to the right place in one read.