Coming soon on 8frame. Placing documents on the board and wiring them into text nodes (LLM nodes) that read them is not released yet. This page covers why briefs get lost between the document and the prompt, how teams bridge that gap today, what you can do on the 8frame canvas now, and how text nodes will work when they ship. No date is announced.
Every campaign starts with words: a brief, a research deck, a client's notes, a shot list in a spreadsheet. Image and video models start with a prompt of a few sentences. The step in between, reading thirty pages and deciding what the first six frames should show, is where most of the thinking happens, and it is usually done by one person in a private chat window.
If you are writing the brief itself, see what a creative brief is and how to brief an AI video generation.
How teams bridge brief and prompt today
The common pattern is to paste or upload the brief into a general chat assistant, ask for shot ideas, and copy the results into the image tool. Chat assistants are good at the reading part. Anthropic's documentation, for example, says Claude can answer questions about "text, pictures, charts, and tables in PDFs" you provide.
The weak points are around the reading, not in it:
- The summary lives in one person's chat. The team sees the prompts, not the reasoning.
- Copy and paste breaks the chain. Nobody can tell which page of the brief a frame came from.
- Long documents get trimmed quietly. When a chat runs out of room, you do not always know what it left out.
- The whole brief ends up in the image prompt. Pasting long text into an image model's prompt rarely helps; image prompts work best short and visual.
What works on 8frame today
The canvas does not take documents yet: uploads today are images (PNG, JPG/JPEG, WebP, GIF, AVIF, BMP, up to 50 MB), video (MP4, WebM, MOV) and audio (MP3, WAV). A workable routine:
- Pull the brief's key lines into a comment on the board. Audience, the one message, mandatory elements, formats. Everyone on the board then works from the same short version.
- Write one prompt per frame, short and visual. Subject, setting, light, framing. Keep the brief's words out of the prompt unless they describe something visible.
- Wire reference images in. The Nano Banana node takes up to 10 references at no extra charge. A Nano Banana 2 image is 11 credits at 1K, Nano Banana Pro 21 at 1K or 2K.
- Storyboard a shot list in one go. ShotGrid (beta) makes a 3x3 contact sheet, nine frames in one 4K image, for 42 credits, from a prompt of up to 5,000 characters and 1 to 10 reference images.
- If your brief already lives in an AI assistant, the 8frame MCP connector at mcp.8frame.co lets that assistant start generations on 8frame. Results come back as file links, not onto a board.
What is coming: text nodes that read documents
Text nodes will put the reading step on the board, where the team can see it.
Documents on the board
You will be able to place a document on the board, such as a PDF brief or a CSV shot list, which will read as a table. A document will connect to text nodes and agents only. If you try to wire it straight into an image or video generator, the board will tell you to connect a text node first. The text a text node writes, however, can feed a prompt.
The text node
A text node will appear in the «+» catalog under Text. Its field is the task, in plain words: "Summarise these three briefs into one page for the director". Standing instructions go in its settings and travel with the node into every run.
- Several documents into one node, in the order you wire them, since order matters to the model.
- Pages per connection. You will choose which pages a node reads on the wire itself, so the same document can feed two nodes with different pages.
- Limits you can see. The node will show how much is connected, in pages and words. If it is more than the model can read, the node will say so and offer two ways on: choose pages, or summarise first (a quick text node in front of the biggest document). Nothing is planned to be cut silently.
- The price before you run. Launch will show the run's credit price before you press it. We are not publishing text-node prices before launch.
- Streaming output. The answer will appear in the node's output as it is written.
Turning text into work
Two actions will connect the reading to the making:
- Make prompt nodes: a numbered list in the output (say, six shot ideas) becomes one prompt node per item, ready to wire into image or video generators.
- Make a text object: copies the answer onto the board as editable text.
And on any document, Ask will let you type a question; it will create a text node wired to that document, with your question as the task, and run it.
Agents will read documents through the same mechanism. See AI agents as teammates on a creative canvas, and for brand rules, brand kits for AI image and video generation.
Three workflows that text nodes are designed for
| Start | Text node task | Then |
|---|---|---|
| Three research PDFs | "Summarise into one brief" | An editable text object the team comments on |
| One brief | "Write six shot prompts" | Make prompt nodes, wire each into an image generator |
| A CSV shot list | "One ShotGrid prompt per row" | Wire the prompts into ShotGrid for contact sheets |
Each of these can be run by hand today, with the reading done in a chat assistant and the prompts typed into nodes. The difference coming is that the document, the reading and the prompts will sit on one board.
FAQ
Can I upload a PDF to an 8frame board today?
Not yet. The canvas accepts images, video and audio today. Documents on the board are coming soon with text nodes.
Will the whole document go into my image prompt?
No. Documents are designed to connect to text nodes and agents only. A text node's output can then feed a generator's prompt.
What happens if a document is too long?
The text node will show how many pages and words are connected, and if that is over its limit, offer to choose pages or summarise first. It is not designed to cut text without telling you.
What will text nodes cost?
Each run's credit price will be shown on Launch before you run it. Prices will be published at launch, not before.
Can I storyboard a shot list today?
Yes. ShotGrid (beta) makes nine frames in one 4K contact sheet for 42 credits, from a prompt of up to 5,000 characters.
Sources
- Anthropic, PDF support ("text, pictures, charts, and tables in PDFs"): platform.claude.com/docs/en/build-with-claude/pdf-support, checked 2026-10-21
- 8frame prices, defaults and limits: the 8frame app, checked 2026-10-21
- Upcoming 8frame features: the 8frame product roadmap, 2026-10-21
Try it without signing up: the 8frame canvas opens on a real board. The canvas is free and unlimited; generation is paid from $19/month.