Show HN: Tabu, NSFW image and video API for explicit content moderation

1 point by jofo_s 3 minutes ago

Hi all,

I am Jozef and in last wees I was building Tabu on my own. Now I am seeking advice on how to scale it further and also how it works. I would like to join Ycombinator, but firstly want to iterate it and get some more feedback. Is I have no engineering background, I am also considering to look for technical co-founder, but thats the future.

I built Tabu because I was rejected on Apple App Store with previous app I created. The reason was Guideline 1.2 for user-generated content, where users content should be checked for nsfw content. Existing solutions like AWS Rekognition, Google Cloud Vision, Sightengine etc. require heavy devops or hosting own ML models. Thats all increasing costs and slowing down development.

I just needed a simple moderation filter with single request, so I came with idea to built it for others as well.

Right now its quite simple, backend uses a NSFWJS model running in-memory with 200ms latency per image. Buffer is destroyed immediately after classification so no images are stored (but user have an option to turn off privacy-first mode). Then JSON payload is returned with confidence scores across 5 categories: porn, hentai, sexy, drawing, neutral. I also built a dashboard to adjust sensitivity thresholds per image and video. In future I want to expand models to cover more categories and have whole nsfw coverage.

I have free tier with 5000 requests/month so its more than enough to test.

Any feedback, and ideally real test, will be very helpful. I am open for any discussions, calls, getting in touch etc.