Microsoft is testing an experimental feature that literally puts a “face” to its assistant: it is called Portraits and it is coming to Copilot Labs with 40 stylized human avatars capable of supporting voice conversations in real time. The goal isn't just to beautify the interface: the company wants to reduce the friction of speaking to an AI and make the interaction more natural, especially for those who feel more comfortable addressing a face.
This experiment leverages Microsoft Research technology to generate fluid facial animations from an image and audio, with Live lip-syncing, expressions, and head movementAt the same time, Microsoft has raised several safety barriers: the portraits are not photorealistic, there are usage limits, and the feature is limited to adults and a few countries for now, with a controlled rollout to learn from real-life user experiences.
What exactly is Portraits in Copilot?
Portraits is a new option of Copilot Labs which allows you to choose from over 40 human characters with varied styles, from 2D designs to 3D options, all intentionally stylized so as not to cause confusion with a real person. Each avatar displays gestures and microexpressions during dialogue, contributing to a sense of presence while talking to the voice assistant.
The idea stems from a usage pattern: some users have commented that they feel more relaxed and confident if they can “look” at someone, even an avatar, when interacting with AI by voice. That's why Microsoft has decided to experiment with a face that reacts in real time to the conversation, instead of the classic bubble or abstract avatar of always.
As shared by the Copilot team—and picked up by outlets like The Verge, as well as messages on Copilot's Discord server—these portraits offer a higher degree of naturalness in facial mimicry, reinforcing speaking turns and the emotional nuances common in a conversation. It's not just the mouth: the head, the gaze, and the expression contribute to the feeling of fluency.
As a curious detail, among the variety of designs there are very diverse options by race, gender and age; some users have pointed out that one of the avatars resembles “Coco's grandmother” for their aesthetics. Other portraits are more modern or minimalist, and there are differences in stylization that range from the almost cartoonish to the semi-realistic, but always far from photorealism.
To personalize the experience, the user can link the portrait to a voice. This combination allows for a closer "conversational partner" feel, without losing sight of the fact that we are dealing with a AI with visible indicators that it is not human, something that Microsoft has put at the heart of the design.
The technology behind: VASA‑1 from Microsoft Research
Portraits is powered by VASA‑1, a technology developed by Microsoft Research that generates talking faces and Realistic facial animations from a single image and audioOne of its strengths is that it doesn't require complex 3D modeling: the system synthesizes lip sync, head movements, and expressions in real time.
This approach allows Copilot's portraits to respond fluidly in conversation without incurring the lengthy preparation times typically required by traditional 3D pipelines. In other words, Low latency is part of the design, something key so that the sensation of “talking to someone” is not broken by waiting or visual stuttering.
Microsoft introduced VASA‑1 over a year ago as a research milestone, and now it's putting it to practical use in Copilot. The decision to maintain a stylized—and not photographic—aesthetic isn't a whim: the company is seeking a balance between expressive naturalness and ethical clarity, avoiding the illusion of a real person speaking on the other end.
Among the early impressions shared by testers, it is highlighted that the portraits adapt well to voice inflections and handle the lip syncing and gestures, even when switching between languages. Although it's still a prototype, feedback suggests the technical foundation already enables fairly believable interactions.
Unlike other approaches on the market, VASA‑1 is not aimed at celebrity deepfakes or hyperrealistic recreations, but at AI-assisted conversations with social cues enough to make the dialogue more human, but without crossing the line into “looking” like a person.
Microsoft's position: naturalness yes, illusion of humanity no
Mustafa Suleyman, head of Microsoft AI, has underlined the objective of this experiment: to make it easier for those who prefer to speak to "a face" to do so, but without falling into the trap of photorealism. In the words of the executive, we must Building AI for people, not a “digital persona”The risk of designing systems that are indistinguishable from humans is that users end up attributing consciousness or even rights to them.
Along these lines, Microsoft makes it clear that the portraits are “intentionally designed” to not being photorealisticThey also include visual alerts indicating that you are interacting with an AI, so the experience is transparent and boundaries are clearly marked for all types of users.
The company wants to learn in the open. Suleyman points out that this feature is a prototype in development, a controlled trial to understand How users perceive a conversation with an AI “with a face”, what it offers them, and where the risks lie. Hence, the decision to launch it in a restricted and progressive manner, with visible controls and daily limits.
The approach harmonizes two often conflicting goals: increasing the closeness of the conversational experience while simultaneously preventing the system from being perceived as a person. On that frontier, transparency—AI signatures, visual style, and boundaries—is the pillar Microsoft wants to maintain. as a guarantee of security and confidence.
Availability, requirements and first tests
Portraits is part of Copilot Labs, so it is in an experimental phase and has limited access. It can currently be tested in the United States, the United Kingdom, and Canada, and only for users over 18 years oldMicrosoft has also implemented daily and session limits to limit usage while gathering feedback and measuring the impact.
Access is via the Labs section in Copilot. Some how-to materials shared by the community note that enabling the feature may require a Microsoft account and, in certain cases, a Copilot Pro subscription$20 per month is mentioned in some guides, although the cost and requirements may vary as the trial progresses.
The marketing process includesseveral phases that are reflected below: Boot It's simple: if the option appears in your Copilot Labs, just press "Try now", choose a portrait from the available catalog and assign a voice to itFrom there, you'll be able to have real-time voice conversations with the AI, which will animate the chosen face with mouth, head, and gesture movements that match the audio.
Initial tests have garnered positive feedback from those who wanted a less “cold” interaction than text chat. The following stand out: agility of language switching and the overall naturalness of the gestures. As with any prototype, there are still details to fine-tune: some voices sound better than others, and some portraits work better depending on the context.
According to Microsoft, availability could be expanded with There if the experience meets expectations and the results are confirmed usability and satisfaction benefits. There are no public dates for a global rollout yet.
From Copilot Appearances to Portraits: What Changes
Before Portraits, Microsoft had already tested the waters with Copilot Appearances, a previous proposal presented in July that featured animated cartoon-like figures or "blobs" to mark emotions and conversational turns. That approach was more conceptual and less human in its aesthetic.
The big difference now is the use of VASA‑1 and the jump to stylized human avatars that better maintain coherence between voice and gesture, rather than an abstract character. This serves to enhance the sense of naturalness without entering into photorealism, a boundary that Microsoft says it doesn't want to cross in this context.
Community sources and specialized newsletters have speculated about a possible future convergence between these portraits and concepts such as Copilot Characters or Live Portraits, as well as integrations related to OneDrive or "Pages/Memories" features. For now, all of this is undated and remains experimental and prospective.
In parallel, some interviews and opinion pieces have gone further, talking about a Copilot with a "persistent identity," extended memory, or even the avatar "aging" over time. Although they sound suggestive, these ideas should be taken as a long-term visions or hypotheses, not as features officially announced for Portraits today.
Market context: attendees with faces and visual trends
The industry is moving toward multimodal assistants with visual presence. Competitor Grok, for example, has introduced 3D avatars with styles ranging from anime to adult options (NSFW). In contrast to this approach, Microsoft insists on a controlled, non-photorealistic aesthetic, with clear usage limits and warnings, trying to avoid problematic drifts.
The strategic difference is evident: prioritize the trust, security and usefulness Above the flashy. Hence the age- and country-specific launch, and the collection of usage data to iterate cautiously. Microsoft's ambition is for Portraits to complement real-life tasks, not become an end in itself.
Potential use cases include internal training, customer service, and accessibility for those who prefer the voice channelThe presence of a face that nods, emphasizes, or displays emotional nuances can help understand instructions, reinforce empathy, and minimize the feeling of speaking “into a blank screen.”
Of course, the real value will depend on how Portraits integrates with the rest of the Microsoft ecosystem and whether the visual interaction contributes measurable improvements in satisfaction or efficiencyThe experiment will serve precisely to measure it in specific scenarios.
Security, Privacy, and Boundaries: The Game Framework
Microsoft has implemented daily and session limits for the use of Portraits, in addition to unambiguous visual indicators that you're dealing with an AI. The idea is twofold: to prevent abuse and to set clear expectations for the type of interlocutor.
The fact that the portraits are stylized and not photorealistic reduces the risk of misunderstandings and potential malicious use. In a context where some platforms with avatars have suffered problems - from inappropriate content to attempts to simulate real people—, this design decision introduces barriers to deception.
In terms of privacy and regulatory compliance (e.g., GDPR), controlled deployment makes it easier for Microsoft to strengthen safeguards before scaling. The company wants to ensure that interactions don't get out of control and that transparency signs are visible at all times.
The Labs experimentation model allows you to iterate on real feedback, adjust limits and, if necessary, correct drifts in useThe combination of visual style, usage controls, and visible warnings forms the initial security perimeter while the impact of the feature is assessed.
Business adoption: potential and frictions
In companies, the maturity to incorporate assistants "with a face" is uneven. Innovative sectors—technology, digital, R&D or customer experience teams— can see value in the closeness that portraits provide: better onboarding, more enjoyable training, or guided support with social cues.
In more traditional sectors – banking, insurance, industry or administration – AI is still largely associated with pure and simple productivityHere, an emotional interface can be perceived as unnecessary, frivolous, or invasive if its usefulness isn't explained with numbers. The demand for short-term ROI imposes barriers to face-to-face experiments.
In addition, there is a perception challenge: providing the assistant with gestures, voice and memory can generate distrust if you do not communicate well What it's for and how it protects data. Regulatory compliance (e.g., GDPR) is especially important when it comes to sensitive conversations or personal data.
There is also the challenge of integration. Many organizations are not technically or culturally ready to work with agents that manage voice and non-verbal signalsThis involves changes in processes, experience design, and sometimes in the skills of the team orchestrating the AI.
- The value proposition must be clear: better service, less friction or greater retention.
- Compliance and governance must be up to date: controls, auditing and transparency.
- A limited pilot is advisable to measure real impact and calculate ROI.
Relationship with other projects in the Copilot ecosystem
While Portraits explores the visual layer of conversation, Microsoft pushes other lines to expand Copilot's reach, such as DALL·E 3 with Copilot. An example is Gaming Copilot, an assistant designed for PC gamers that integrates into the Game Bar of Windows and on devices like ROG Ally.
This assistant can interact by voice or text, offering Tactical recommendations, mission assistance and strategies in real time. It includes a “Mini” mode that stays on screen without interrupting the game and push-to-talk options for quick queries, designed to follow the flow of the game without taking you out of the action.
According to Microsoft, Gaming Copilot cross-references user profile information with public Bing data to tailor responses to the exact context of the game, with the possibility of content creators provide recommendations in the future. Although it's a different field, it reflects the same approach: contextual, useful AI with a polished interface.
quick questions
- How many portraits are there? More than 40 stylized human avatars with a variety of styles, ages, and aesthetics.
- Is it photorealistic? No. They are designed to avoid confusion with real people and maintain ethical clarity.
- Where is it available? Currently available in the United States, the United Kingdom, and Canada, and for those over 18.
- Do I need Copilot Pro? Some guides mention that it may be required in certain cases; access is managed through Copilot Labs.
- How does it work technically? With VASA‑1, which generates facial animation, head movements, and lip sync from an image and audio.
- What is? To make voice conversation more natural, reducing the coldness of the chat without giving the impression of talking to a human.
Portraits in Copilot is a measured step towards more human conversational interfaces without crossing the line into photorealism: avatars that gesture, lip-sync, and nod naturally for those who prefer to speak to a "face" rather than an empty box. With VASA-1 as the driving force, clear usage limits, visible warnings, and restricted deployment, Microsoft aims to test the extent to which a visual presence improves the experience and trust, and in what real-life scenarios—from home to work—this layer can add tangible value.
Passionate writer about the world of bytes and technology in general. I love sharing my knowledge through writing, and that's what I'll do on this blog, show you all the most interesting things about gadgets, software, hardware, tech trends, and more. My goal is to help you navigate the digital world in a simple and entertaining way.