Theres been plenty of speculation about whether generative AI could replaceor perhaps work alongsidehuman creativity. Indeed, many already use popular generative AI tools to create illustrations and even audio and video.But at the same time, other forms of machine learning and artificial intelligence have quietly come to help creative professionals and skilled amateurs shoot photos and videos, record audio, and write prose. Generally, these tools arent creating content themselves; rather, they help users execute their existing creative processes more efficiently and accurately to make new work.Photographers, they dont want hallucinationsthey want things that look like what they saw when they shot the picture, says Bruno Liege, cofounder of photography software company DxO. So we are focusing on enhancing the quality without inventing anything in the image.DxO, based in Paris, has built image processing software for more than two decades, today focusing on tools for photography enthusiasts who work with the raw image files generated by digital cameras. Over the years, the companys scientists have amassed what Liege believes is an unparalleled collection of digital photos taken with a wide range of cameras, lenses, lighting conditions, and aperture settings. The company has used that data to train machine learning systems, built into DxOs PhotoLab software, that can outperform traditional hand-coded algorithms at tasks like accurately detecting colors from the limited data captured by digital camera sensors and eliminating noise from images, even when photographers are working under low-light conditions.Because of our product, they are able to shoot things they believed they couldnt shoot at all, Liege says. They get better quality, and theyre also able to do things they were not able to do before.Liege says the company benefits indirectly from the push for large language models, since it makes powerful hardware and software tools for AI more widely available. But DxOs goals and methods are still different from those of generative AI companies; it still relies extensively on the human eye to verify that its technology is producing accurate photos, even when processing images humans would have previously struggled to make usable.Similarly, in the audio world, Adobe Podcast includes AI tools that can help podcasters and others produce clean-sounding speech from recordings taken in noisy environmentsand configure their microphone and recording setup for optimal quality, even if theyre not expert sound engineers.We started with this idea of, how do we just make things sound good for people who arent audio professionals, or even for professionals who dont have the time or the budget to spend time fiddling [with] the knobs of traditional audio tooling, says Mark Webster, head of Adobe Podcast.Using AI technology developed at Adobes research division, the softwares Enhance Speech feature can strip background noise and echoes from recorded sound, allowing for otherwise potentially unusable sound to be incorporated into podcasts. Feedback from users, who can optionally tell Adobe how well the software worked with their audio samples, has helped improve the tools accuracy, Webster says.A Mic Check feature additionally helps users (or their podcast guests recording audio from remote locations) configure their microphone settings and recording environments to optimize quality.Adobe Podcast also pioneered letting users edit audio using an AI-generated transcript of what people said in recordings, rather than manuallyand, often, tediouslyclipping exact portions of audio they want to use in podcasts. In general, Webster says, the team has thought about how AI-powered automation can help redefine and simplify audio-editing interfaces, although users can export audio to do additional edits in other, more traditional software (including other Adobe tools) if they wish.Weve basically been using the same approach to editing audio for 30 years, and it was all kind of, how do we digitally re-create the confusing interfaces of mixing boards and stuff, Webster says. So its been fun to basically use AI to focus on the important part of the audiothe contentversus having to worry about the characteristics of the knobs and sliders.The software has helped make it possible for new podcasters, hobbyists, and people creating audio for their businesses quickly get up to speed, says Sam Anderson, lead designer for Adobe Podcast.Even fifth graders working on a science fair podcast project can throw something together in a classroom environment, Anderson says. They dont have a professional sound studio, theyre in a noisy classroom, and suddenly you have fifth graders creating professional-sounding podcast content, which is always just incredible to hear.The latest version of the speech-enhancement feature includes the option to reincorporate some background sounds into the audio, once the speech component has been cleaned up, for users who want a more atmospheric sound. And it seems likely that in the near future, part of the creative process will entail deciding what degree of AI assistance (and which AI parameters) makes sense for a particular project.The startup Lux makes iPhone camera apps, including Kino, a video camera app that Apple recently named iPhone App of the Year for 2024. Cofounders Ben Sandofsky and Sebastiaan de With started Lux with the goal of giving users a level of control over their iPhone cameras similar to what theyd have with a professional DSLR (digital single-lens reflex) camera. That led to the companys first product, an iPhone camera app called Halide.We loved using these beautiful, tactile objects that are cameras, de With says. The tactile nature was a little bit lost in the translation to these flat panes of glass we have in our pockets.Lex later built Spectre, an AI-powered app for taking long-exposure photosthe kind that can effectively erase moving crowds or cars from a busy tourist site, or turn nighttime city traffic into artful streams of light. Spectres AI technology can automatically adjust settings for time of day and light exposure, and stabilize the captured image so that it can be used without a tripodcritical for photography in many busy spots where its not practical or permitted to set up a camera stand.But since August 2024, Luxs Halide has included a mode called Process Zero, which actively eschews AI, including systems turned on in iPhone photography by default. That gives photographers more control over the images theyre capturing, as they manually adjust camera settings and try for an ideal shot. Sandofsky says Lux developed the feature after noticing users were already utilizing options to turn down the level of AI in their photos, often looking to make adjustments themselves in editing software after the fact or looking to have a more unprocessed, vintage-style photo.Basically, the thesis around Process Zero is that by having dumber tools, you have more control over the tools, Sandofsky says.Kino is also largely AI-free, which can help beginning photographers learn to avoid common mistakes that the technology will try to compensate for, like shooting an image backlit by the sun, de With explains.If youre trying to learn as a photographer, sometimes you need to be able to get a stinker under your camera, he says. You need to be able to get a shot that just says, oh, what youre trying to do here is impossible, because photographically, you need to worry about the way light impacts what youre doing.But the Lux cofounders emphasize that they have no ideological opposition to AI, which they say can be quite useful in certain photographic situations, and Halide still makes it easy to toggle between different levels of AI assistance. Their philosophy, de With says, is to support AI doing the tedious parts of creative work, so humans can focus more on the elements they enjoy.If we were to be a tool that creates a painting, we dont want to do a painting for you, he says. We want to help you choose a better brush.