In 2025, AI is revolutionizing video production at an unprecedented speed.

Late last year, an AI-generated parody video titled "Vacation on Epstein Island" went viral on Platform X. It featured lifelike visuals of celebrities like Michael Jackson and Diddy on a tropical island, amassing over 1 million shares within just 24 hours.

Few realized that the driving force behind this viral sensation was Higgsfield, an AI video startup founded merely two years ago. Amidst fierce industry competition, Higgsfield stands out with its 'creator-centric' business strategy.

Higgsfield recently announced an $80 million follow-on funding round, bringing its Series A total to $130 million and valuing the company at $1.3 billion, thus earning it the status of a new unicorn. Its growth trajectory has been nothing short of explosive: 15 million users in 9 months, 4.5 million videos generated daily, and annual revenue soaring to $200 million in just two months.

Higgsfield's success is largely attributed to its precise targeting of users. According to Reuters, 85% of its users are social media marketers who primarily leverage the platform to create branded content, short video ads, and marketing materials.

Its core strategy is evident: identify creators and brands with commercial monetization needs, define clear B2B payment demands, and meet them with comprehensive tools deeply integrated into professional workflows, offering advertising-grade control. Through cross-model scheduling, Higgsfield ensures sustained performance leadership. This closed-loop system of 'creation entry + workflow + distribution incentives' builds long-term revenue potential.

Today, let's delve into this new unicorn in the AI video sector.

/ 01 /

Building the Premier 'Cinematic' AI Video Production Pipeline

In 2025, AI-generated video is on the cusp of explosive growth.

The film, advertising, and short video industries have stringent demands for visual storytelling: continuous content production, evolving aesthetic standards, and consistently substantial budgets. Investors recognize that if AI systematically reconstructs video production as it did with images, the value of entry-level products will be immense.

Higgsfield is positioned at this pivotal juncture.

Unlike typical 'video generation models,' Higgsfield's narrative has shifted from 'can it generate video' to 'how to serve genuine AI video creators.'

Simply put, Higgsfield functions as a comprehensive AI video workflow tool for creators and marketing teams.

In Higgsfield's official descriptions, keywords such as brand creativity, marketing workflows, and content collaboration repeatedly emerge. The aim is to unite brands and creators in a shared workspace, collaborating around modular AI video tools.

The platform's centerpiece is a workspace called Canvas, the core area for user creation and collaboration.

Here, creators design shot sequences and camera movements, manage visual structure and style uniformly, and consolidate multiple creative stages into a single workflow.

Furthermore, Higgsfield integrates a 'multi-agent collaboration system,' akin to a virtual film crew:

The screenwriting agent handles narrative structure, the director agent controls emotional pacing, and the cinematographer agent decides camera movements and positions.

For users, Higgsfield eliminates the need for complex prompt engineering, offering multiple intuitive entry points for diverse needs.

For creative explorers, the 'Paint-to-Video' feature is available.

Simply sketch a subject's outline on the canvas, annotate direction and speed with motion arrows, and indicate camera intent with surrounding arrows to transform sketches into dynamic shorts. After generation, motion effects can be fine-tuned by adjusting arrow curvature and line thickness, achieving professional-grade tracking or surround shots.

For efficiency-focused professionals, text, image, or sketch-to-video conversions follow a unified controlled workflow: ① Define style and base movements; ② AI automatically matches camera motions.

The platform provides a library of over 50 preset professional camera movements, from dolly pushes and FPV spirals to 360-degree surrounds and bullet time, all callable with a single description, enabling precise, modular 'camera and lens combination' deployment.

Its tool design deeply aligns with film and advertising industry creative norms:

Professional Camera Movement Library: Offers over 50 preset camera motions, including realistic dolly pushes, orbital surrounds, FPV flight perspectives, and bullet time, all callable via description.

Cinematic Quality Control: The proprietary 'HCS Module' embeds color science from six top film cameras (ARRI, Red, etc.) and 11 classic lenses, enabling one-click cinematic color grading and dynamics.

Precision Editing Capabilities: 'SOUL Inpaint' provides pixel-level editing, supporting object replacement, background changes, and dynamic tracking optimization for advertising-grade control.

Value-Added Ecosystem and Multi-Device Coverage: The 'Nano Banana' ecosystem offers over 2,000 mini-tools for 4K image generation, video restoration, and style transfer. Beyond desktop, Higgsfield's mobile app Diffuse enables 'zero-threshold creation' (e.g., uploading selfies for dynamic videos), while AI virtual idol KION showcases its virtual human technology.

These features form a complete chain from concept to finished product, positioning Higgsfield as a 'virtual production team' for professionals, not a toy for amateurs.

Technologically, unlike foundational model developers, Higgsfield combines multiple off-the-shelf AI models into 'practical tools + workflows' to meet specific industry needs.

The platform utilizes its self-developed 'Soul' image model for visual consistency but excels at aggregating and refining external mainstream models like Sora 2, Kling, and Minimax. It can parallelize different models in a single generation, offering users multiple stylistic shot options.

This strategy cleverly avoids the AI video sector's fatal flaw of 'your model is good, but next month someone else's will be better.' Higgsfield encapsulates technical complexity in the backend, delivering stable, user-friendly, and scalable creative tools upfront.

Higgsfield's user base comprises advertisers, brand content teams, designers, editors, and social media creators. Their core demand is enhancing efficiency and producing directly commercializable, conversion-driven content.

The platform strategy revolves around this:

Higgsfield prioritizes showcasing works on platforms like X and builds its ecosystem through 'creator bonuses' and official reposts of high-quality cases. It directly incentivizes content creation with 'up to $100,000 weekly bonuses,' attracting users through large prize pools.

An e-commerce user noted, 'Used Higgsfield AI to produce 50 product videos in 2 hours without hiring traditional cameramen.' These videos directly enhanced product page appeal and conversion rates, precisely addressing commercial clients' pain points: cost reduction, efficiency gains, and conversion boosts.

Its commercial success validates this approach. Within five months of launch, Higgsfield's annualized revenue reached $50 million, surpassing $200 million ARR in nine months, doubling from $100 million in just two months.

/ 02 /

Valuations Approaching $10 Billion: 11 Companies Secure Major Funding in a Year

Looking back at late 2025, it was a year of technological breakthroughs for video generation models and intensified market competition. The AI video sector saw surging investment enthusiasm, with multiple core companies disclosing major funding rounds. Capital prioritized technical implementation and scenario adaptability over mere concepts.

Higgsfield AI broke through with creator incentives and social media strategies, while others leveraged technical strengths or commercialization capabilities to capture market share.

AISPEAK secured two funding rounds through global product expansion, raising $60 million in Series B led by Alibaba and RMB 100 million in Series B+, totaling over $70 million.

Its products focus on ease of use, covering short video creation and commercial ad generation. It boasts over 100 million global users, 16 million MAU, and $40 million ARR.

Shengshu Technology, a benchmark for Chinese video and 3D multimodal models, originated from Tsinghua University's Institute for Artificial Intelligence. In 2025, it completed a Series B round worth hundreds of millions of RMB.

Its core product, the Vidu video model, uses the original U-ViT architecture to support 16-second 1080P video generation. It topped both VBench-1.0 and VBench-2.0 rankings, surpassing Sora and Runway. With video generation costs at one-tenth of peers, it achieved $20 million ARR in eight months, generating over 400 million videos, with commercial content exceeding 50%.

Pollo AI, a lightweight tool, set a record for early-stage AI video tool funding in China with a $14 million seed round in December 2025.

Its product integrates scriptwriting, video production, and post-editing, emphasizing automation. With just seven months of launch, it reached 6 million MAU and achieved break-even.

SandAI, a pure video model company, secured over RMB 100 million in funding.

Focusing on general video model R&D, it excels in subject consistency across frames and complex scene generation, providing B2B technical services and customized model training for film studios and content platforms.

LiblibAI, a Chinese multimodal creation platform, completed a $130 million Series B round in October 2025, co-led by Sequoia China and CMC Capital. It integrates image, video, and 3D generation capabilities, fostering a 'tools + community' ecosystem with 20 million AI creators.

Video Rebirth, a professional AI video company, secured $50 million in funding in November 2025. Led by a former Tencent Distinguished Scientist, its team introduced a physical attention mechanism to address physical realism and controllability in video generation.

Globally, the overseas sector saw further concentration among top players.

Runway initiated a new funding round in late Q4 2025, targeting $500–$1,000 million at a valuation aiming for $10 billion, still ongoing as of January 2026.

Earlier, it closed a $308 million Series D round in April 2025, backed by NVIDIA and SoftBank, valuing it at $3.55 billion.

Its technology gained Hollywood recognition, with the Gen-3 Alpha model used by Netflix and Paramount for trailers. The December 2025 Gen-4.5 model further broke physical realism barriers, topping Video Arena's independent rankings.

Luma AI exceeded funding expectations, announcing a $900 million Series C round in November 2025, valuing it at $4 billion. Backed by AMD Ventures and a16z, its core Ray 2.0 model, with exceptional 3D content generation, doubled its user base in 30 days, becoming a Hollywood staple for replacing traditional storyboards and shortening pre-production cycles.

Pika Labs closed an $80 million Series A+ round in late 2025, with continued support from a16z. Its strength lies in precision editing, offering 'one-sentence scene revisions' for localized redrawing without disrupting overall style. This differentiation earned it a spot on Adobe's creative suite shortlist, potentially integrating into mainstream tools.

Stability AI secured its first $45 million round in October 2025, with Disney participating strategically and demanding theme park short film pipelines within a year, binding technology commercialization to cultural tourism scenarios.

Israel's Lightricks completed a $60 million strategic round in July 2025, funding AI film studio LTXV and generative video model R&D. Its breakthrough lies in redirecting TikTok advertiser budgets toward AI-generated videos over traditional shoots.

Through Higgsfield's case, we can foresee that standalone text-to-video models will gradually merge into unified multimodal 'grand unified' models. AI video competition will shift from single-point tools to platform ecosystems integrating generation, distribution, and monetization. Future competitiveness will lie in specialized fine-tuned models for e-commerce, education, healthcare, and other verticals.

By Lang Lang