Automate the Boring Parts of Your Creative Workflow

AI tools made generation fast. A ComfyUI workflow produces an image in seconds. But everything that happens after generation, the organising, tagging, exporting, and cleaning, still happens one asset at a time. The Pipeline Builder turns that repetitive work into something you define once and run forever.

The Tax on Creative Output

For prolific creators, the after-generation overhead is significant. At a thousand images per week, even thirty seconds of manual processing per image adds up to more than eight hours. That is an entire working day spent on tasks that follow the same pattern every time. Tag the new uploads. Strip the metadata. Move the approved ones. Export the batch. Tag the new uploads. Strip the metadata. Move the approved ones. Export the batch.

The problem is not laziness. It is that these tasks are genuinely repetitive: the same selection criteria, the same transformations, the same output destinations. They are patterns. And patterns should be automated. Creative time is better spent generating the next hundred images than post-processing the last hundred.

The batch processing patterns we have been writing about are now a first-class feature inside Numonic. You do not have to script them. You do not have to remember them. You compose them, save them, and let them run.

The Pipeline Builder

Numonic's Pipeline Builder lets you define asset processing workflows as a sequence of stages. Each stage has a clear purpose and a clear input-output contract, which means you can read a pipeline top to bottom and know exactly what it does.

Select, Filter, Transform, Output

Four stage categories, each with a specific role:

• Select stages choose which assets enter the pipeline. Select by collection, tag, date range, search query, or any combination. “All ComfyUI images from the last twenty-four hours with no tags” is a valid selector.
• Filter stages narrow down the selection. Filter by file type, dimensions, metadata presence, tag match, or custom criteria. “Only images wider than 1024px that have a prompt but no title” refines the set.
• Transform stages modify assets or metadata. Add tags, strip metadata, generate thumbnails, update fields, apply AI tagging, or run custom transformations. This is where the actual work happens.
• Output stages deliver results. Export to a folder, create a collection, send a webhook notification, generate a report. The pipeline produces something tangible.

Twenty-nine stage types cover every common operation in the AI-creative workflow. You build pipelines by dragging stages from a picker in the visual builder, or by writing them as text using the DSL shorthand:

select collection:"Client Work" | filter type:image | transform strip-metadata preset:share | output export

Pipe characters separate stages. Each stage has a type and parameters. Power users tend to find the text DSL faster for complex pipelines, because the same grammar that drives the search bar drives the selector. The visual builder is there when you want to see the shape of the flow.

Templates for Common Workflows

Not everyone wants to build from scratch. Eight built-in templates cover the most common automation needs, each one composed of real stages you can inspect, edit, and extend:

• Auto-Tag New Uploads — select untagged assets, apply AI-generated tags
• Weekly Client Deck — select this week's approved assets, strip metadata, export as a batch
• Metadata Cleanup — find assets missing titles or descriptions and generate them with AI
• Privacy Strip for Sharing — select a collection, strip sensitive metadata, export with IPTC disclosure
• Archive Old Assets — move assets older than ninety days to an archive collection
• Tag Standardisation — find tag variations (“landscape”, “Landscape”, “landscapes”) and normalise them
• Collection Summary Report — generate a report of collection contents and metadata completeness
• Export Batch — select, transform, and export in one step

Clone any template and customise it. Save your customised version as a new team template. The templates are not black boxes. They are starting points, and every stage inside them is the same stage you would drag into a pipeline from scratch.

Scheduling and Event Triggers

Pipelines are useful when you run them manually. They become powerful when they run themselves.

Cron scheduling lets you set a recurring schedule. Run the Weekly Client Deck pipeline every Friday at five o'clock. Run Metadata Cleanup nightly. Run Archive Old Assets on the first Sunday of every month. Any standard cron expression works, and the scheduler understands your tenant's timezone so the clock does not drift when the seasons change.

Event triggers fire pipelines in response to specific events: new asset uploaded, tag added or removed, asset moved to a collection, asset deleted. The Auto-Tag New Uploads pipeline is most useful as an event trigger: every new upload gets tagged automatically, with no manual intervention and no delay between upload and searchability.

Combine both approaches. A cron job handles the weekly export. Event triggers handle real-time tagging. The two modes cover the two shapes of creative work: the scheduled deliverables your clients expect on a cadence, and the continuous flow of new output that needs to be organised the moment it lands.

Pipelines Compose with the Rest of Numonic

The Pipeline Builder is not a standalone tool. It reuses the primitives the rest of the product is built on. The search grammar that powers the search bar is the same grammar that drives the select stage. The auto-curation and AI-tagging features are available as transform stages. The export presets you configure for the Gallery are available as output stages. If you know how to use the rest of Numonic, you already know how to build a pipeline.

This is how Numonic is designed to work. Every new capability ships as a composable stage, because the agent-first API that underlies the product makes composition the default. The same pipelines you build in the visual builder can be triggered by an AI agent, a webhook, or a cron expression. Creative automation is not a separate layer. It is the shape of the platform.

Key Takeaways

• The Pipeline Builder automates repetitive asset processing with visual or text-based workflows
• Twenty-nine stage types across select, filter, transform, and output cover every common operation
• Eight built-in templates handle the most common workflows out of the box, and every one is fully editable
• Cron scheduling and event triggers let pipelines run without manual intervention
• The text DSL gives power users a fast alternative to the visual builder, using the same grammar as the search bar