CLASS 8June 17 · Node-Based Workflows in Krea

Class 8: Node-Based Workflows in Krea. "Krea is Microsoft Paint. Comfy UI is free Photoshop."

Taught bySeb SagaSebastian Sagastume · Founder, Seb Saga LLC
Pratt After Effects & video editing instructor since 2022

The class in brief

Guest instructor Seb Saga taught node-based AI workflows in Krea: chaining single-purpose blocks into a reusable pipeline instead of re-typing one prompt at a time. Seb is a video producer and animator whose clients include Disney and the New York Islanders.

Six archetypes, a live build from blank canvas to finished video, and a blunt account of where the tool breaks. After this page you can name the right archetype for a job and let an agent build a workflow's bones from one sentence. You can also know when to walk away from Krea for ComfyUI instead.

This class moves. Watch the four recovered demo clips with audio →

From prompts to pipelines

Why this matters · 6:04 PM

A node does one job. Chain enough of them and the workflow runs itself.

Seb's core thesis, opening the night : nodes are "smart LEGO blocks" that chain into a reusable pipeline, "basically programming, but with visual blocks."

What the room would build that night fell into three shapes: scalable product placement, a video chain, and multi-layer compositing .

Before reaching for a node canvas at all, Seb gave six diagnostic questions to ask first :

Does it need to scale?
Can it be reused or handed off?
Are you comparing models?
Is it repetitive?
Do you need several models orchestrated together?
Do you need absolute control? If the answer to this question is yes, the real answer is ComfyUI, a free, more advanced node tool built for full control, not Krea.

questions to ask before you build a node workflow at all. Answer no to all six, and one prompt still does the job.

things the room would build that night: product placement, a video chain, and multi-layer compositing.

The framework · 6:17 PM

Six ways to build a workflow

Lab benchcompare models or settings

Same prompt, several models, side by side. Seb's test: a skateboard trick, where one model nailed it and others missed entirely. Also test aspect ratios; models perform differently across them, a training-data artifact.

Factoryone input, same process out, run over and over

Multi-angle character generation, turnaround sheets, a product photo becoming a model wearing it plus a 360 view. The same output shape, run over and over.

Multiplierone asset, many scenarios

A single product shot placed into several environments at once: the product-placement pattern the night's live build used.

Relayoutput feeds the next model

Image into edit into video, one model's output becoming the next model's input, down the line.

Compositorassemble the pieces, then unify

Generate scene layers separately, remove backgrounds, assemble, then run the flat composite through an editing model to match shadows and light.

Control rigstructural prompting

Pose skeletons and depth maps, called ControlNets, work as a third type of prompt, alongside text and images. This is ComfyUI territory, not Krea's. Seb summed up the split with this page's hero line: Krea as Microsoft Paint, ComfyUI as free Photoshop.

The Factory, product photo to turnaround set

The exercise · 6:52 PM

Building the first chain live

The build

Blank canvas to a finished squirrel-in-a-tuxedo video.

Seb built the first chain from nothing: a Z-Image node with a plain sunset prompt .

Then a Text node fed an LLM Call node, which fed the image node, three blocks long . From there, a Qwen edit put the squirrel in a tuxedo and top hat, and an LTX-2 node turned the result into a short video.

"The Krea agent builds the bones. You still write the prompts that matter."

The preset that saves the step

Seb's LLM-node system prompt, saved and reused across every workflow he builds: "Expand on the user's image prompt. Ensure to describe settings, tone, feel, angle, color palette, and any other descriptions when relevant. Respond with only the detailed prompt." One block, written once.

The agent builds the bones

Asked in plain language to place a sweater into four environments, the Krea agent built the prompt-writer, the splitter, and four generation nodes on its own . Scaled up to a cardigan and eight environments, the pattern held .

Cheap first, then swap

The Photoshoot App demo ran a skateboard product photo through cheap models first, four style variations for the cost of testing. Only then did anyone touch a frontier model .

Compositing, then the video chain

The craft · 7:45 PM

Assemble the pieces separately, then let one model unify the light.

Post-break, the Compositor archetype got its full demo: a fox, a stone bowl, and a zen-garden background generated as three separate layers, backgrounds removed. They were then handed to a single editing model with one instruction, integrate the layers with realistic shadows and light .

The video chain closed the demo block: a subway phone ad built from a green-screened phone, three Seedance clips stitched end to end . A 2x upscale finished it on the way out .

Subway ad video chain — The subway ad, stitched from three clips

Methods and prompts

Five methods to take with you

METHOD 01 · THE SIX QUESTIONS part of: keep the judgment human

Decide if you need nodes at all

Before building anything, answer the six diagnostic questions yourself. A single prompt still does the job for most tasks; nodes earn their complexity only when the answer is genuinely yes.

Where it came fromTaught by guest Sebastian Sagastume, known as Seb Saga, at the start of class 8, as six diagnostic questions to ask before building any node workflow in Krea.Use it whenUse this before opening a node canvas, when you are not sure whether a task actually needs a multi-step workflow or a single prompt would do.

Same categoryJudge AI output with your own trained eye (class 1) · The sea-of-sameness check (class 2) · Pick the relationship on purpose (class 2)

Working prompt

Here's the task I'm about to automate: [describe it]. I'll answer these six questions myself first, then you check me.

Does it need to scale past a handful of outputs?
Could it be reused or handed off?
Am I comparing models or settings?
Is it repetitive?
Does it need several models orchestrated together?
Do I need pixel-level control?

My answers: [yours]. Tell me honestly whether I need a node workflow, or a single prompt still does the job.

You will know it worked whenit tells you plainly whether a single prompt still covers the task, based on your own six answers, not a default push toward nodes.

METHOD 02 · THE AGENT part of: build tools not chats

Let the agent build the bones

Describe the outcome in plain language and let the agent lay out the nodes; you review and fix the prompts it writes, rather than wiring every connection by hand.

Where it came fromDemonstrated by Seb Saga in class 8's live build, where he described an outcome in plain language and let Krea's built-in agent lay out the node structure before reviewing the prompts it wrote.Use it whenUse this when starting a new node workflow and you would rather describe the goal in words than wire every node connection by hand.

Same idea elsewhereStand up a reusable Project or Gem (class 4) · Weight your references on purpose (class 5) · Let one model write the prompt for another (class 6)

Working prompt

I want a node workflow that [describe the outcome, e.g. "places this product photo into 4 lifestyle environments"]. Build it: the input node, a prompt-writer step, and one generation node per output. Use the cheapest capable model for a first pass. Show me the prompts you wrote before I run it.

You will know it worked whenit shows you the prompts it wrote for each node before running anything, so you can catch a bad one before it generates.

METHOD 03 · THE PRESET part of: context beats prompts

Save a reusable prompt-enhancer preset

Seb's own system-prompt block, saved once and applied to any image node. One saved instruction, reused across every workflow instead of rewritten each time.

Where it came fromSeb Saga's own saved system prompt, shown in class 8 as a reusable LLM node instruction he applies to any image-generating node instead of rewriting it each time.Use it whenSave this once you find an instruction that reliably improves your image prompts, so every future node workflow can reuse it without retyping.

Same idea elsewhereControl the input and build gradually (class 2) · Show it your voice and not just the output (class 3) · Evolve the reference instead of restarting (class 5)

Working prompt

Expand on the user's image prompt. Ensure to describe settings, tone, feel, angle, color palette, and any other descriptions when relevant. Respond with only the detailed prompt.

You will know it worked whenevery image node using this preset expands your short prompt into settings, tone, angle, and color palette without you typing them each time.

METHOD 04 · COST DISCIPLINE part of: reject the first draft

Build cheap and then swap the model up

Verify the whole chain works on the cheapest model that can do the job, then upgrade only the one node that needs it. A two-click dropdown change, not a full rebuild.

Where it came fromSeb Saga's cost-discipline habit was taught in class 8. Build and verify a full node chain on the cheapest capable model, then upgrade only the one node that actually needs a stronger model.Use it whenUse this when setting up a new workflow and you want to control cost, testing the whole chain cheaply before spending on a frontier model.

Same idea elsewhereSet up a debate partner (class 4) · Critique mode (class 10) · Find where the workflow breaks (class 11)

Working prompt

Build this workflow with [name the cheapest capable model, e.g. Z-Image]. Once every node runs clean end to end, tell me which single node would benefit most from a frontier model, and swap only that one.

You will know it worked whenit names exactly one node worth upgrading to a stronger model, not a blanket recommendation to swap the whole workflow.

METHOD 05 · THE 2X2 TRICK part of: generate multiples

The 2x2 grid trick

Ask for four variations in one grid instead of four separate generations. Better scene consistency, and four options for one generation's cost.

Where it came fromShared by Seb Saga in class 8 as a way to get four scene variations from a single generation instead of four separate ones, with better consistency between the results.Use it whenUse this when you want several options from one image generation instead of paying for and comparing four separate runs.

Same idea elsewhereRefine with constraints (class 9)

Working prompt

Generate a 2x2 grid showing four variations of this scene: [describe it]. Keep the composition and framing consistent across all four so I can crop out whichever one wins.

You will know it worked whenthe four variations share the same composition and framing, so you can crop any one out without the scene shifting underneath it.

Where it broke · all night

The line splitter and the honest downsides slide

Seb put Krea's weaknesses on screen before anyone hit a bug :

Buggy.
Less scalable.
Credit-hungry.
Less control than the alternatives.

The line splitter, a utility that fans one output into several prompts, was the night's biggest live failure. Two workflows silently produced nothing, and the fix was asking the agent to abandon the splitter and write a prompt into each node by hand .

The shelf

Tools and references

Tools that night

krea.ai/nodes, plus the built-in Krea agent (the "K" button)
ComfyUI (comfy.org), local and free; Comfy Cloud, ~$20/month hosted
Named as peer node platforms: Runway nodes, Weavy, Flora
Cheap tier: Z-Image, LTX2, Qwen Image Edit, Gemini 2.5 Flash. Frontier tier: Nano Banana Pro, Seedance, Claude Sonnet/Opus
Topaz, Magnific (upscalers); Photopea (free Photoshop-alike, compositor round-trips)

Named in the room

Seb Saga: sebsaga.com · [email protected], open to questions and connections
The skateboard "Benihana" trick, the Lab Bench's test case across models
This page carries the frames that teach the lesson. The full run of stills from the recording is not included.

Frames from this class

40 captures, in order. Click any one to see it full size.

From prompts to pipelines

Bio slide for Sebastian Sagastume with a photo, owner and operator credit, client list, and contact links. — Guest instructor Sebastian Sagastume, who goes by Seb Saga, runs a video, animation, and drone production business and has done client work for Disney, the New York Islanders, and a Netflix documentary.

Workshop outline table with four time blocks across the bottom of the slide. — The night's agenda ran from 6pm to 9pm: an intro to nodes and workflows, a break, a ComfyUI segment, then wrap up and Q&A.

Slide showing jeans product photos composited into several skateboarding action scenes. — One of three example workflows Seb previewed: a scalable product placement pipeline that takes two input images and one prompt and turns a single pair of jeans into skate photography across multiple scenes.

Node diagram showing a chain from Generate Image through Change Screen, Add Foreground, Remove Elements, Rotate Phone, into an Image to Video step. — The video chain workflow diagram: generate an image, change the screen, add or remove foreground elements, rotate the object, then feed the result into an image to video model using a start and end frame.

Slide showing a fox and a stone bowl composited into a Japanese zen garden image. — The third example workflow, compositing: separate layers, a fox, a stone bowl, and a garden background, combined into one photorealistic zen garden scene.

Slide with three stacked boxes describing what a node is next to an example three step node chain. — Seb's core definition: nodes are like smart Lego blocks that snap together into a workflow, illustrated with a chain of a realistic image model feeding a creative upscaler feeding a video generator.

Slide listing six yes or no questions in a two column grid. — The six diagnostic questions Seb gave for deciding when to use nodes instead of a single prompt: does it need to scale, can it be reused, and are you comparing models? Is it repetitive, does it need several models, and do you need absolute control?

Six ways to build a workflow

Slide with a skateboard prompt on the left and three skateboarding photos labeled by model on the right. — The Lab Bench workflow archetype: the same skateboard trick prompt run across three models side by side, Nano Banana Pro, Luma UNI-1, and ChatGPT, to compare results.

Slide titled The Factory showing a man on a sidewalk linked by an arrow to a generated street photo. — The Factory workflow archetype: one input photo of a person automatically becomes the same multi-step generated output, here a street scene, over and over.

Slide showing a sweater photo linked to a model wearing it and a four angle turnaround sheet. — A Factory example built for apparel: a Halloween sweater product photo goes in, and a model wearing it plus a full 360 degree turnaround sheet come out automatically.

Slide showing a storefront photo linked to an enhanced version of the same storefront, partly obscured by foreground leaves. — A Factory example using a storefront photo: the same node chain automatically produces an enhanced, upscaled version of the shot.

Slide showing a cardigan product photo linked to four lifestyle photos of it worn in different settings. — The Multiplier workflow archetype: one product photo, a cardigan, placed into four different lifestyle scenes and lighting setups in a single run.

Slide showing a squirrel image linked through Generate Image, Edit Image, and Generate Video steps to a tuxedo squirrel scene. — The Relay workflow archetype: output feeds the next model down the line, shown here as a squirrel image generated, then edited into a tuxedo, then turned into video.

Slide titled The Compositor showing the fox and garden layers on the left and the full art director style prompt text on the right. — The Compositor workflow archetype, with Seb's actual system prompt shown: it asks the model to write four separate compositing layer prompts for one zen garden scene, background, subject, and foreground, before assembling the final image.

Slide titled The Control Rig, ComfyUI, with a grid of six example generated images. — The Control Rig archetype uses ComfyUI for structural prompting with poses and depth maps, shown here in a grid of relighting, product shot, and pose control examples.

Slide titled Comfy is King with a grid of advanced example outputs. — Seb's case for ComfyUI's ceiling: advanced node graphs powering 3D and Gaussian splat generation, pose transfer, relighting, and upscaling, work he says Krea cannot yet match.

A single Z Image node panel with a squirrel prompt and generation settings. — The live build begins: a Z Image node set to generate a squirrel sipping a latte in a forest, the cheap open source model Seb starts every workflow with before swapping up.

A Text node connected by a line to the Prompt input of a Z Image node. — Seb demonstrates the basic mechanic of node linking by wiring a Text node's output directly into the Z Image node's prompt input.

Three connected node panels, Text, Z Image, and Seedance, shown side by side. — The full relay chain in miniature: a Text node feeds a Z Image node, whose output feeds a Seedance video node, turning a still prompt into a video generation setup.

Slide-like panel showing a sneaker product photo and four generated lighting variant images in a chat interface. — The Krea agent (the K button) built this from a plain language request: one input sneaker photo, four different lighting conditions generated automatically.

Node graph with a central hub connected to many small environment image nodes, plus an agent chat panel on the right. — The Krea agent building a bigger pipeline from a single instruction, an LLM node plus a line splitter write nine different product photography environment prompts, then wire them out to separate generator nodes.

Same node graph as before with the agent chat panel showing a Stage 2 progress message. — The agent reports its own progress in the chat log: it has wired the line splitter's nine outputs into nine separate edit nodes for stage two of the pipeline.

Node graph and chat panel with a user message asking the agent to arrange and group the nodes. — A student then asks the agent to tidy the sprawling graph, and the agent responds by arranging and grouping the nodes into labeled sections, one of the capabilities Seb called out.

Slide with an eight box grid listing agent capabilities. — Seb's summary of what the Krea agent can do: build full pipelines from scratch, write prompts, group and organize nodes, preview cost, batch edit settings, suggest models, and answer how-to questions as you work.

Where it broke

Slide with a four box grid reading Buggy, Less Scalable, Credit Hungry, Less Control. — Seb's honest rundown of Krea's weak points: it is buggy, less scalable, credit hungry, and offers less control than an alternative like ComfyUI.

Slide showing the text sebsaga.com centered on a gradient background with contact links below. — Sebastian Sagastume (Seb Saga) closed with an open invitation to connect, sebsaga.com, his LinkedIn, and [email protected], repeated more than once during the class.

Slide with Goal, Tools or Info, Deliver, and Remember sections for the in class activity. — The hands on activity brief: use Krea nodes to build a workflow of your choice in 60 minutes, deliverables were five photos, one final composited image, and a five to ten second video.

Building the first chain, live

Close up of a Z Image node showing a generated photo of a squirrel holding a latte cup in a mossy forest. — The finished result of the opening demo: the squirrel and latte image the Z Image node generated from the enhanced prompt, shown in the node's result panel.

A squirrel image connected to a Qwen Image Edit node with the tuxedo prompt and a small result thumbnail. — Qwen Image Edit node applies the prompt "give the squirrel a tuxedo and top hat" to the demo squirrel image, Seb Saga's live example of turning one generated asset into an edited variant inside a Krea node chain.

Two full images side by side, a squirrel in a forest and the same squirrel edited into a tuxedo and top hat. — The finished edit side by side with the original, showing Krea's Qwen Image Edit node turning a plain forest squirrel into one wearing a tuxedo and top hat.

Compositing & the video chain

A node canvas with a bowl image, a fox image, and a garden background image linked toward composited result thumbnails. — The zen garden "Compositor" workflow: separate fox, stone bowl, and garden background nodes feeding into a single composited scene, Seb Saga's example of the Compositor archetype from the six workflow types.

A node canvas with a chat panel showing a message about replacing a prompt splitter with individual nodes. — A student asks the Krea agent to swap the buggy line splitter for individual prompt nodes, and the agent's chat log confirms it deleted the LLM and splitter nodes and wired dedicated prompts directly into each clothing node.

The finished Band Logo Mockup Workflow: the bardetta band logo wired into four separate mockup nodes producing a t-shirt, vinyl sleeve, and concert poster all carrying the same logo, Seb Saga's live "Multiplier" example.

An LLM Call node with visible System Prompt text about expanding an image prompt, plus a Presets button. — The LLM Call node fully expanded, showing the saved Prompt Enhancer system prompt that tells the model to expand a user's image prompt with setting, tone, and camera details before passing it downstream.

Seb Saga walks through a band logo mockup workflow in Krea: upload a band's logo into one node, and the connected chain generates a t-shirt, vinyl sleeve, and concert poster mockup automatically.

Krea LLM node with a system prompt field showing the added instruction to place all mockups in Alexandria, Egypt. — Jina Alhenawi, a classmate in the course, built a logo mockup factory that customized every generated scene with one line in the system prompt: make sure all mockups are placed in Alexandria, Egypt.

ComfyUI node interface with a video upload panel showing footage of boats and mountains over water. — Sebastian Sagastume closes the class with a quick ComfyUI demo, loading his own shaky handheld footage of boats on the water in Japan into a video stabilization workflow.

ComfyUI split screen comparing the original shaky footage to the stabilized version of the same boats and mountains scene. — The ComfyUI workflow's result: a side by side comparison of Seb Saga's original shaky footage against the stabilized version, the same object removal and cleanup technique he uses in his own production work.

ComfyUI export settings panel with format and frame rate options next to the stabilized boats and mountains footage. — The final step of Seb Saga's ComfyUI stabilization demo: export settings for the cleaned up footage, set to compare the outpainted result against the original at h264-mp4 format.

Class 8: Node-Based Workflows in Krea. "Krea is Microsoft Paint. Comfy UI is free Photoshop."

From prompts to pipelines

Six ways to build a workflow

Lab benchcompare models or settings

Factoryone input, same process out, run over and over

Multiplierone asset, many scenarios

Relayoutput feeds the next model

Compositorassemble the pieces, then unify

Control rigstructural prompting

Building the first chain live

Blank canvas to a finished squirrel-in-a-tuxedo video.

The preset that saves the step

The agent builds the bones

Cheap first, then swap

Compositing, then the video chain

Five methods to take with you

Decide if you need nodes at all

Let the agent build the bones

Save a reusable prompt-enhancer preset

Build cheap and then swap the model up

The 2x2 grid trick

The line splitter and the honest downsides slide

Where this class connects

Tools and references

Tools that night

Named in the room

Frames from this class