The first time you boot up Pickle, the camera light blinks on, the app studies your face for a few seconds, and then a version of you, recognizably you, appears in the preview window. The eyes track. The mouth moves when you talk. The hair sits roughly where your hair sits. You wave. The avatar waves a half-beat later, and that half-beat is the entire product thesis: real-time, not rendered overnight.
Pickle, founded in 2024 and headquartered in San Francisco, is building infrastructure for AI-generated personas that behave live on Zoom and Google Meet [WOWTALE, June 2025]. The company closed a $4.35 million seed round in June 2025 with backing from NFX and Krew Capital, after going through Y Combinator's Winter 2025 batch [WOWTALE, June 2025] [TechCrunch, 2025]. The founding group, Sanio Jung, Emmett Kim, Daniel Park, Ho Jin Yu, SJ Lee, and Chris Kang, brings a mix of medical and computer engineering backgrounds [Crunchbase] [Y Combinator].
The bet
Pickle's wedge is narrow and specific: replace the camera feed in a video call with a lip-synced clone of the user, generated and animated in real time [WOWTALE, Feb 2025]. The pitch is for the meeting you have to attend but do not want to dress for, the early standup before coffee, the back-to-back calendar where being on camera is social tax rather than communication. Reviewers who have tested the product describe an avatar convincing enough to fool a spouse glancing at the screen [Tom's Guide, 2026], and Inc. has written about it as a contested answer to Zoom fatigue [Inc., 2026].
That framing matters, because it positions Pickle differently from the obvious comparisons. Synthesia and HeyGen built large businesses on asynchronous avatar video, the kind a marketing team renders for a training module. Tavus has pushed into conversational video. Pickle is going after the live, two-way meeting itself, which is a harder real-time problem and a more personal use case. The product replaces you, in your meetings, while you are sitting there.
Why it could be big
The tailwinds are real. Hybrid work normalized the video call as the default unit of white-collar communication, and the fatigue research that followed normalized the idea that being on camera all day is a cost rather than a courtesy. If a meaningful share of knowledge workers would prefer to send a presentable proxy to a low-stakes meeting, the addressable surface is enormous, because the meetings themselves are not going away.
NFX and Krew Capital are credible backers for a consumer-adjacent AI infrastructure bet, and Y Combinator's W25 cohort placement gives Pickle distribution into the early-adopter developer audience that tends to seed these tools inside companies before procurement notices [TechCrunch, 2025]. The company's own positioning, visible on its site and in third-party catalogs, treats the avatar as the front end and the real-time lip-sync model as the underlying infrastructure that other products could eventually call [Pickle] [VideoSDK, 2026].
Seed round (June 2025) | 4.35 | $M
The team and what they are hiring for
Pickle lists 15 employees in San Francisco [Y Combinator]. The open roles tell their own story. The company is recruiting an Applied AI Engineer and, more revealingly, a Hardware Engineer for something called AI Glass [AshbyHQ, 2026]. That second listing, paired with public references to a focus on AI wearables [pr.ai], suggests the avatar product is the visible layer of a broader ambition: capturing personal context through a worn device and rendering it back through a synthetic self. Whether that hardware ships or stays a research track, it tells you the team is not framing itself as a video-call utility company.
Founders Sanio Jung, Emmett Kim, and Ho Jin Yu maintain public profiles tied to the YC W25 cohort [LinkedIn, 2026]. The engineering bench skews technical and specialized, which fits the problem: real-time facial generation at meeting-grade latency is closer to a graphics and inference optimization challenge than a content pipeline.
What bears say, what bulls answer
The most credible concern is the deepfake overhang. Inc. has already covered the avatar-in-meeting use case as ethically contested, raising questions about consent, authenticity, and the social contract of a video call [Inc., 2026]. Enterprises tightening policy on synthetic media could constrain Pickle's path into the workplaces where the product is most useful. The bull answer, supported by the same coverage and by Tom's Guide's hands-on review, is that the technology is convincing enough that the demand is already there [Tom's Guide, 2026], and that a vendor with first-mover positioning, real-time infrastructure, and a clear identity-verification story can shape the norms rather than fight them. Competitors including Tavus, HeyGen, and Synthesia have larger footprints but are built around different primary use cases, which gives Pickle room to define the live-meeting category before the incumbents pivot toward it.
What to watch
The next twelve months should answer three questions. First, does Pickle's avatar quality hold up at the scale of paid usage, not curated demos. Second, does the AI Glass hire signal a hardware launch, or remain an R&D bet inside a software company. Third, how do Zoom, Google, and Microsoft respond, given that any one of them could ship a native avatar feature and reset the competitive map. A Series A in 2026, if it comes, will likely be priced on whether Pickle has converted demo virality into recurring seats inside companies rather than novelty downloads.
The deeper cultural question Pickle is putting on the table is the one every reviewer keeps circling without quite naming: if a synthetic version of you can sit through the meeting convincingly enough that no one notices, what exactly were we asking the real you to be doing in there in the first place?