# Input nodes Input nodes hold what a run starts from: text, images, video, audio and settings. When you publish, most of them become the space's form. Last updated: 2026-10-09 **Key facts** - Spacelab input nodes hold what a run starts from, text, images, video, audio and settings, and every input node is free. - A Spacelab Multi Image Input holds up to 10 images, and Background Input, Characters and Poses take up to 10 picks per node. - A Spacelab Document node reads one PDF, DOCX, PPTX, Markdown, TXT or CSV file up to 25 MB, with a PDF up to 50 pages and a CSV up to 500 rows and 100 columns. - A Spacelab Video Input's form field takes one clip of up to 20 seconds, and an Audio Input's form field takes one file of up to 50 MB. - Prompt, Image Pool and Pinterest Scraper are Spacelab input nodes that never become form fields, so what you set in them is the same for every visitor. Web page: https://sloosh.ai/docs/nodes-input Each entry says what the node does, then gives the same lines in the same order: Best for, In, Out, Settings, Cost, Limits, Example and Tip. A line that doesn't apply is left out. To find a node on the canvas, search Add node by name. Input nodes are free, so they have no Cost line. Form field, after Settings, says what the node becomes when you publish the canvas as a space. ## At a glance - [Prompt](https://sloosh.ai/docs/nodes-input.md#prompt): text you write, with `{{variables}}` other nodes fill in. - [Text Input](https://sloosh.ai/docs/nodes-input.md#text-input): words a space visitor types. - [Document](https://sloosh.ai/docs/nodes-input.md#document): a PDF, DOCX, PPTX, Markdown or CSV file for a model to read. - [Image Input](https://sloosh.ai/docs/nodes-input.md#image-input): one image, usually the main subject. - [Reference Image Input](https://sloosh.ai/docs/nodes-input.md#reference-image-input): one supporting image: a style, a pose or a mood. - [Multi Image Input](https://sloosh.ai/docs/nodes-input.md#multi-image-input): up to 10 images sent as one set. - [Background Input, Characters and Poses](https://sloosh.ai/docs/nodes-input.md#background-input-characters-and-poses): picks from your library. - [Image Pool](https://sloosh.ai/docs/nodes-input.md#image-pool): a folder of images, one drawn at random per run. - [Pinterest Scraper](https://sloosh.ai/docs/nodes-input.md#pinterest-scraper): images from a Pinterest search. - [Video Input](https://sloosh.ai/docs/nodes-input.md#video-input): one video clip. - [Audio Input](https://sloosh.ai/docs/nodes-input.md#audio-input): one audio file. - [Aspect ratio](https://sloosh.ai/docs/nodes-input.md#aspect-ratio): one ratio for every node it feeds. - [Number](https://sloosh.ai/docs/nodes-input.md#number): a number, with an optional min and max. - [Select](https://sloosh.ai/docs/nodes-input.md#select): a list of choices you set. ## Text ### Prompt Node: Prompt (`prompt`) - Inputs: none - Outputs: `text` (Prompt, text) Text you write on the node, passed on to a model. Press Add Variable to give it an input port, write `{{name}}` in the text, and whatever is wired into that port lands there when it runs. - Best for: the prompt of every model node. Image, video, music and text models have no prompt box of their own. - In: one port per variable. A variable takes any output, such as text, a number or a choice. - Out: the finished prompt text. - Form field: none. What you write is the same for everyone. To let visitors type part of it, wire a Text Input into a variable. **Example** ```text Prompt: A {{product}} on a marble table, soft window light, shot at eye level. Text Input named Product → the product variable ``` Wired into a model that takes references (Image Model, Video (references), Video edit on Seedance 2.5), a Prompt can tag them: type `@` and pick from On this node (`@Image 1`, `@Video 1`, `@Audio 1`) or From library, which adds the element as a new input wired into the next free slot. Tags renumber when you rewire. See [Tag with @](https://sloosh.ai/docs/tagging.md#in-spacelab). ### Text Input Node: Text Input (`text-input`) - Inputs: none - Outputs: `text` (Text, text) Plain text with no variables. It's how space visitors give you words: a product name, a headline, a colour. - Out: Text. - Form field: a text field. - Example: Text Input named Headline → a Prompt variable. ### Document Node: Document (`document-input`) - Inputs: none - Outputs: `document` (Document, document), `text` (Text, text), `images` (Images, image) A file for a model to read: a brief, a deck, a spec or a sheet. - Best for: letting a brief or deck steer the run, in place of typing it into a prompt. - Out: Document (into the Documents input of Any LLM or Prompt Enhancer, which read the whole file), Text (the file as text: a CSV becomes a table, each slide a section) and Images (the pictures inside it, 10 by default, icons skipped). - Settings: Max images, under advanced (up to 20). - Form field: a file upload for one PDF, DOCX, PPTX, Markdown, TXT or CSV. - Limits: 25 MB; a PDF up to 50 pages; a CSV up to 500 rows and 100 columns. Password-protected files are refused. - Example: a brief PDF → Any LLM that pulls out the headline and offer → a Prompt → Image Model. - Tip: Claude and Gemini read a PDF up to 14 MB as it is, charts and layout included; a bigger PDF, and every PDF for other models, goes as its text. A scanned PDF has no text, so keep it under 14 MB and read it with Claude or Gemini. ### URL Input Node: URL Input (`url-input`) - Inputs: none - Outputs: `text` (Text, text), `images` (Images, image) A web page read from its link: a product page, a brand's about page, a public Google Doc. - Best for: building from a page that already says what the product is, in place of copying its text and saving its photos. - Out: Text (the page's title and main text, menus and footers left out) and Images (the photos on the page, 10 by default; icons, logos under 256 px and repeats skipped). - Settings: Max images, under advanced (up to 20). - Form field: a link that starts with https://. - Limits: public pages only. A page up to 10 MB, a linked PDF up to 25 MB; text is cut at 100,000 characters. A page that needs a sign-in, or builds itself with JavaScript, can't be read: the node says so. - Example: a product page → Any LLM that writes the ad copy, and its Images → the reference images of an Image Model. - Tip: it reads the page as it is. It doesn't search or summarise; wire its Text into Any LLM for that. ## Images ### Image Input Node: Image Input (`image-input`) - Inputs: none - Outputs: `image` (Image, image) One image. Upload it, drop or paste it onto the node, pick one from your library or Unsplash, or paste a link to an image under Link. - Best for: the main subject, such as the product, the garment or the person. - Out: Image. - Form field: an image upload, with the same Upload, Unsplash and Link choices. - Limits: a link must be a public https link to a JPEG, PNG, WebP, GIF, AVIF or HEIC image up to 25 MB. It is copied to our storage when you add it, so runs never depend on the original site. Links that Ask AI or an MCP agent put on the node are copied the same way. - Example: Image Input named Product photo → Image Model's reference images. ### Reference Image Input Node: Reference Image Input (`reference-image-input`) - Inputs: none - Outputs: `image` (Image, image) Wires exactly like Image Input. Use it for supporting images so the canvas reads clearly: anyone opening it sees which image is the subject and which is a reference. - Best for: a style, a pose or a mood. - Out: Image. - Form field: an image upload, labelled Reference image unless you rename the node. ### Multi Image Input Node: Multi Image Input (`image-input-multi`) - Inputs: none - Outputs: `images` (Images, image) Several images on one node, sent down one wire as a set. Wire it into a port that takes several images, such as an Image Model's references. - Best for: several views of one subject. - Out: Images. - Form field: an upload that takes several images. - Limits: 10 images. - Example: front, side and back of a sneaker → Image Model, so the model sees the whole shoe. - Tip: for one image at a time from the set, put an Image Iterator or an Image Randomizer after it. ### Background Input, Characters and Poses Node: Background Input (`background-input`) - Inputs: none - Outputs: `images` (Images, image) Pickers for backgrounds, people and poses. Choose from your library or upload new ones. - Out: Images. - Form field: a picker, so the visitor chooses from the same kind of list. - Limits: 10 picks per node. - Tip: add an Image Randomizer after one to use a different pick on each run. ### Image Pool Node: Image Pool (`image-pool`) - Inputs: none - Outputs: `image` (Image, image) A whole folder of images, with no limit on how many. Each run draws one at random. - Best for: variety from run to run without wiring anything else. - Out: Image. - Form field: none, so the pool stays yours. - Example: a folder of 200 backgrounds → Image Model. ### Pinterest Scraper Node: Pinterest Scraper (`pinterest-scraper`) - Inputs: `query` (Query, text) - Outputs: `images` (Images, image) Searches Pinterest for a keyword and returns the pins it finds. - Best for: mood and style references. The pins are other people's images, so don't ship them as your own. - In: Query. Wire a Text Input or Prompt into it. - Out: Images. - Settings: Limit, 1 to 100 (25 by default). - Form field: none. ## Video and audio ### Video Input Node: Video Input (`video-input`) - Inputs: none - Outputs: `video` (Video, video), `duration_seconds` (Duration (s), text) One video clip. - Out: Video. - Form field: a video upload that takes one clip of up to 20 seconds. - Limits: MP4, MOV, WebM, AVI, MKV, WMV, MPEG or 3GP, up to 500 MB on the canvas. - Example: Video Input → Video edit, or → Video Analyzer to describe what happens in it. ### Audio Input Node: Audio Input (`audio-input`) - Inputs: none - Outputs: `audio` (Audio, audio) One audio file. - Out: Audio. - Form field: an audio upload that takes one file of up to 50 MB. - Limits: MP3, WAV, OGG, M4A, AAC, FLAC, AIFF or WebM, up to 200 MB. - Example: your brand jingle → Merge audio + video. ## Settings ### Aspect ratio Node: Aspect ratio (`aspect-ratio-input`) - Inputs: none - Outputs: `value` (Aspect ratio, text) Pick a ratio and wire it into the Aspect ratio port of Image Model or a video node. One control then sets every node it feeds. - Out: the ratio, such as 9:16. - Settings: the ratio, 3:4 by default. - Form field: a dropdown. ### Number Node: Number (`number-input`) - Inputs: none - Outputs: `value` (Number, text) A number, with an optional min and max. Wire it into a Prompt variable. - Out: Number. - Settings: the number, Min and Max. - Form field: a number field that keeps to your min and max. - Example: Number named Age → A portrait of a {{age}}-year-old runner. ### Select Node: Select (`select-input`) - Inputs: none - Outputs: `value` (Value, text), `prompt` (Prompt, text), `values` (Values, text), `prompts` (Prompts, text) Your own list of choices. Add each with Add option in the node's settings panel, and the node sends on the one picked. - Out: the chosen value. - Form field: a dropdown, so visitors can only pick what you allowed. - Example: Select named Season with Spring, Summer, Autumn, Winter → a Prompt variable. ## Frequently asked questions ### Can a Sloosh space read a PDF brief? Yes. Add a Document node in Spacelab and wire it into Any LLM or Prompt Enhancer, which read the whole file. Claude and Gemini read a PDF up to 14 MB as it is, charts and layout included; a bigger PDF, and every PDF for other models, goes as its text. ### What is the difference between Image Input and Reference Image Input in Spacelab? Image Input and Reference Image Input wire the same way in Spacelab. Use Image Input for the main subject, such as the product, and Reference Image Input for a supporting style, pose or mood, so anyone opening the canvas sees which image is the subject. ### How do I use a different background on every Spacelab run? Use an Image Pool node in Spacelab: it holds a whole folder of images with no limit and draws one at random each run. Or put an Image Randomizer after a Background Input, which picks a different background from your picks on each run. ### Can I add an image to Spacelab from a link? Yes. A Spacelab Image Input takes a public https link to a JPEG, PNG, WebP, GIF, AVIF or HEIC image up to 25 MB, under Link. Sloosh copies the image to its own storage when you add it, so runs never depend on the original site. ### How do I give Sloosh space visitors a fixed list of choices? Add a Select node in Spacelab and add each choice with Add option. When published, Select becomes a dropdown, so visitors can only pick what you allowed. Wire it into a Prompt variable, for example a Season select with Spring, Summer, Autumn and Winter.