PonPon Flow: Build Visual AI Pipelines Without Code

If you've ever found yourself repeating the same sequence of steps — generate an image, remove the background, upscale it, add text — then Flow is built for you. It turns manual multi-step workflows into automated pipelines you can run with one click.

What Flow is

Flow is a node-based visual pipeline builder. You drag nodes onto a canvas, connect them with wires, and the data flows from input to output. Each node represents an action: generate with a specific model, apply a tool, branch based on a condition, or output to a destination.

If you've used tools like ComfyUI, Unreal Blueprints, or even Zapier, the paradigm will feel familiar. But Flow is specifically designed for AI media production — the nodes map directly to PonPon's models and tools.

Node types

Source nodes

Source nodes are where data enters your pipeline:

• Text Input — A prompt or text string. Can be static or parameterized so you can swap prompts without rebuilding the pipeline.
• Image Upload — An uploaded image that feeds into image-to-video models or processing tools.
• Batch Input — A list of prompts or images. The pipeline runs once per item automatically.
• Canvas Selection — Pull items directly from a Canvas board into your pipeline.

Model nodes

Every PonPon model has a corresponding node:

• Sora 2 — Text-to-video or image-to-video generation with full parameter control.
• Kling 3.0 — Video generation with motion intensity and camera control parameters exposed as node inputs.
• Veo 3.1 — High-fidelity video generation with aspect ratio and duration controls.
• Seedance 2.0 — Dance and motion-focused video generation.
• Nano Banana Pro — Fast image generation with style and composition controls.

Each model node exposes all parameters as adjustable inputs on the node itself. No need to remember syntax — just adjust sliders and dropdowns.

Tool nodes

• Upscaler — Increase resolution to 2x or 4x. Takes an image or video frame as input.
• Background Remover — Remove backgrounds from images. Outputs a transparent PNG.
• Text Editor — Add text overlays with font, size, color, and position controls.
• Multi-Angle — Generate multiple viewing angles from a single product image.

Logic nodes

• Condition — Branch the pipeline based on metadata (e.g., if resolution > 1080p, skip upscaling).
• Loop — Repeat a section of the pipeline N times with incrementing parameters.
• Delay — Add wait time between steps (useful for rate-limiting or staged outputs).
• Merge — Combine multiple pipeline branches back into one stream.

Output nodes

• Save to Gallery — Store results in your PonPon gallery.
• Save to Canvas — Place results onto a specific Canvas board and section.
• Download — Package outputs for local download.
• Webhook — Send results to an external URL for integration with other tools.

Building your first pipeline

Here's a practical example: creating a product image pipeline that takes a raw product photo and produces social-ready assets.

Step 1: Add an Image Upload source node. Upload your product photo.

Step 2: Connect it to a Background Remover node. The product gets cleanly isolated.

Step 3: Connect the remover output to a Multi-Angle node. Set it to generate 4 angles (front, 3/4 left, 3/4 right, top-down).

Step 4: Connect each angle output to an Upscaler node set to 4x.

Step 5: Add a Text Editor node to overlay your brand name on each upscaled image.

Step 6: Connect all outputs to a Save to Canvas node pointed at your "Product Assets" board.

Hit run. The entire sequence executes automatically, and you get 4 high-res, branded product images from a single upload.

Batch processing

Flow's batch capability is where it gets truly powerful. Instead of a single Image Upload, use a Batch Input node and drop in 50 product photos. The entire pipeline runs for each photo automatically. Walk away, come back to 200 finished assets.

Batch processing respects your credit balance — the pipeline pauses if credits run low and resumes when you top up. A progress bar shows completion percentage and estimated time remaining.

Templates

Flow ships with pre-built templates for common workflows:

• Social Content Kit — Prompt → generate video → add text overlay → export in 3 aspect ratios
• Product Photography — Upload → remove background → multi-angle → upscale → brand overlay
• Video Storyboard — Multiple prompts → generate clips → sequence → preview
• Style Exploration — Single prompt → generate with 4 different models → compare side by side
• Thumbnail Factory — Video → extract best frame → upscale → add title text

Templates are fully editable. Start from a template, customize the nodes, save as your own reusable pipeline.

Advanced techniques

Parameterized prompts

Use variables in text nodes: "A {product} on a {background} surface, {lighting} lighting." Then feed in a spreadsheet of product/background/lighting combinations via Batch Input. One pipeline, hundreds of unique outputs.

Feedback loops

Connect an output back to an input with a Condition node in between. For example: generate an image → check if it meets a quality threshold → if not, regenerate with adjusted parameters. This creates self-refining pipelines.

Model comparison pipelines

Split a single prompt to multiple model nodes running in parallel. Compare Sora 2, Kling 3.0, and Veo 3.1 outputs side by side in a single run. Useful for finding which model handles specific scene types best.

When to use Flow vs. Canvas

Canvas is best for freeform creative exploration — brainstorming, comparing, organizing. It's manual and spatial.

Flow is best for repeatable production — batch processing, multi-step automation, standardized output. It's automated and sequential.

Many users combine both: explore ideas on Canvas, then build a Flow pipeline to produce final assets at scale. Canvas for creativity, Flow for production.

Flow turns PonPon from a generation tool into a production system. Once you've built a pipeline that works, you can reuse it indefinitely — same quality, zero repetitive manual work.