When I work with 8-bit images in Photoshop I never achieve to get smooth gradients when working with dark colors. I looks bad and it's a clear line between dark gray and black. My guess is that is because I work 8-bit files but is there any reason at all to not use 32-bit images on the web? Sure, they have a greater file size but is that the only reason? It doesn't have anything to do with web browsers and incompatibility with 32-bit images?

Dani AI

Generated

Banding in dark gradients is a quantization artifact: there simply aren’t enough tonal steps in the final image, so smooth ramps turn into visible bands. The term “8‑bit” is often used loosely — it can mean 8 bits per channel (the usual web case), an 8‑bit indexed palette, or Photoshop’s special 32‑bit float HDR mode. That distinction matters because Photoshop’s 32‑bit float is an HDR working mode, not something browsers display directly. See Color depth and for background.

A practical workflow that avoids the pitfalls described is to do heavy editing in 16 Bits/Channel (Image > Mode > 16 Bits/Channel). That preserves most Photoshop filters while giving far more tonal precision than 8‑bit. When you’re ready to export, reduce visible banding by enabling the Gradient Tool’s Dither option when creating ramps, or by adding a very subtle monochrome noise layer (Filter > Noise > Add Noise — Uniform, Monochromatic — around 0.5–2%). Then flatten, use Edit > Convert to Profile to sRGB, convert to 8 Bits/Channel, and export for the web.

Where possible, avoid rasterizing gradients: CSS or SVG gradients render smoothly on the device and sidestep quantization entirely. Modern compressed formats (WebP/AVIF) also tend to preserve ramps better than aggressive JPEG compression; choose the format that fits your quality-vs-size needs. See MDN’s notes on CSS gradients and image formats: linear-gradient and Image types.

’s suggestion to work at higher precision is sound, and ’s observation about mode limits explains why you saw grayed-out features in 32‑bit Photoshop. 16‑bit is the usual compromise: keeps most tools working, reduces rounding artifacts, and — with dither/noise on export — gives smooth dark gradients on typical web displays.

Recommended Answers

All 10 Replies

Why not design your images locally in 32-bit mode, then convert them to some other format for publishing.

There are a number of different image formats, which allow you to trade quality for size. Pick the one which fits your needs.

That's true, I can always convert them later on. But is there any reason why the standard is 8-bit (or is it even a standard? do I just have a distorted world view?)

If you're using GIF as your file format, then yes you're stuck with 8-bits and that can get pretty crude at times.

By and large, you should be using
PNG - for computer generated images.
JPG - for photographic images.

IE was the last of the browsers to support PNG properly, but that's been fixed for the past 4 or 5 years.

There are several sites (search:IE PNG fix) which allow you to fix PNG in older versions of IE. But check your server logs to see how many browsers connect with IE versions earlier than IE7

But when in Photoshop about every option you can choose with layers are grayed out when working with 32 bit and a great number of filters are grayed out when working with 8bit. So am I limited to 8bit if I want to use any creative built in functions in Photoshop?

Member Avatar for Member #334542

Yes you limited to 8 bit, bcoz it refers to the color management,filters, effects and gradients will not allow you to apply because of referring to lower bits.Image mode will give you the options enabled in higher bits.

Is there any way around this? Like working with 2 psd documents, one 8-bit and one 32-bit, and copy pasting into the other?

Member Avatar for Member #334542

Not Posible, while you try it will ask for the mode change or automatically converted to lower bits I think.

Then how do people work with dark colors and still get good result?

Member Avatar for Member #334542

For that not only the solution is 8 bit or 32 bit. They will also use image works like curves, brightness darkness, illuminations etc. That will give them the output as you say.

Ok, thanks for the good answers, it has helped me a lot! I guess I have no more questions :) thanks again Rajarajan!

Be a part of the DaniWeb community

We're a friendly, industry-focused community of developers, IT pros, digital marketers, and technology enthusiasts meeting, networking, learning, and sharing knowledge.