Build a Flow you can read, run, and reuse.
Apollo Flows connect models, creative apps, and media tools on one visual canvas. Learn what every node, port, color, and status means, then build a pipeline that runs the same way every time.
How a Flow works
A Flow is a saved graph of work. Nodes perform each step, connections control order and move data, and every run keeps its own inputs, outputs, status, and credit usage.
Start
Collect the text or media that changes on each run.
Do the work
Generate with models or process media with focused apps.
Route results
Branch, collect, split, merge, or pass outputs downstream.
Review
Open Results to inspect every node and saved output.
Read the canvas at a glance
The canvas uses two color systems. A node's accent color identifies the kind of work it does. A port and its line color identify the value moving between nodes.
Anatomy of a node
Every node follows the same layout, so a large Flow stays readable.
Image Studio
Studio node
Left in, right out
Inputs are on the left edge. Outputs are on the right edge.
Accent means node family
The stripe on the node's left describes its role, not its output type.
What each line color carries
A connection can only join compatible ports. Follow the color to see which value is moving.
Trigger
Gray. Controls execution order without carrying creative input.
Text
Blue. Prompts, scripts, captions, and other text.
Image
Green. Uploaded, generated, or edited images.
Video
Purple. Source clips and generated video.
Audio
Orange. Speech, sound, and extracted audio tracks.
Any value
Slate. A flexible value used by Collect and Split.
Trigger lines control order
A gray trigger connection makes the control path explicit. A node waits for its incoming dependencies before its layer can begin.
Colored lines move values
A data connection carries an upstream output into a compatible downstream input. Its color always matches the payload type.
Drag from an output dot on the right to a compatible input dot on the left. Built-in templates pair their data path with an explicit trigger path.
Node types and what they do
The colored stripe on each node identifies one of six families. Mix families freely as long as their ports are compatible.
Flow
Start defines run inputs. End marks the completion path and saves the finished run.
Text
Generate or transform prompts, scripts, captions, lists, and structured creative directions.
Studio
Create or edit images and videos with a selected AI model and its settings.
Audio
Turn connected text into speech with a selected voice and audio model.
App
Run focused creative jobs such as product shots, lip sync, background removal, and enhancement.
Tool
Store reusable assets, collect or split items, merge media, or extract audio without another model.
Create your first Flow
Build one readable path first. Add branches and batching after the main result works.
01
Choose a starting point
Open a template to study a working graph, or create a blank Flow with Start and End.
02
Add nodes
Pick text, studio, audio, app, or tool nodes from the palette. Each node should have one clear job.
03
Configure the work
Choose models, prompts, app options, output counts, and other settings. The node shows its estimated credit cost.
04
Wire the graph
Connect matching data ports, then make the execution path clear with trigger connections.
05
Choose where inputs live
Put changing values on Start. Put reusable files and text in Storage.
06
Run, inspect, refine
Resolve any validation issue, press Run, then inspect outputs, errors, duration, and credits in Results.
What happens when you press Run
Apollo saves a snapshot of the graph, plans the dependency order, and records the result of every node. You can keep editing the Flow without changing an older run.
- 01
Collect inputs
The run dialog asks for every field defined on Start and uses the files already pinned in Storage.
- 02
Plan layers
Connections decide which nodes must finish first. Independent nodes in the same layer run in parallel.
- 03
Fan out or gather
A list can run a downstream node once per item. Collect gathers items into a batch, while Split routes them to indexed outputs.
- 04
Save results
Each node records its status, outputs, error details, duration, item count, and credits used.
How to read node status
Waiting for its dependency layer.
The node is processing now.
Outputs are ready to inspect.
Open the node to read the error.
Required upstream data failed.
Results shows the saved canvas for each run. Select any node in that snapshot to inspect its exact output and credit usage.
Start fields or Storage?
Put changing information at Start and reusable source material in Storage.
Use Start for each run
Product photos, campaign briefs, scene descriptions, and source clips can be requested when someone presses Run.
Use Storage for reusable assets
Brand references, identity stills, fixed copy, and other files can stay pinned to the Flow across runs.
Learn from a working template
Open a built-in Flow to trace its node colors, typed data lines, trigger path, branches, and expected outputs. Copy it when you are ready to customize.
Fake influencer
A full AI creator pipeline. Copy, Image Studio, voice, and a lip-sync app, wired so you do not hop between tools.
Describe the influencer in Storage, type the scene at run, invent the face, write the UGC hook, speak it, and lip-sync.
Viral short
A vertical-video pipeline. Invent the bit, paint the first frame, then a 1080p clip that follows the scene and camera path from that still.
Invent a talking-head or WTF short (a yeti in Paris, a cat on a bike), paint the first frame in 2K, then animate it in 1080p with that scene, camera path, and the still.
Ad Maker
A product-ad pipeline. Add your product photo at run. The writer invents copy plus scene and camera, the reshoot follows that scene, then the clip follows the camera path plus the still.
Invent campaign copy plus scene and camera, reshoot the product into that scene, then animate the still along the camera path.
No Flow overhead
A Flow costs exactly the credits used by the paid nodes that actually run. Apollo does not add a setup, connection, per-step, or orchestration surcharge.
The canvas itself costs nothing
Building the graph, connecting nodes, routing values, and saving run history do not add credits to the total.
Node costs stay visible
Each paid node shows an estimate before the run. Results records the credits it actually used after processing.
Run total
Credits used by the generation and app nodes that ran
Flow overhead
0 credits
Flow questions
- Do Flows cost more than running tools separately?
- No. A Flow adds zero overhead. You only spend the credits used by the paid generation and app nodes that actually run.
- What do the connection colors mean?
- Gray is trigger, blue is text, green is image, purple is video, orange is audio, and slate is a flexible value used by tools such as Collect and Split.
- Why use both a trigger and a data connection?
- The trigger makes the execution order explicit. The data connection supplies the value the node uses. Built-in templates show both paths so the graph is easier to read.
- Should an input go on Start or in Storage?
- Use Start for values that can change on every run. Use Storage for reusable files or text that should remain pinned to the Flow.
- What happens when one node fails?
- The run records the failed node and its reason in Results. Nodes that require data from the failed node are skipped.
- Can I begin with a template?
- Yes. Viral short, Ad Maker, and Fake influencer are ready to inspect and copy from the Flows page.
Ready to build a Flow you can trust?
Open a template, trace its colors and execution path, then make a copy and change one node at a time.