Creative Workflows
Connect prompts, references, LLMs, and creative models on one canvas. Explore several directions, develop the one you love, and turn the whole process into a workflow you can use again.
The first result is a starting point. Connect one prompt to several image models and see how each interprets the same idea. Keep the results together, compare the details, and choose what deserves another pass.
Same moth. Three different interpretations.
A single sculpture brief feeds FLUX, Recraft, and Ideogram. Translucent wings, color, and fine structure take a different form in each result.
Give each branch a question to answer, so every generation helps you make a decision.
One Text node can supply several prompt inputs. Edit the shared wording in one place, then run the branches you want to compare.
Duplicate a branch and change its prompt, reference, or model settings. Keep the earlier result beside it while you judge composition, texture, and detail.
Connect the chosen result to a compatible editing or video node. The other directions stay on the canvas, ready to revisit.
Your references contain more direction than a few keywords can capture. Connect them to a vision-capable LLM, ask it to describe what matters, and use that response in the next stage of your work.
Read the visual language of a moodboard.
Claude compares the three moth studies and identifies shared traits: translucent wings, pale colors, gemstone details, and a sculptural quality. Those observations give the next brief something specific to build on.
Give the chat node a clear creative job. Its text output can become input for the rest of the flow.
Ask for the palette, materials, lighting, and composition in your references. Review the response against the images and keep the details that express your intention.
Ask the LLM to turn that analysis into a focused image or video prompt. Connect its text output to a compatible prompt input to carry the direction forward.
Request a shot list with one description per line. Text Split can turn those lines into separate outputs for individual storyboard branches.
Build prompts from reusable pieces. Keep the lighting and composition consistent while you change the product, explore materials, or send each shot down its own branch.
Prompt Template
Start with “Studio photograph of {product}, on a {material} plinth, soft side light, generous negative space.” The named slots become inputs. Connect a new product or material while the rest of your direction stays in place.
Inline choices
Use {travertine|brushed steel|frosted glass} inside the template to create a choice of materials. Switch the selection, then run the image model when you are ready to explore that version.
Text Concat
Keep a product description, campaign direction, and LLM-generated visual notes in separate nodes. Text Concat joins their text into one prompt, with the separator you choose. Each source remains easy to change.
Text Split
Give an LLM’s response a predictable format, such as one shot per line. Split on new lines, then connect each part to its own generation branch. A wide establishing shot and a close-up can develop independently.
A reference sets the visual direction. An image model helps you explore it. A video model gives the chosen frame movement. Build those stages together so you can see where the result came from and decide what to change next.
A connected canvas gives the creative process a visible structure, with a place for each decision.
Leave the source images alongside your working branches. Compare a new direction with the original without reconstructing the brief from memory.
Use one prompt to establish the image and another to direct motion. The video brief can describe camera travel and subject movement without rewriting the visual concept.
Connections carry inputs between compatible nodes. Run the node or Recipe when you are ready; editing a prompt does not itself launch every connected generation.
When a flow works, turn the selection into a Recipe. Choose the inputs someone can change and the outputs they receive. Your model choices and connections stay inside, ready for another brief.
A complete process, ready to run again.
This Botanical Campaign Recipe puts the brief beside the image result. Change the product direction, run the Recipe, and review the next interpretation without rebuilding the flow.
A useful Recipe has a clear interface. Expose the creative choices that matter, keep the internal setup together, and let the result become the next input.
Start with a prompt and one model. Add a branch when you want another interpretation, or connect an editing or video node when you want to develop a result. You can also start from the Recipe library and bring your own inputs.
Yes, when the output and destination input are compatible. For example, text can feed a prompt, an image can feed an image-to-video model, and a vision-capable chat node can accept image references. Check each node’s inputs or read the socket guide.
No. Choose the node or Recipe you want to execute. Fuser resolves the inputs it needs for that run. Finish an upstream generation before running a downstream node that needs the new result.
Share the project using its sharing controls, or package a tested flow as a Recipe with clear inputs and outputs. Multiplayer and collaboration lets you work together around the same canvas.
Bring a brief, a reference, or an unfinished idea. Make the first connection and see where it takes you.