I like the idea, good luck! But I uploaded a very short sound effect (2s) instead of music, hoping it would complete faster but 15 minutes later and still waiting... A cancel button would also be nice. How do you render the video on the server? Do you open selenium and take screenshots of each frame and then stitch them together using ffmpeg?
Hi, sorry, looks like the server was down because it couldn't handle the load. I am currently running a "poor man's setup", everything is on EC2 instance. I am not so big on DevOps side of things, so I have an API that does all the FFmpeg processing and the frontend calls that API. I am looking for a better way to do things and been considering Lambda functions too. I would love to hear if you have any tips on handling a long processing job like this.
I've worked on something similar for a project of mine and used AWS Lambda which has been nice because it scales well and didn't cost me money while it wasn't running. The limit is 15 minutes which should be fine for something like this.
As others mentioned, I would go with a client-side approach if it works for you though.
I don't recommend lenovo laptops. Last year I bought a thinkbook 15 g4 and have been struggling with performance issues from day one. Lagging, stuttering, low fps, thermal throttling etc. Tried everything but couldn't fix it. Not possible to overclock or undervolt. Very poor performance and poor tech support.
I have always had good experience with Thinkpad T series. But anything outside Lenovo's business range laptops is usually terrible. It's like an entirely different company! Had a terrible experience with IdeaPad and it died just after a year :(
It would be helpful if your tool had a unique feature that differentiates it from competing tools, but even if you came up with such a unique feature, its going to be copied eventually. There are so many Link in Bio tools these days and all seem to be more or less the same. But competition isn't necessarily a bad thing. Despite the competition, I've seen social media profiles use different tools, not just the best out there. Even if you had built the best tool, its not going to make much of a difference if no one knows about your product. Maybe you could focus on go-to-market strategy. You could narrow down your target audience and redesign it for a specific audience instead of something too general. Something just for artist, designers, musicians, doctors, lawyers, students etc. and try to get one influencer who has similar followers to use your tool in their bio. You should drop a link here to get more feedback.
It's no secret. OpenAI isn't the only player in the field. The whole world has seen the power of Generative Pre-Trained Transformers and many organizations and even governments are going to develop and fine-tune their own custom LLMs. Of course it demands huge amount of training data and computing power but that's no secret.
It could do with a heavy perf pass tbh. Haven’t had time juggling all the feature work, but it’s planned for 1.0. Mind letting me know what machine you’re on?
