NewFace combines a node-based canvas, reusable ad formats and third-party video models for sellers and agencies.
By RuntimeWire Staff · Published
Primary source: PR Newswire
Why it matters #
LibAI Lab is using NewFace to move from consumer creative tools into recurring ad production, where workflow ownership may prove more durable than access to any single model.
Yong Tang, the computer-vision researcher behind Cutout.pro and PromeAI, is moving LibAI Lab closer to the advertising budget. LibAI Lab launched NewFace on August 20, pitching an AI production workspace that turns product photos, reference videos and written briefs into e-commerce images, UGC-style clips and batches of short-form ads. (prnewswire.com)
The launch extends a product portfolio that has largely centered on general-purpose visual creation into a market where customers produce creative continuously and can measure the resulting sales. LibAI Lab co-founder Jefferey T. described the thesis in the company's launch announcement: "Creative teams need more than access to powerful models - they need a system that understands intent without taking away control." (prnewswire.com)
Tang brings a researcher's background to that orchestration problem. A Visual 1st biography identifies him as LibAI Lab's founder and director, with a 2007 Ph.D. in computational science and civil engineering from Pennsylvania State University. His work has spanned computer vision, large language models, optimization and generative-AI applications for e-commerce, architecture and interior design. Before LibAI Lab, he built Cutout.pro and PromeAI. (visual1st.biz)
From visual utility to campaign system
NewFace's main interface pairs a conversational agent with a node-based canvas. A user can describe a campaign in plain English, and the agent breaks the request into connected production steps, retaining the original brief, reference material and prior choices. The user can approve changes one at a time, let the agent continue independently or edit individual nodes. (prnewswire.com)
That design reflects the commercial bet beneath the launch. Model access is becoming interchangeable as creative platforms offer overlapping menus of image and video generators. NewFace is trying to own the project layer above those models: the brief, shot selection, creative structure, revisions and handoff into a finished campaign.
LibAI Lab calls its reusable workflows "Skills." Viral Remake applies the storyboard, pacing and opening hook of a reference video to a user's product. Batch Product Video produces as many as 10 short clips in one run, giving marketers multiple creative angles to test. Users who already know the exact asset they need can bypass the canvas and open separate tools for image generation, video generation, video-to-audio conversion or product-detail-page assets. (prnewswire.com)
The product page says general video generations can run for up to 15 seconds, while projects using the Remake Skill can exceed that limit. The launch announcement also says NewFace has added Seedance 2.5, ByteDance's video model, for single generations of up to 30 seconds. ByteDance introduced Seedance 2.5 on July 31 with 30-second generation, multimodal references and timestamp-level editing. The different limits appear to depend on the workflow and model selected rather than a single platform-wide ceiling. (newface.ai)
LibAI wants to become part of the stack
NewFace also includes a command-line interface and a Skill that Pro and Max users can call from agent environments including Cursor, Codex, Claude, Gemini and Perplexity. A customer's agent can hand a generation job to NewFace and pass the output into another step, making LibAI Lab's product a production service inside a broader automated workflow. (prnewswire.com)
That integration matters because the canvas alone puts NewFace in direct competition with a long list of browser-based creative tools. An API and callable Skill give LibAI Lab another route into agencies and technical marketing teams that want to automate asset production without moving every job through a separate interface. NewFace's Team Plan adds shared creation, centralized credits and project controls for organizations producing at higher volume. The launch announcement says new users signing up this month can claim up to 160 free credits.
LibAI Lab enters this market with some evidence that it can attract a broad creative audience. Andreessen Horowitz's sixth Top 100 Gen AI Consumer Apps list included Cutout among its 50 most-used web products. LibAI Lab says Cutout appeared in every edition of the ranking from September 2023 through March 2026, though the ranking is based primarily on third-party web traffic and mobile usage rather than LibAI Lab's revenue or customer retention. (prnewswire.com)
The ad generators already have money
NewFace is arriving after several rivals raised capital around the same premise: marketers need many short videos, and generative AI can make testing them cheaper. Creatify said it raised a $15.5 million Series A in May 2025. Arcads said it raised a $16 million seed round led by Eurazeo in December 2025 after building a product around AI actors, product demonstrations and localized marketing videos. (creatify.ai; arcads.ai)
LibAI Lab's proposed distinction is control. NewFace combines the canvas, persistent project context, reusable creative structures and batch output instead of centering the experience on an avatar library or one-click product-link conversion. That approach could suit teams that want automation without surrendering every production decision to a black box.
The commercial proof still has to come from the ads. LibAI Lab describes its Skills as informed by high-performing apparel, skincare and product-showcase content, but has not published campaign benchmarks tying those workflows to conversion rates or lower acquisition costs. NewFace launches with a broad toolset and Tang's track record of shipping visual-AI products. Its next test is whether sellers treat the canvas as a daily production system rather than another place to sample the newest video model.