Lumen Labs
Generative image + video, one platform

Lumen Labs turns a sentence into a 4K image in under three seconds and a 10-second 1080p clip in under a minute. One API, two media types, frame-perfect control — built for teams that ship visuals every day.

  • 4K images in ~3s
  • Commercial license included
  • No GPUs to manage

Powering creative pipelines at studios, agencies, and product teams

NorthlightCadencePixelaLumen FilmsStudio VoltBrightframe
The model suite

Every model you need,behind one endpoint.

Stop juggling five vendors and six checkpoints. Lumen unifies text-to-image, text-to-video, and editing under a single typed API and one billing line.

Aurora — text to image

Our flagship diffusion model renders photoreal and stylized 4K images in about three seconds, with accurate hands, legible text, and faces that hold up at full crop.

Motion — text to video

Generate 10-second 1080p clips at 24fps with stable subjects, real camera moves, and temporal coherence that survives the cut.

Reframe — inpaint & edit

Mask, extend, and relight any image or frame. Swap backgrounds, remove objects, and outpaint to any aspect ratio without re-rolling the whole scene.

Style Lock

Pin a look with three to ten reference images and reproduce it across thousands of generations — brand-consistent every time.

Director controls

Steer lighting, lens, camera angle, and motion with structured parameters instead of prompt-engineering roulette.

Built for production scale

2.8s
Median 4K image
120M+
Frames rendered
99.95%
API uptime
190+
Countries served
Made on Lumen

A single prompt. A directed result.

Every piece below was generated and graded entirely on Lumen Labs — no external retouching. Hover-worthy at full resolution.

Aurora · 4K image

Neon Monsoon

Cinematic street scene, anamorphic flare, rain-soaked reflections.

Motion · 1080p video

Cold Open

10s dolly push through a fogged forest at first light.

Style Lock

Studio Portraits

200 on-brand headshots from a single reference set.

Reframe

Product Hero

One pack shot relit and outpainted for six ad formats.

Aurora · macro

Liquid Chrome

Abstract metallic textures for a fashion title sequence.

Motion · loop

City in Motion

Seamless 8s aerial loop over a generative skyline.

Built for developers

An image API thatfeels like infrastructure.

Predictable latency, streaming results, and webhooks that actually fire. Integrate generation in an afternoon, not a sprint.

One REST endpoint

Image, video, and edits all flow through a single consistent, idempotent API surface.

Typed SDKs

First-class SDKs for TypeScript, Python, and Go with streaming, retries, and full type safety.

Async webhooks

Kick off long video renders and get a guaranteed-delivery callback the moment the frame is ready.

Usage you can predict

Per-second video and per-image pricing with hard spend caps, so a viral launch never surprises finance.

Pricing

Pay per render, not per seat.

Start free, scale linearly, and keep full commercial rights on every output. No platform fees, no annual lock-in.

Free

For exploring and prototyping.

$0/mo
  • 200 image credits/mo
  • 5 video credits/mo
  • Up to 1080p output
  • Personal-use license
  • Community support
Most popular

Studio

For creators and small teams shipping daily.

$49/mo
  • 5,000 image credits/mo
  • 300 video credits/mo
  • Up to 4K image · 1080p video
  • Style Lock + Reframe
  • Full commercial license
  • API access + webhooks

Scale

For platforms and high-volume pipelines.

Custom
  • Volume credit pricing
  • Dedicated GPU capacity
  • Private model fine-tunes
  • 99.95% uptime SLA
  • SSO + audit logs
  • Named solutions engineer

Questions, answered.

Who owns the images and video I generate?

You do. Every output on a paid plan ships with a full commercial license — use it in ads, products, and films with no per-use royalties or attribution required.

How fast is generation, really?

Median is about 2.8 seconds for a 4K image and under a minute for a 10-second 1080p clip. Latency is published live on our status page and backed by the Scale SLA.

Can I keep a consistent style or character?

Yes. Style Lock pins a look from three to ten reference images, and private fine-tunes on Scale let you train a dedicated model on your own brand or characters.

What about safety and provenance?

Every output carries invisible C2PA content credentials, and prompt and output moderation is on by default with configurable thresholds for enterprise accounts.

Do you train on my prompts or outputs?

No. On Studio and Scale your data is never used to train models, and you can enable zero-retention so generations are deleted from our systems immediately after delivery.