Freepik Spaces Tutorial: EVERY Node Explained (Magnific AI)

Summary of the tutorial (Freepik “Spaces” / Magnific AI): “EVERY Node Explained”

The video is a guided walkthrough of Freepik Spaces (an AI workflow builder) focused on node-based visual content creation. The creator emphasizes that using nodes and connections lets you build scalable, repeatable “content systems”—not just single outputs. With a properly designed workflow, you can generate large batches (e.g., ~100 consistent assets) from a small set of inputs.

---

Core concept / workflow scaling

• Spaces uses a drag-and-drop canvas where you connect nodes to create pipelines for text → images → video (and further transformations).
• The creator claims—after testing similar platforms—that Freepik Spaces is the favorite for this kind of workflow.
• Workflows can be designed so you can tweak initial inputs and regenerate many outputs with one click.

---

Main “basic” nodes explained

1) Text node

• Used to write prompts for downstream nodes (image/video generators).
• By itself, it doesn’t generate anything until connected.

2) Image generator node

Creates images using:

• A direct prompt (text inside the node), or
• A connected Text node as the prompt source.

Configurable features include:

• Number of creations (batch size)
• Choice of AI image generator model (example mentioned: “Google image gen 3”)
• Aspect ratio selection

Reference-image capability

• The image generator can take an image output from earlier nodes as a reference to generate “matching style/composition” results.
• It can also accept multiple reference images into a single image generator node (example: an AI headshot workflow with multiple references).

3) Video generator node

Creates videos from:

• Text prompts (e.g., “create a video of a cat”), or
• An image input (image → video).

Video control via ports:

• Start image port: defines the starting frame
• End image port: defines the ending frame
• A video prompt describes what happens in between (example: “cat staring at the lens… and then walking away”).

Video output/ports behavior:

• After generation, outputs can include:
  • Video
  • Audio, which can be routed to other nodes
• The tutorial suggests you can feed generated video into other video steps (mentioned conceptually as video-to-video).

---

“Most crucial scaling” node

4) Assistant node

• Functions like a ChatGPT-style chatbot node for text-based tasks.
• Supports using other LLMs (examples named: Gemini 3 Pro, Claude Sonnet 4.5).
• UI clarification: a toggle/indicator shows whether the content shown is prompt vs result.
• Can output in different formats, including:
  • Text
  • List objects (used heavily for scaling)

5) List node (key batching/scalability mechanism)

The creator calls the List node the most important for scaling workflows.

• Instead of manually creating many separate nodes/prompts, the list contains multiple items (e.g., animals, environments).
• When connected to an image generator:
  • The image generator automatically runs for each list item
  • The number of generated results corresponds to the list length (example: list of monkey/dog/cat → generates for each)

Practical advantage

• If you needed to create 100 items, List avoids repeating edits like “put on a white background” for each one.

Real workflow example: AI headshots

• Input: a selfie
• Assistant generates a list of environments/scenes (example: 20 professional environments)
• List feeds into the image generator to produce headshots in each environment
• Output is again a list of resulting images

---

Post-processing node

6) Image upscaler node

• Enhances image quality after you already have acceptable results—helping avoid wasting credits on overly-perfect initial generations.
• Configurable:
  • Upscale factor: 2x / 4x / 8x / 16x (example used: 4x)
  • Mode: precision (clarity) vs creative
  • Model choice (default mentioned)
• The creator demonstrates that the upscaled image appears much clearer than the original.

---

Other nodes / tools reviewed quickly

Upload / Assets / Stock

• Upload node: add files into the workflow.
• Assets node: view prior creations/history and reuse previous outputs.
• Stock node: choose from a stock media library (some appear AI-generated; can include usable real photos).

Image editor node

Connected to an image input; supports:

• AI expand (aspect ratio expansion)
• Camera angle control (example tried, but results looked mostly unchanged/wrong)
• Remove background
• Basic adjustments: exposure, brightness, contrast, saturation, temperature, etc.

Variations node

• Generates multiple variations from a base photo.
• Produces a batch (described as categories like back view, close-up, long shot, extreme close-up, etc.).
• Quality is noted as sometimes not great; recommendation is to use Image upscaler afterward.

Video combiner node

• Concatenates multiple sequential video clips into a longer video.
• Tutorial notes constraints:
  • Clips should share aspect ratio and size
  • Otherwise it may produce undesirable padding/letterboxing.

Audio nodes (brief)

• Music generation (examples: “11Labs music” and “Google lia”)
• Sound effects:
  • Adjustable duration (example: 3–30 seconds)
  • Loop toggle for continuity
• Text-to-speech:
  • Supports voice models (examples: 11 Labs models, Gemini 2.5 Pro)
  • Output described as natural human voice

Workflow UI organization nodes

• Sticky note: documentation/commenting inside the workflow
• Stickers: visual labels
• Group: neatly group multiple connected assets/nodes and name the group
• Also mentions renaming creations by double-clicking

---

Final takeaways emphasized

• The “main building blocks” covering most use cases:
  • Text, Image generator, Video generator
  • Assistant, List
  • Image upscaler
• The List node is positioned as the core step for turning a simple workflow into a scalable batch system (e.g., generating 100 images from structured inputs).

---

Main speakers / sources

• Primary speaker: the tutorial creator (no name provided in the subtitles)
• AI models mentioned as sources inside the Assistant node:
  • GPT/ChatGPT-style LLM
  • Gemini 3 Pro
  • Claude Sonnet 4.5
  • 11Labs
  • Gemini 2.5 Pro