Pika Labs

Pika Labs is a United States-based artificial intelligence startup founded in April 2023 by Stanford University alumni Demi Guo and Chenlin Meng. The company develops generative video models that produce short animated clips from text descriptions, still images, or short video prompts. Pika has emerged as one of the most prominent independent companies in the generative video sector alongside Runway ML and OpenAI's Sora, and its consumer-oriented mobile experience launched in 2025 has positioned the firm at the intersection of creative tooling and short-form social media.

Origins and team

Demi Guo and Chenlin Meng founded Pika after meeting at Stanford's AI Lab, where Meng had previously contributed to denoising diffusion implicit models and other foundational diffusion research. The early team drew from researchers and engineers with backgrounds at OpenAI, Stability AI and Meta AI, and the company began with a focus on accessible creative tools rather than enterprise post-production.

Products

Pika's main product, accessible at pika.art and through native applications, generates short videos from prompts in 720p or 1080p resolution with typical clip lengths between 5 and 10 seconds. Successive model releases — Pika 1.0 in late 2023, Pika 1.5 in 2024 and Pika 2.0 in late 2024 — improved temporal coherence, motion fidelity and the ability to follow complex prompts. Pika 2.2, released in February 2025, expanded resolution support, increased maximum clip length and introduced more controllable scene editing.

Pika has developed creative tools including Scene Ingredients, which allows users to combine specific objects, characters and locations into a single shot, and Lip Sync, which animates a generated character speaking arbitrary audio. A consumer mobile application launched in July 2025 reoriented part of the product toward short-form social video creation, with templates and remix workflows aimed at Gen Z users on iOS and Android.

Funding and partnerships

Pika has raised approximately USD 135 million across multiple rounds at a reported valuation of USD 470 million as of 2025, with investors including Lightspeed Venture Partners, Spark Capital, Adam D'Angelo, Elad Gil and Andrej Karpathy. Press reports in mid-2025 indicated that Meta Platforms had held acquisition discussions with the company. The firm has reported a user base of approximately sixteen million across its products by mid-2025.

Technical context

Pika's models are based on diffusion architectures conditioned on text, image and structural inputs. The company has not published detailed technical reports, but the product roadmap reflects the broader 2024–2026 evolution of video diffusion from short, low-resolution clips toward longer, higher-fidelity outputs with finer prompt control. Competitors in this category include Runway ML's Gen-3 and Gen-4 families, OpenAI's Sora, Google DeepMind's Veo, Kling AI from Kuaishou, Hailuo AI from MiniMax, and Luma Labs' Dream Machine.

Industry significance

Pika is regularly cited as a case study for how a small, research-led startup can win a niche in generative media against well-capitalised incumbents by combining accessible UX with iterative model improvements. The company has also drawn attention to the policy and labour implications of high-quality AI video for advertising, film and short-form social platforms, with debates over copyright, training data provenance, deepfakes and creator economics intensifying through 2025.

References

1. Fortune. (2025). This 26-year-old's TikTok-like AI app makes playful short videos from just a few words.
2. Sacra. (2025). Pika valuation, funding and news.
3. Pika Labs. (2025). Pika 2.2 Release Notes.
4. Malaysian Communications and Multimedia Commission. (2024). Industry Guidelines on Synthetic Media.