SialFlow
Transform images at request time, queue heavier image, video and document work as durable jobs, and keep the status, retries and output of each run.
Platform
Image manipulation, video processing and document extraction run against a shared API and job model, so a team does not have to operate three separate toolchains.
The platform is shaped by media and delivery systems we have had to run ourselves.
POST /v1/images/{id}/bg-remove { "format": "png" } → 202 job kind=bg_remove poll GET /v1/images/jobs/{id}
Capabilities
One job model and one set of events. Different workers underneath.
Smart crop, background remove, responsive variants, bulk jobs and versioned transform profiles at catalogue scale.
VOD encoding, batch processing, thumbnails and clip workflows on the same job and telemetry model.
OCR, conversion, structured extraction and review states, run as workflows rather than one-off scripts.
Background remove is live. Broader enrichment and moderation queues are in development on the same job model.
Execution
From ingest to delivered variants, each step recorded as a durable job.
Object upload completes; the asset enters the workflow.
MIME inspect and policy checks before expensive work runs.
Background remove, smart crop and named transform profiles run as durable jobs.
Variants and signed delivery URLs, with the job history attached to the run.
On-demand manipulation
Fast image operations can run through the delivery URL. Compute-heavy work becomes a durable asynchronous job with status, retries and history.
Resize, crop, change format, apply overlays and generate responsive variants through a compact URL grammar, without preparing every derivative in advance.
Entropy-aware smart crop and background removal help turn one master image into channel-ready outputs. Alpha PNG and WebP delivery can be requested synchronously where supported.
Bulk image processing, VOD encoding, thumbnails, OCR and document extraction run asynchronously so clients do not have to hold open fragile requests.
Input, transform profile, workflow version, retries and output stay attached to the run, so a team can see what produced a given asset.
GET /i/{org}/{image}/w800-h600-fit-cover-g-auto-f-webp/product.webp → resize · smart crop · WebP · delivered at the edge
Architecture
Requests are checked against policy and tenant quota, queued, processed by the relevant worker pool, then closed out with an event.
Client ↓ SialFlow API (bearer auth, project scope) ↓ Policy & quota (tenant concurrency limits) ↓ Job queue ├─ image workers ├─ video workers └─ document workers ↓ Object storage (yours or ours) ↓ Events / webhooks
Solutions
Three common shapes: e-commerce, agencies and publishers.
E-commerce
Ingest → moderate → transform → variants → deliver. PDP-ready outputs with auditable preparation jobs.
Agencies
Templates, brand policy gates and bulk runs in client-scoped workspaces.
Publishers
Image, video and document work in one place, from ingest to publish-ready.
Status
FAQ
Image pipelines, background remove, video encoding and document extraction are live. Broader AI enrichment and moderation queues are in development.
Image delivery supports request-time transforms such as resize, crop, format conversion, overlays and responsive variants. Heavy work such as bulk processing, encoding or extraction runs as asynchronous jobs with status and retries.
Yes. Bearer auth, project scope, async jobs and outbox events follow the same model across capabilities.
Tell us the first workflow: an image pipeline, a video job or a document flow. We can point you at portal access and documentation.
Next step
An image pipeline, a video job or a document flow is enough. Full API detail is on sialflow.com.
