Vision AI.
Detect, classify and read images and video through one API — grounded in the multimodal model, not a fixed label set.
Objects, scenes, faces, text and defects in images and video, through a single API. Because it's grounded in the multimodal Aether model rather than a frozen label set, it recognizes what you describe — and you can specialize it from a handful of examples.
- Category
- Applied AI
- Deployment
- Managed → air-gapped
- Governance
- IAM · encryption · audit
Vision APIs ship a fixed label set. Your world usually isn't in it.
Classic vision services classify against a catalogue someone else decided on — useful for generic objects, useless the moment you need 'this specific defect on our part' or 'this document stamped paid'. You end up training a bespoke model, which is its own project. Aether's vision is grounded in the multimodal model, so it does open-vocabulary recognition — you describe what to look for in words — and specialises to your categories from a handful of examples, not a labelling campaign.
OCR, detection and video, through one API grounded in the model.
Objects, scenes, attributes, text in photos and video frames, including handwriting, and tracking across frames — one interface, not a different product per task. Because it's the same model that can also reason over what it sees, the output isn't just boxes and labels; it can describe, compare and decide, so a defect isn't only detected but characterised and routed.
Your images, your boundary.
Inspection footage, medical images and document scans are exactly the data that shouldn't leave the building. Aether's vision runs where your data lives — managed, in your VPC, on-prem or fully air-gapped — and on the same platform as your storage and the rest of the model, so frames don't cross a vendor boundary to be analysed.
No training job: describe the defect in words and detect it, grounded in the model.
import { aether } from "@aether/sdk";
const r = await aether.vision.detect({
image: frame,
find: "hairline crack near a weld seam", // open-vocabulary
fewShot: "lake.defect_examples", // 6 labelled images
});
if (r.found) routeToQA(r.box, r.severity);A fixed-label API can't find 'hairline crack near a weld seam' unless that label exists. Here you describe it, optionally anchor it with a few examples, and the multimodal model finds and characterises it — no bespoke model to train first.
Three steps to running.
An image or a video frame via API or SDK.
Open-vocabulary recognition, OCR or your own few-shot labels.
Structured output flows straight into your workflow.
Vision AI, in full.
Objects, scenes and attributes, plus open-vocabulary recognition from a prompt.
Read text in photos, scans and video frames, including handwriting.
Specialize to your parts or defects with a handful of labelled images.
Track objects and events across frames, with timestamps.
Provision it in a few lines.
Every service is reachable from the same SDK, CLI and infrastructure-as-code — one identity, one bill, one audit trail across the whole catalog.
# Provision vision ai on Aether Cloud
aether applied-ai create \
--service vision-ai \
--name app \
--region us-1 \
--deploy managed # or vpc | on-prem | air-gappedAt a glance.
- Inputs
- Image · video
- Recognition
- Open-vocabulary + custom
- OCR
- Print + handwriting
- Custom
- Few-shot
- Latency
- Real-time + batch
Built for real work.
Quality inspection
Content moderation
Document & receipt capture
On one model, not stitched together.
The usual stack runs vision ai in one product, the model in another and the data in a third — and the seams between them are the cost. Aether Cloud runs it on the same platform that serves the model, governs your identity and deploys into your boundary, with the rest of the catalog one hop away.
No stitching a vector DB to one place, a warehouse to another and a model to a third — vision ai sits next to the rest of the catalog, one identity, one bill.
The provider that runs Aether runs your vision ai — so the data and the model never leave the same governed boundary to talk to each other.
Need a capability that isn’t here yet? The model writes and deploys it into the same boundary — the catalog is a starting point, not a ceiling.
Good to know.
Detection, OCR and classification like them — but grounded in the multimodal model, so it recognizes what you describe and specializes from a few examples, not a frozen label set.
Yes — specialize it from a handful of labelled images instead of training a model from scratch.
Yes — object and event tracking across frames with timestamps.
Your boundary, your choice.
Pairs well with.
Speech-to-text and text-to-speech in one service — transcription, diarization and natural voices, on the model that reasons over what's said.
Real-time and batch translation across 100+ languages, with glossaries and the context the model carries — not phrase-book substitution.
Turn documents into structured data — forms, tables, invoices, contracts — grounded in the model, not brittle templates.
Time-series forecasting as a service — demand, capacity, load — with uncertainty intervals and what-if scenarios.
Real-time recommendations and next-best-action from your interaction data, served at low latency.
Run Vision AI on Aether Cloud.
Detect, classify and read images and video through one API — grounded in the multimodal model, not a fixed label set. Deployable managed, in your VPC, on-prem or fully air-gapped — talk to us about the configuration your workloads and your boundary require.