Create incredible AI portraits and headshots of yourself, your loved ones, dead relatives (or really anyone) in stunning 8K quality. (Get started for free)
The idea of machines being able to create art may seem like science fiction, but artificial intelligence (AI) is quickly proving itself capable of remarkable creative feats. Apps like Dream by WOMBO and systems like DALL-E 2 can generate astonishingly convincing images and artworks after being trained on millions of photographs and artworks. While AI art generators are still reliant on human-created training data, they are reaching new levels of quality and imagination.
According to Ahmed Elgammal, director of the Art & AI Lab at Rutgers University, "Artificial intelligence is about having software that can do tasks which used to require human intelligence and creativity." Machine learning algorithms are able to analyze formal qualities of art like color, composition, and style and then generate new images based on statistical patterns. The results are often indistinguishable from works by human artists.
While some may see AI art as a threat to human creativity, proponents argue it is simply a new tool. Artist Mario Klingemann uses AI in his Generative Adversarial Network (GAN) artworks, describing it as "an apprentice that I can teach." Others, like artist Refik Anadol, see AI as a collaborator for creating otherwise impossible images and experiences.
Regardless of whether it is viewed as a rival or partner, AI is undeniably reshaping artistic possibilities. Systems like AICAN developed by Ahmed Elgammal have passed "visual Turing tests" by fooling humans into thinking algorithmically-generated art was created by a person. This suggests AI may soon match or surpass human creative abilities in the visual arts.
As AI algorithms generate increasingly photorealistic images of human faces, they are crossing into the uneasy realm known as the uncanny valley. This term, coined by robotics professor Masahiro Mori in 1970, refers to the unsettling feeling people experience when exposed to humanoid objects that appear almost"but not quite"fully realistic.
When it comes to computer-generated portraits, imperfections in rendering subtle facial expressions and skin textures can make images seem strangely artificial and corpse-like rather than lifelike. Even tiny anomalies can trigger an instinctive revulsion, akin to seeing a prosthetic hand that doesn't move quite right. This effect is amplified when the image is of a specific person, as we have innate familiarity with how faces should look and move.
AI portrait systems clearly illustrate the uncanny valley phenomenon. As demonstrated by apps like Reface and Gradient, swapping celebrity faces onto user selfies creates disconcerting hybrid visages. Efforts to generate fictional portraits, like NVIDIA's StyleGAN algorithm, produce unnervingly glassy-eyed results. The generated people simply don't look quite human.
However, some computer graphics researchers believe AI will eventually conquer the uncanny valley by achieving true photorealism. As machine learning models are trained on exponentially more data, including 3D scans of human subjects, algorithms may learn to replicate the nuances of facial appearance and expression.
This could raise new ethical issues, though, if personalized AI avatars become indistinguishable from reality. Services like Hour One already offer AI-generated portraits of deceased loved ones, which some find comforting while others consider it ghoulish. Realistic fictional faces could also be abused for misinformation campaigns.
One frontier in AI portraiture is training algorithms to accurately depict and elicit human emotion. While early AI portraits often featured blank expressions, researchers are now focused on enabling algorithms to analyze and replicate the incredibly nuanced ways humans convey feelings visually.
Being able to synthesize emotional expressions in AI art could allow for portraits that genuinely connect with viewers. However, perfecting this capability requires massive datasets linking facial musculature and features to emotional states. It also demands advanced face generation systems that can dynamically model how emotions morph the face.
Startups like Pinscreen have built 3D facial models and datasets to teach algorithms emotional expressions. Pinscreen's AI avatars can automatically mimic the emotions of real people filmed by a smartphone camera. This technology has been used by Lucasfilm's ILMxLAB studio to create digital humans for virtual reality.
Another pioneer in emotional AI faces is researcher Daniel McDuff at the University of Colorado Boulder. His system uses machine learning to classify emotions in YouTube confession video portraits, then generates new realistic portraits reflecting those emotions. This allows studying how expressions affect human perception of emotions.
According to McDuff, "Our digital humans can help communicate feelings of anger, joy, surprise, fear, disgust, sadness and more...We have an opportunity to use them to explore how emotions are conveyed and improve understanding between people."
However, ethical concerns exist around exploiting emotional data from videos. NYU researcher Timothy Lee warns datasets dating back years may not represent current social norms and values. He argues that training datasets should better represent diversity and only use voluntarily shared data.
As algorithms grow more adept at manipulating emotional expressions in AI art, they raise issues of consent and authenticity. Virtual avatars like those created by HereAfter could be programmed to always smile, while deepfake videos distort emotions and intent. As this technology advances, regulations and transparency will likely be needed to maintain portraits as honest representations.
AI portrait systems are making high-quality photography accessible to everyone, not just professional photographers. Apps like Meero and services like NVIDIA Maxine now empower anyone with a smartphone to generate stunning AI-enhanced portraits. This democratization of photography through AI has the potential to transform self-representation and memory-preservation.
In the past, capturing flattering, well-composed portraits required access to expensive equipment, studios, and expert skills. Professional headshots or nice family photos were costly and time-consuming. Many people lacked the means to preserve visual memories in an aesthetic, personalized way.
AI changes this by acting as an automatic photographer and retoucher. Allison Huynh, Product Manager at NVIDIA, explains how they designed Maxine's AI capabilities to be "inclusive and empowering." The app's camera, lighting, and processing features allow anyone to produce professional-quality selfies using just their phone.
Democratized photography can be especially meaningful for marginalized communities and developing regions. Artist and researcher Natsai Audrey Chieza sees AI photo tools as a way to challenge exclusion in the arts. She highlights how apps like Dream by WOMBO give underrepresented artists access to technology for creative exploration.
Ijeoma Oluo, journalist and author of "Mediocre: The Dangerous Legacy of White Male America," also emphasizes photography's role in social justice. In discussing her AI-generated book cover portrait by Faces of V, she explained how for marginalized groups, "complete control over your public image is really empowering."
Easy access to flattering AI portraits could also benefit mental health and self-esteem. Psychologist Linda Kaye notes how seeing an idealized or enhanced portrait of yourself positively impacts self-perception and confidence. Similarly, the ability to digitally preserve beloved memories can provide comfort and closure.
However, this rapid democratization via apps does raise concerns about data privacy, representation, and misuse. Services relying on datasets with bias could disproportionately alter marginalized faces. And uncanny AI portraits circulating online could enable harassment or misinformation campaigns.
As AI generates increasingly impressive photographic portraits, concerns arise that algorithms could make human photographers obsolete. However, experienced artists argue this is an oversimplification. While AI may transform aspects of photography, human vision and creativity remain irreplaceable.
Pioneer digital artist Mario Klingemann believes that AI should be seen as an assistant rather than replacement for artists. He uses generative adversarial networks to create mesmerizing AI artworks, describing the process as collaboration between human and machine. In photography too, Klingemann sees AI as a tool to expand creative possibilities, not supersede them.
Photographer Denise Matias shares this perspective, saying AI apps help her craft distinctly human art. She states, "I don"t think any sort of artificial intelligence could replace my style and vision." Matias uses editing apps like FaceApp to alter portraits of her subjects, imparting an imaginative, surreal quality. The AI becomes an ingredient in her unique aesthetic.
While AI can generate technically excellent photographic portraits, human photographers provide the contextual understanding and storytelling that creates meaningful art. As photographer Kendrick Brinson explains, "What matters most is the relationship between me and the person I'm photographing." An empathetic human connection allows Brinson to craft emotionally resonant images.
Briana Berglund, a product manager at creative platform Unsplash, argues AI image generation lacks context and intent. She states, "Photography is so much more than accurate images. It"s about someone deciding a precise moment matters." The human eye curates the flood of visual data, elevating points of interest into artful photographs.
Technical skill is also required to execute creative vision. Photographer Samantha Casolari uses AI tools like Lensa AI avatars in her work, but believes the "crucial role of photographers is far from being made obsolete." Casolari's expertise in lighting, angles, composition and editing combines with AI to achieve her desired aesthetic.
As AI art generators create increasingly sophisticated and striking images, questions arise about copyright protections and ownership. When an algorithm is trained on millions of photographs and artworks to develop its artistic style, who holds claim to the new AI art it produces? This issue grows more complex as AI becomes capable of highly creative, imaginative output with minimal human prompting.
OpenAI, creator of DALL-E 2, asserts that any art generated by its system is owned by the user who created it. However, critics argue the company itself should not own broad rights to art produced by its AI. Artist and researcher Anna Ridler asks, "Does that mean OpenAI could exhibit work I've made with DALL-E commercially?" She and others want clarity on individuals retaining commercial rights to AI art they generate.
Compounding this is the question of how much creative credit the AI system deserves. Artist Refik Anadol collaborates with AI, but says the "copyright should be shared between the artist and the tools." Others insist the AI is just a tool, claiming sole copyright like a photographer using Photoshop. But AI can autonomously create novel images unforeseen by users.
Legal scholar Boyden Gray argues current copyright law does not adequately address this tension. Protections were designed assuming human authorship, while AI systems have some autonomous creative capacity. Ryan Abbott, professor of law and health sciences, recommends granting AI partial copyright reflecting its creative contribution.
However, many artists simply utilizing AI tools reject this idea. Artist Kelly Bourdet is emphatic that she holds full rights to her AI artworks, stating: "I created them, not some amorphous AI. The AI was just my collaborator." She argues each image reflects her vision and style, not the AI's.
Some researchers advocate for AI art to enter the public domain, allowing free use and remixing. But for commercial artists like Bourdet, their livelihood depends on preserving exclusive rights to sell AI art they create. Open legal questions around AI authorship threaten these artists' ability to economically benefit from their creations.