Build and scale creative products with the most direct and expressive video and image generation models available — through the Orrery API.
Start Building
Access frontier image and video generative features through one straightforward endpoint. Build and iterate generation for your business. Start plans include:
Scale your creative AI products with higher rate limits and hands-on support from the Orrery research team.
gen = create({ prompt: 'impossible camera flight through a slot canyon', model: 'vela-2', aspect_ratio: '16:9', });
Vela 2 opens a new generation of video models: fast coherent motion, dense material detail, and event sequences that follow cause and effect. The result is a much higher share of usable generations and footage that survives an actual edit.
Generate action-heavy shots from a plain sentence. From a high-speed chase to close human movement, Vela 2 raises the bar on motion fidelity — smooth, cinematic, and unnervingly physical.
Tell a story with real cinematic grammar. Vela 2 handles precise camera movement — sweeping panoramas, intimate close-ups, and continuous tracking shots — all specified in text.
Extend your content toolkit. Whether it is a product promo, a service showcase, or serial storytelling, generate broadcast-usable clips in seconds and keep the pipeline moving.
Fold generative work into an existing VFX pipeline. Vela 2 produces previsualization, plate replacement, and first-pass sequences without booking a location.
Prism delivers category-specific visual quality, crafting images that hold up against professional standards — not generic model output. From campaign design to editorial, every generation is tuned to the requirements you actually ship against. Learn more
Prism sets a new bar for visual precision. Its handling of your prompt holds every element in place — color, composition, and rendered text — with accuracy that matches what you described.
| Resolution / Model | Driftwave 3.5 | Inkwell | Muraldream | Halcyon 1.1 | Orrery Prism |
|---|---|---|---|---|---|
| 1080p | N/A | ¢ 6.2 | ¢ 5.4 | ¢ 6.1 | ¢ 1.7 |
| 1080p fast | N/A | N/A | N/A | N/A | ¢ 0.4 |
| 720p (coming soon) | ¢ 6.3 | N/A | N/A | ¢ 5.2 | ¢ 0.9 |
| 720p fast (coming soon) | ¢ 3.8 | N/A | N/A | ¢ 2.4 | ¢ 0.2 |
The Orrery image model runs on the same universal fusion architecture as our video model, which is where the efficiency comes from — a research result, not a quality trade you pay for later.
Generate consistent character variations from a single reference photo. Prism holds facial structure and identity steady across poses, expressions, and scenes.
Apply and blend style references with exact control. Prism reads the essence of each reference image and lets you combine distinct visual elements without losing finish quality.
Vela and Prism share an instruction-following system, so users do not need to learn prompt engineering. That means teams can build generative products that reach new markets.
Build workflows that take a static image and return motion. Instruct Vela in ordinary language and get a narrative back, not a slideshow.
Control the narrative Vela generates by pinning a start and end image keyframe.
…and extend those narratives into stories, all without any pixel-editing complexity inside your app.
Create continuous loops for engaging interfaces, product marketing, and backgrounds.
A generative camera that lets even a first-time user get a shot looking right with one plain sentence.
Your app can produce content sized for every platform without any video or image editing UI.
Video carries culture, ideas, and connection around the world. Video is a universal language. Unlike text, making video is a physical act, and editing high-dimensional video data is genuinely hard. So the few who hold the means of production create, and the rest of us consume. They tell the stories and we all listen.
Video can be the most effective medium of visual storytelling. Video — if it were as universally manipulable as text — could be the most effective medium of thought.
To get there we have been training Vela, a family of generative models that produce and manipulate video. Adoption of Vela 1 outran everything we projected, and in directions we did not predict. Which tells us we should go faster.
We are launching the Orrery API in beta today with the newest family of Vela 1.6 models. It covers high-quality text-to-video, image-to-video, video extension, loop creation, and our camera-control capabilities.
Through our research and engineering we aim to bring about an age of visual literacy in which anyone can make the thing they can picture, and the distance between an idea and an artifact keeps shrinking.
Video is a powerful and persuasive medium. To prevent misuse we built a layered moderation system that pairs AI filters with human oversight. The API lets you tune that system to the norms of your market and your users. With ongoing feedback and learning we keep refining the approach so our models and products stay inside legal standards and get used constructively.
Enterprise Terms of Service
