Palette

Palette (also presented as Palette Labs, product at studio.palettelabs.com) is a San Francisco-based AI-native media platform that consolidates multimodal content production into a single generation engine. Users start from a prompt, image, document or reference, and the platform produces video, visuals and structured content from the same source without rebuilding the idea for each output format. Editing is done in natural language — restyling a scene, replacing an object, revising a sequence — while the platform maintains the connection between source and context as formats change.

The product includes storyboarding controls (audience, format, pacing, visual intent), character definition for consistent identity across shots and formats, a video editor, motion transfer, prompt templates and editing, shot continuation, 3D-guided shots and training-course generation. A model-routing layer lets users select a specific third-party generation model or delegate selection to Palette; models referenced include PixVerse V5.5, ByteDance Seedance 2.0, Google Veo 3.1, Kling 3.0 Pro, MiniMax Hailuo 02 and LTX Video. Routing applies specification and capability gates and then cost gating to pick the cheapest trusted option. A REST API (base URL https://studio.palettelabs.com/api/v1) exposes image, video, music and image-tool endpoints plus job polling, model listings, asset management and credit balance, authenticated with bearer API keys and drawing on the same credit balance as the web Studio. Roadmap items listed on the site include brand kits, storyboard-to-sequence, real-time collaboration, performance direction, auto-repurposing and a developer platform.

Palette positions itself for both individual creators and enterprise creative, marketing, learning-and-development, customer support, product, manufacturing, healthcare, retail, field service and safety-training teams, framing its value as reducing tool and context fragmentation and increasing the share of generations that reach a usable result.

Founding story

Founders Josephine Lee (CEO) and Serena Pei (CTO) met at an art program and studied computer science at MIT. They cite prior experience in multimodal research and computer vision as well as industry roles at Scale AI, Datadog and Amazon. The company site describes the team as MIT computer science graduates and multimodal AI researchers.

Business model

Two-track model: a self-serve individual tier priced at $0.01 per credit covering native image and video generation, preset workflows and other generation tools, and an enterprise tier sold via custom quote offering end-to-end services and custom self-improving workflows. A developer API bills against the same Studio credit balance.

Consumption-based credits ($0.01/credit) for self-serve and API usage, with credits consumed only on creation, automatic refunds for failed jobs and partial refunds when a cheaper fallback model is used; custom-quoted enterprise engagements for services such as ad generation and runs, launch programs, influencer UGC and L&D platforms.

Traction

Public evidence of traction is limited. Y Combinator lists the company as Active in the Summer 2026 batch with a team size of 2. Self-reported platform metrics on the company site include 35% faster generation, 50% lower generation cost, 100% task-capable routing on an internal benchmark and 130+ automated regression tests. No customer counts, revenue or usage figures are disclosed in the sources.

Latest developments

Y Combinator Summer 2026 batch participation and the accompanying launch post describing the platform, its enterprise service lines and the public Studio product; the site lists brand kits, storyboard-to-sequence, real-time collaboration, performance direction, auto-repurpose and a developer platform as forthcoming capabilities.