Show HN: Open-source AI video editor (github.com)
It all started as an internal experiment but as we tackled some of the issues it was clear there could be some value sharing it with the open source community.
Some of the key points and tech stack details:
- It uses IndexedDb, so all data is local (i.e. no auth, no cloud db)
- Multiple AI models for video, image, music and voice-over. APIs are provided by fal.ai
- Built with the typical React+Next.js, Shadcn front-end
- Used remotion.dev for the realtime video preview (this is such a great project, without it the codebase would be twice as large)
- File uploads so you can bring your own media by uploadthing.com
- ffmpeg for encoding the final video and also some ui tricks, like the audio waveform
We deployed a version of it and for now it's free to use. We do plan to add some rate limiting and a bring your own API Key next, but it's open source and I'm curious about what the community will build on top of it, or derive from it. Customize your own video app and if you do, please share.
If you have any questions, hit me up!
61 comments
[ 485 ms ] story [ 2930 ms ] threadA small bug report: if you have both a still image and a video in the timeline, the video does not play in the preview.
Some eta on how long a generation will likely take.
We will keep improving the UI, including shortcuts. Thanks a lot for the feedback.
They think they've got a leg up on their model partners. That the foundation model companies will mostly become generic copies of one another, servants of the compute layer. There are so many foundation video models now, and they'll battle it out over dwindling margins. Pika, Runway, Kling - they're all the same. And there's also growing open source foundation models.
The thing that stands in Fal's way is that the future of AI video for artists is local. Hunyuan and Comfy can run on desktop machines, and real artists hate the SaaS model and non local stuff. It doesn't look like we'll even need A100s to reach Pixar levels of quality.
The ones to watch in the art space are Comfy and Invoke. And Adobe.
Fal probably has a future in powering the API-driven YouTube Shorts slop, though there's probably an upper limit to the number of talking head startups.
But there's no way they win artists over to cloud. Sophisticated tools are going to be local. Any SaaS art tools that do achieve escape velocity will buy and manage their own compute and won't go through Fal as a reseller.
Seems to me like you're not watching a growing trend.
I'm deeply embedded in this community and comfy is the biggest thing in it.
https://trends.google.com/trends/explore?date=today%205-y&q=...
https://trends.google.com/trends/explore?date=today%205-y&q=...
https://trends.google.com/trends/explore?date=today%205-y&q=...
Everyone is tired of paying for generations they don't use. Everyone wants local.
Local isn't some perfect little fisher price UI that some product manager approved. Local can be mixed and matched and pushed to the limits.
Topaz Labs' run rate is more than Runway and Pika, so don't write off local tools.
Internet creators are not "everyone". 99% of people aren't going to waste their time setting up ComfyUI instances and screwing with github nodes.
As I said, which you haven't disputed, Midjourney's revenue is orders of magnitudes more than any of these tools.
You don't know.
Midjourney might have tons of revenue for being the first mover, but they're busy building VR headsets and generative world models for their metaverse thing.
Meanwhile everyone who is using these models for work - ie, making money with these tools - is clamoring for local execution and deep controllability. Midjourney won't give them that.
The top AI studios are all building on Comfy.
Don't write off Topaz, either. It's the last stage in the pipeline for everyone's workflow.
Stable Diffusion came before Midjourney. What first mover advantage?
Flux has been around almost a year, by your reasoning everyone should have switched. They didn't. Everyone who used Midjourney a year ago is still using it.
The Midjourney sub on reddit has 1.6M subs. The Stable Diffusion sub has 600k.
I don't care what AI studios you know want to use. They use Premier Pro. Just like every film school in America does. Despite plenty of open source alternatives.
90% of people don't give a shit about running locally as long as it's high quality, easy to use and not expensive. And that's exactly why Midjourney is dominating.
Oh, and Runway is projecting to hit $150M in ARR this year.
You've got your facts wrong.
> Flux has been around almost a year
It's been around half that time.
And you're also underestimating how many billions of images have been generated with Flux. The marginal cost and value of these models will trend towards zero.
> The Midjourney sub on reddit has 1.6M subs. The Stable Diffusion sub has 600k.
So you're saying that they're within the same order of magnitude?
> I don't care what AI studios you know want to use. They use Premier Pro. Just like every film school in America does. Despite plenty of open source alternatives.
Premiere is local software. There's also Final Cut and lots of other alternatives. They're not using CapCut to edit their films.
We're also super early in this game. You're mentioning 20+ year old mature software.
> 90% of people don't give a shit about running locally as long as it's high quality, easy to use and not expensive.
They absolutely do care.
Right now 9 out of 10 generations are garbage, and that's being generous.
> Oh, and Runway is projecting to hit $150M in ARR this year.
That's a really ambitious target they set for themselves, and I think they're getting ahead of things. The field is full of much better competition now.
They also failed to raise their ambitious last round. They've raised too much and have too high a valuation, meanwhile smaller companies have caught up.
Figma destroyed Photoshop & Sketch.
Whatever, you have your opinion, none of the facts today support it, maybe you'll be right in the future (you won't be).
But also makes sense. If you wanted e.g. music, you could either look at catalogue of what's out there or try to make your own. But if no one is making what tickles you and you yourself have no talent to make what you would like to listen, the AI comes to rescue.
People don't want to watch two chess computers play even if they can both smoke Magnus Carlsen.
Art is an even more extreme version of this. Part of what makes a Dali painting a Dali painting is the mimetic desire of the general weirdness of the human Salvador Dalí.
AI can't displace human mimetic desire. AI art is just worthless muzak playing in the background at Dennys. Non-artistic/non-creative people right now are amazed at how they can make muzak but muzak is shit by its very nature.
Anyone can replace the AI layer with their own local models, other services... whatever suits your use case and preferences is fair game.
You can replace anything, deploy on your own server, port it to other stacks... whatever brings value to you.
We're also open to PRs, cut an issue in the repo and we can get the conversation going.
https://news.ycombinator.com/item?id=42789451
https://news.ycombinator.com/item?id=42742537
https://news.ycombinator.com/item?id=42742525
That's not what this site is for, and destroys what it is for, so we end up having to ban such accounts.
If you wouldn't mind reviewing https://news.ycombinator.com/newsguidelines.html and taking the intended spirit of the site more to heart, we'd be grateful.
It's a quite complex UI, so it's not easy to export it as a single component.
Just seeing the interface gives a lot of clues as to potential usability as if it's targeted at novices or hardcore editors. Not being able to see if there's any kind of lag when trying to scrub/navigate the video is also something that tells me if I'm going to get upset with it from being web based or if it feels native.
Just food for thought on what people might be looking for instead of having to install something that might not be right for them.
Edit: What do the buttons next to the play/pause button do? Typically, I'd expect to frame advance, or jump to next the next clip. I click and see nothing happening. I dropped two images I uploaded into the timeline, yet there's nothing visible in the timeline. The monitor shows something, but nothing in the timeline indicating how the clip fits in the timeline. There's not indication of what the frame size of the timeline is. The images look like they are being stretched/squeezed to fit, but there's also no indication in the media bin about the media like duration/size/etc.
This really feels targeted for people with no previous knowledge of a video editor. It also just has that feel of "do something quick to make a post with it's AI to HN".
As for the feedback, fair enough. As I said, this is just a demo and it's in its early stages. It has no goal of replacing professional video editors or even matching their features. Some annoying bugs and missing features, we will tackle them.
As for your last comment, feel free to judge, but I can say it was not quick at all to do this just to post on HN. You can check all the commit history and the codebase so you can have an idea how much care was put into it.
This is a very premature demo that does more harm than good in promoting your product. If you said we built a tool that makes it easy to generate content, then you'd have a much more interesting product. Tacking on claims of this being a video editor is extremely disappointing, as the most basic of abilities of a video editor are missing. This is what brings on the criticism of being a rushed product launch just for the PR postings.
scrolling frame by frame through hours of hundreds of tiny videos, fixing their exposure, fixing their dropped frames, fixing transitions, fixing and matching visual narration to textual and audio narration, aligning music to video, making "vlog" like templates. there is just so much to be fixed in this industry, but everyone is concerned with generating assets. that's just not the biggest issue video editors are facing. like not at all.
You can sign up at https://www.video-jungle.com/ and I'll send an invite over.
Remotion does look really cool for in-browser video editing. People should note that you can’t make free-to-use software with Remotion. It is a paid product, and you can download and run the code for free in very limited circumstances that let you experiment with it. You can’t use the output of your own software built on Remotion for any for-profit purpose, nor can anyone else, without a license.
During the days of oss.incredible.dev we found that the biggest friction for our user in video production was themselves, It's extremely hard for anyone to come in front of a camera and record themselves, while we were able to solve for everything else like scripting, story boarding, layouts, animations, collaboration, block based recording etc it was the inability to record themselves that caused them to leave the workspace hanging. With AI, users won't have to record themselves, they can have a saved avatar and just give it a script to make it work! This was the missing puzzle.
I'm planning to revive oss.incredible.dev, but with the intention to make it AI first product. Let me know if anyone here would like to join hands.