Let's Enhance

Image upscaling supercharged

#1 Product of the DayNovember 04, 2017

Let'sEnhance uses machine learning to enhance low resolution JPEG images, boosting resolution 4x times

  • Pros: 

    Results came out way better than expected.


    UI doesn't make the difference between "anti-jpeg", "boring", and "magic" clear. Should have a legend/key for less savvy users.

    Wow. I was skeptical. I know this kind of thing has been around before, but I have never seen it work at this level, especially because artifacts are pretty minimal. I still can't believe the "magic" 4x result I have sitting in front of me, from a 500x400-size starting image. Well done. This is definitely a service I would use.

    B. has used this product for one day.
  • Rodrigo Chiong
    Rodrigo ChiongFreelance Full-stack Web Developer

    Resulting images are great and the upscaling is sometimes mind-blowing! Also it's a free web app, no need to install or compile.


    UI is not the best, saving options aren't clearly labeled. It takes time to process but it's understandable.

    I tried this with some old scanned images, with less than 1000 pixels on each side. The results are amazing and the recovered detail is impressive.

    The tool will give you 3 different options for each file. They are not clearly explained or labeled, but these are my findings with a 872x583 px jpg image:

    "Anti-jpeg" gives you the image in the original resolution but seems to remove jpg artifacts. Resulting PNG is 835kb and 872x583 px.

    "Boring" gives you 4x the resolution but doesn't add much more detail than anti-jpeg. Detail is blurred but it preserves original patterns and edges as much as possible. Resulting PNG is 5.9 MB and 3488x2332 px.

    "Magic" seems to be the best option. It gives you 4x the original resolution and it also enhances edges and details. It looks sharper than the other 2 options but it introduces some noise and slight artifacts (in my case they appear on some patterns in the image). Nevertheless this is the best option of all 3 in my case. Resulting PNG is 13.3 MB and 3488x2332 px.

    YMMV but this is indeed a great tool to try!

    Rodrigo Chiong has used this product for one day.
You need to become a Contributor to join the discussion.
Ryan Hoover
Ryan HooverPro@rrhoover · Founder, Product Hunt
So it IS possible!?
Nikolas Klein
Nikolas KleinHiring@nikolasklein · Product Designer at Figma
@rrhoover Exactly my first reaction as well.
Oleksandr Savsunenko
Oleksandr SavsunenkoMaker@crazy_alex · Head of AI at Skylum
@rrhoover ;-)) we were going to name this thing BoostYourPhoto but few days before alpha launch we found
Simon Chapman
Simon Chapman@simoncchapman · Founder. Need.ie
@rrhoover Oh please!
Andrey Nikishaev
Andrey Nikishaev@creotiv · ML Magician
@rrhoover it was possible 3 year ago)
Jeffrey Wyman
Jeffrey Wyman@jeffrey_wyman · At the intersection of Tech and Business
This is amazing. Should be #1 for today in my opinion! Check out a demo image I used: It's a little hard to see the improvements, but it's definitely there. A little artifacts on the ocean and mountain background, but overall pretty good! I can see myself using this for clients who want potato images on their website
Michal Ptacek
Michal Ptacek@michael_ptacek · founder of Officelovin
Tucker Wash
Tucker Wash@ducktuckgo · Software Developer @ Mobelux
I'm curious how this compares to https://github.com/alexjc/neural... ?
Ivan Pasichnyk
Ivan Pasichnyk@ivan_pasichnyk · CEO, Ecoisme
@ducktuckgo while guys trying to get access to product hunt I will answer from their name: This is pretty good as well, we have different structures of discriminator and generator as well as different training protocol, but general idea is pretty close
Oleksandr Savsunenko
Oleksandr SavsunenkoMaker@crazy_alex · Head of AI at Skylum
@ducktuckgo Finally I got access here ) Yup, @ivan_pasichnyk is right. General idea is there, all the magic is around structure and training. And it always can be better
Chakib Tsouli
Chakib Tsouli@tsoulichakib · Product Designer
[Enhancing Intensifies]
Damien Tsnkff
Damien Tsnkff@tsnkff · Front-End. Sometimes I design
LOVE this one. I was skeptical at first, but once I tried it I was amazed. Good job team 💪🙌