What Linum does

Linum is a two-person AI research lab founded in 2022 by brothers Manu Chopra and Sahil Chopra, headquartered in San Francisco and incorporated in Delaware as Linum, Inc. The company is building open-weight text-to-video generative AI models aimed at reducing the cost and complexity of animation and filmmaking — explicitly targeting the $3-4M indie animation production cost benchmark with generative primitives such as consistent characters, intentional camera movement, and controllable motion.

The company's technical core is Linum v2, a 2B-parameter text-to-video diffusion transformer trained from scratch (not extended from an image model like its predecessor v1, which inflated Stable Diffusion XL into video). V2 generates 2-5 second clips at 360p or 720p resolution and is supported by a custom Image-Video VAE that Linum built over four months in 2024, featuring Self-Modulating Convolution operations adapted for Conv3Ds with 8x spatial and 4x temporal downsampling. Both the v2 model weights and the standalone VAE are released under Apache 2.0, permitting commercial use. The team also maintains ancillary engineering assets such as a GPU-cluster monitoring service built for ~$8/month on Ubicloud.

Commercially, Linum is pre-revenue with no disclosed pricing model and no paying customers. Models are distributed through Hugging Face (model weights), GitHub (source code), and third-party inference providers; Linum itself does not monetize direct access. Capital comes from the Y Combinator W23 batch and venture partners (Adverb Ventures; Eight Capital and Unpopular Ventures also appear in the funding record), alongside compute partnerships with Crusoe (H200 training cluster) and Together AI, plus infrastructure support from Cloudflare and Ubicloud. Go-to-market is product-led and community-led: technical content marketing through the Field Notes blog, an active newsletter, and direct outreach to filmmakers, animators, and secondary users such as musicians and set-visual artists.

Name

Linum

Legal name

Linum, Inc.

Website

https://linum.ai

Company type

Private

Founded year

2022

Operating status

Operating

Headcount range

1–10 employees

Short description

Linum is a two-person AI research lab building open-weight text-to-video diffusion models from scratch to make animation and filmmaking dramatically cheaper and more accessible for filmmakers, animators, musicians, and set-visual artists.

Ownership category

akta.pro rank

Product category

Generative AI Models

NAICS

Software Publishers (5132), Software Publishers (513210)

SIC

Services-Prepackaged Software (7372)

akta.pro primary industry

Open-Source Model Ecosystems & Model Marketplaces (HDAAACAM)

Keywords

• Text-to-video generation
• Generative video models
• Open-source AI models
• Video diffusion models
• Video synthesis AI

Headquarters

HQ city

San Francisco

HQ country

United States

HQ region

North America

Markets served

GTM type

B2B and B2C

Offering type

Software

Cost components

Technology or R&D, Infrastructure, Personnel, Marketing or Sales

Revenue model

1. Pre-revenue / Research Stage: Linum is a research lab in pre-revenue stage. Revenue model has not been established. Models are released as open-source (Apache 2.0) with no direct monetization. Company is backed by Y Combinator and Adverb Ventures.

Go-to-market motion2 records

Distribution channels3 records

Marketing channels4 records

Core offering

Linum is a research lab that trains and open-sources text-to-video generative AI models from scratch. Its core product, Linum v2, is a 2B-parameter Apache 2.0 licensed model that generates 2–5 second video clips at up to 720p resolution from text prompts. The company also open-sourced a proprietary Image-Video VAE used during training, and a low-cost GPU cluster monitoring tool. Models are distributed free via Hugging Face and GitHub for use by filmmakers, animators, musicians, and developers.

Product overview

Linum is a two-person AI research lab offering open-weight text-to-video models. The core product is Linum v2, a 2B parameter text-to-video model available in 360p and 720p variants, licensed under Apache 2.0. The company also open-sourced their custom Image-Video VAE developed during the v2 training process. Supporting the product ecosystem is Field Notes, a technical blog covering research and engineering insights from building generative video models from scratch.

Differentiator

Problem solved

Functional benefit

Brands

• Linum v2: Open-weight 2B parameter text-to-video models capable of generating 2-5 second clips at up to 720p resolution, Apache 2.0 licensed.
• Linum Image-Video VAE
• Field Notes

Products and services

• Linum v2 Text-to-Video Model Open-weight 2B-parameter text-to-video model that generates 2–5 second clips at up to 720p resolution from text prompts. Licensed Apache 2.0 for commercial use, with weights distributed via Hugging Face and source code on GitHub for animators, filmmakers, musicians, and AI developers.
• Linum Image-Video VAE Open-source Variational Autoencoder for compressing both images and videos into a latent space, built in-house to support Linum v2 training. Includes SMC (Self-Modulating Convolution) adapted for Conv3D to avoid artifact problems and supports 8x spatial and 4x temporal downsampling. Released for researchers and developers building video diffusion pipelines.

Quantifiable outcome

• Animation production cost reduction target: order of magnitude decrease from $3-4M

Segments2 records

Ideal customer profiles2 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

AI capability6 records

Feature3 records

Partnerships

One partnership is on record.

• CrusoecoreTechnology or IntegrationInfrastructure partner providing GPU compute (H200 cluster) for training Linum v2. Co-marketed blog post on How Linum built a frontier video model on Crusoe.

Scale indicators5 records

Recent moves6 records

Expansion highlights5 records

Emerging players

• Black Forest Labs: Black Forest Labs builds open-weight generative models (FLUX) for images and is widely viewed as a leading candidate to extend into open-weight video. As a peer in the open-source foundation model category, it represents the most direct competitive model both in philosophy and in Hugging Face-distributed weight strategy that Linum would benchmark against.
• Wonder Dynamics: Wonder Dynamics (Autodesk) builds AI tools that automate character animation and VFX for filmmakers, targeting exactly the $3-4M indie animation budget pain point Linum references. Its focus on animating existing footage rather than generating video from scratch is complementary but competes for the same indie filmmaker wallet and workflow.

Broad incumbents

• Midjourney: Midjourney dominates AI image generation for creative professionals and animators — many of the same users Linum targets with video. Midjourney's moves into video and its established creator community represent both a competitive threat and a distribution reference point for Linum's filmmaker-focused GTM.
• OpenAI (Sora): OpenAI's Sora is the highest-profile closed-source text-to-video model and represents the frontier the entire category is measured against. As a broad incumbent with vastly more compute and capital, OpenAI shapes user expectations around clip length, fidelity, and controllability that Linum's open-weight competitor must match or differentiate against.
• Google DeepMind (Veo): Google's Veo text-to-video model is a frontier competitor backed by Google's research depth and infrastructure, targeting the same professional content creation workflows. Veo sets the bar for cinematic quality, prompt adherence, and duration that small open-weight models like Linum v2 must close the gap to over multiple generations.

Direct peers

• Pika Labs: Pika Labs builds consumer-facing text-to-video and video editing models marketed to creators and marketers, with similar 4-5 second clip outputs and pricing geared toward indie filmmakers. Linum's open-weight 2B model targets functionally equivalent use cases and creator segments with a fundamentally different (open-source vs. hosted) go-to-market.
• Runway: Runway is the commercial category leader in generative video for filmmakers and creative professionals, offering text-to-video, image-to-video, and editing tools. Linum competes directly for the same filmmaker/animator persona and targets the same workflow gap of accessible creative video generation, but takes an open-source rather than hosted-SaaS approach.
• Stability AI: Stability AI pioneered the open-weight foundation model distribution model (Stable Diffusion) that Linum explicitly follows for video, including Hugging Face distribution and community-led commercialization. Both organizations focus on open-source generative media models and downstream developer/maker ecosystems, making them the closest strategic analog despite different modalities.
• Luma AI: Luma AI develops text-to-video and 3D generation models (Dream Machine) aimed at filmmakers and creative professionals, generating short high-fidelity clips from prompts. Linum's targeting of animators and filmmakers, plus its emphasis on creative controls in model training, places it in direct competition for the same creative-producer segment.

Others

• Hugging Face: Hugging Face is Linum's primary distribution channel for open-weight models and the de facto hub for the open-source generative AI ecosystem. While not a competitor, Hugging Face's hosted inference, evaluation, and fine-tuning offerings could either amplify or absorb Linum's reach depending on ecosystem positioning.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat3 records

Key risks7 records

Key highlights7 records

Customer concentration

• Xx.com/LinumInc
• GitHubgithub.com/Linum-AI/linum-v2

Revenue estimate

Valuation estimate

Number of profiles

Profiles2 records

Funding overview

Funding rounds1 record

Investors3 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Linum