Sieve: Process. Understand. Search. Video.

Process. Understand. Search.Video.

Watch Demo(2 min)

Stop wrestling with ML pipelines, video infrastructure, and large datasets. Build magical video AI functionality into your apps with just a few API calls.

/query/intervals

# Get all moments when a dog is running

curl 'https://api.sievedata.com/v1/intervals' \
  -X POST \
  -H "X-API-Key: YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "class": {"$exists": "dog"},
    "temporal.action": {"$in": ["run"]},
    "project_name": "pet_care",
    "source_name": "cam_06-09-1969"
  }'

# Get all moments when a dog is running

curl 'https://api.sievedata.com/v1/intervals' \
  -X POST \
  -H "X-API-Key: YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "class": {"$exists": "dog"},
    "temporal.action": {"$in": ["run"]},
    "project_name": "pet_care",
    "source_name": "cam_06-09-1969"
  }'

Create a workflow

Use a starter workflow with zero configuration: Models & workflows for media, industrial, real estate, news, social, and other use cases available out of the box. Object detection, classification, segmentation, creative video AI generation, etc.
Define a workflow in Python code: Want to run custom image or video AI models on Sieve? Just define your logic & dependencies in a simple interface. We’ll build & deploy it for you.

POST /v1/push

{
	"project_name": "video_art_generation",
	"source_name": "something.mp4",
	"source_url": "some_signed_url",
	"user_metadata": {
		"prompt": "an astronaut riding a horse"
	}
}

{
	"project_name": "video_art_generation",
	"source_name": "something.mp4",
	"source_url": "some_signed_url",
	"user_metadata": {
		"prompt": "an astronaut riding a horse"
	}
}

Response <200>

{
	"description": "published video something.mp4 to queuing",
	"job_id": "4206-9f5f-43df-afqd"
}

{
	"description": "published video something.mp4 to queuing",
	"job_id": "4206-9f5f-43df-afqd"
}

Push data

Submit data already hosted in the cloud to Sieve using signed URLs or upload data directly. Simple HTTP API with available Python wrappers.

Hyperfast processing

Lightning-fast results: Define workflows that process video in parallel by scaling out to thousands of machines. Hours of video processed in minutes, just one API call away.
Stream and batch support: Sieve’s hybrid model works for submitting thousands of hours of video at once and making queries after-the-fact or for building interactive user experiences that require you to poll Sieve in near real-time.

hello.py

import sieve
cli = sieve.Client()
# Get people who were jumping
objs = cli.query(
  project="wizard_security",
  q={
    "class": {"$exists": "person"},
    "temporal.velocity.y": {"$gte": 10}
  }
)
print(objs)

import sieve
cli = sieve.Client()
# Get people who were jumping
objs = cli.query(
  project="wizard_security",
  q={
    "class": {"$exists": "person"},
    "temporal.velocity.y": {"$gte": 10}
  }
)
print(objs)

run hello.py

> python hello.py
[
  {
    "class": "person",
    "object_id": "f3d3-4206-9b8b-4b9f",
    "temporal": {
      "bounding_box": {
        ...
]

> python hello.py
[
  {
    "class": "person",
    "object_id": "f3d3-4206-9b8b-4b9f",
    "temporal": {
      "bounding_box": {
        ...
]

Make structured queries

Everything is an object: A person, a car, the entire frame, the audio — they’re all treated as objects within Sieve that each have their own temporal properties.
You already know how to query Sieve: It’s the same as writing a MongoDB query. You can query for every object and its properties with the same flexibility MongoDB offers.

Continual feedback loop

Improve your workflows and models continually by submitting corrections, additions, or deletions to the structure generated by Sieve. Define custom routines, or use out-of-the-box routines that run periodically to improve models based on this feedback.

Predictable Pricing

Credit-based pay-as-you go and volume plans available. Credits allow users to process video at given FPS, retraining models, and more.

No Vendor Lock-in

Run models and workflows built with Sieve without Sieve’s infrastructure. Take full ownership over your own model weights.

Intuitive API + CLI

Use Sieve’s API, CLI, or dashboard to interact with product. Get support from Sieve’s community.