Script and reference documents
Use these tools to read and change the words and pictures that sit beside your scenes.
Reference documents
Section titled “Reference documents”A reference document holds a title, markdown notes and a small image bank. reference_document_ids on generate_image, refine_image and animate_scene names the documents a generation must match. A scene prompt that writes a document’s exact title also attaches that document.
A reference for one subject (a character, an object, a place or a style) is a sheet document. create_document takes sheet with the subject’s kind and name. The document’s ## Consistency and ## Avoid sections say what must hold and what to avoid. Every picture later added to the document joins the sheet as a named view.
A document attaches to a generation as its sheets, one picture each and three at most. A document without a sheet attaches its pictures as one grid picture. Its text goes whole when short and as a brief when long.
list_documents
Section titled “list_documents”List a project’s documents, each with the folder it sits in, plus the project’s folders.
On ChatGPT app v2, library: true lists your own library instead: the documents you keep outside any project, and the library’s folders.
- Surfaces: ChatGPT app v2, Manan
- Parameters:
project_id(required),library
get_document
Section titled “get_document”One document’s content and image bank.
It reads every document kind. A storyboard returns its panels and a breakdown its rows. item_ids narrows the read to some items. On a storyboard or whiteboard, contact_sheet adds labelled sheets of the pictures. On a whiteboard, group_id and include_captions narrow the read, and view: "snapshot" returns pictures of the board as it looks in Screel.
On ChatGPT app v2, revisions: true returns the document’s saved versions instead of its content. include_comments adds the open comment threads.
- Surfaces: ChatGPT app v2, Manan
- Parameters:
document_id(required),project_id(required),block_ids,contact_sheet,fields,group_id,include_captions,include_comments,item_ids,max_blocks,revisions,scene_ids,start_block,view
create_document
Section titled “create_document”Start a reference document with a title and markdown blocks, or a sheet document with sheet. folder_id puts it in a project folder.
kind: storyboard makes a storyboard instead, and from_scenes fills it with one panel per scene. In Manan, kind also makes a whiteboard or an empty workflow, and ops put the first items on a new whiteboard.
- Surfaces: ChatGPT app v2, Manan
- Parameters:
project_id(required),title(required),blocks,folder_id,from_scenes,kind,sheet
edit_document
Section titled “edit_document”Replace a document’s blocks.
On ChatGPT app v2, a storyboard takes panels and a breakdown takes assets instead. Send the whole list from get_document with your change applied. A listed item keeps its id, and an item without an id is new. The write removes an item you leave out. restore_version brings back a saved version as a new version, so the history keeps every version. Whiteboards and workflows change only in Screel.
- Surfaces: ChatGPT app v2, Manan
- Parameters:
document_id(required),project_id(required),assets,base_version,blocks,panels,restore_version,title
set_document_image
Section titled “set_document_image”Add or replace a picture in a text document’s bank: a chat file attachment or the URL of a picture Screel already stores. It refuses a whiteboard, storyboard or breakdown.
- Surfaces: ChatGPT app v2, Manan
- Parameters:
document_id(required),project_id(required),image,image_id,image_url,name
get_playbook
Section titled “get_playbook”One of Screel’s playbooks for reference work, as text: consistent-characters, props-and-objects, references-from-uploads, scene-continuity or style-and-world; also workflows in Manan. All but workflows also ship as skills with ChatGPT app v2. It spends no credits.
- Surfaces: ChatGPT app v2, Manan
- Parameters:
topic(required)
open_project
Section titled “open_project”A link that opens the project in the Screel web app. Use it for work with no tool here, such as fine-tuning scene timing, sharing or previewing voices. It changes nothing and spends no credits. In Manan, it works in a conversation with no project open.
- Surfaces: ChatGPT app v2, Manan
- Parameters:
project_id(required),document_id,scene_id,scene_position,view
Storyboards and breakdowns
Section titled “Storyboards and breakdowns”A storyboard holds panels, each with a title, camera, action, dialogue, notes and a picture. A breakdown reads one picture and lists the assets in it as rows. get_document reads both, and edit_document changes them on ChatGPT app v2.
generate_panel_image
Section titled “generate_panel_image”Draw the pictures of up to 12 storyboard panels from each panel’s title, camera and action. Each picture lands on its panel when its job finishes. It spends credits.
- Surfaces: ChatGPT app v2, Manan
- Parameters:
document_id(required),panel_ids(required),project_id(required),instructions,request_id
breakdown_image
Section titled “breakdown_image”Read a picture into proposed breakdown rows for your review. The source is a scene’s still, an upload, a picture of the project or a picture in a document. Without document_id it creates the breakdown; with one it reads the picture again. It spends no credits.
- Surfaces: ChatGPT app v2, Manan
- Parameters:
project_id(required),document_id,instructions,source,title
generate_asset_references
Section titled “generate_asset_references”Draw one reference picture for each accepted asset row that has none, cut from the source around the asset’s box. asset_ids picks some rows. Proposed rows wait until you accept them. It spends credits.
- Surfaces: ChatGPT app v2, Manan
- Parameters:
document_id(required),project_id(required),asset_ids,request_id
Comments, library and exports
Section titled “Comments, library and exports”create_comment
Section titled “create_comment”Start a comment thread on a document, or on one of its blocks, pictures, rows, panels, items or nodes. It spends no credits.
- Surfaces: ChatGPT app v2, Manan
- Parameters:
body(required),document_id(required),project_id(required),anchor_id,anchor_kind
reply_comment
Section titled “reply_comment”Reply to a comment thread. On ChatGPT app v2, take the thread’s id from get_document with include_comments. It spends no credits.
- Surfaces: ChatGPT app v2, Manan
- Parameters:
body(required),comment_id(required),document_id(required),project_id(required)
resolve_comment
Section titled “resolve_comment”Mark a comment thread resolved. It spends no credits.
- Surfaces: ChatGPT app v2, Manan
- Parameters:
comment_id(required),document_id(required),project_id(required)
attach_document
Section titled “attach_document”Bring a library document into a project, so its tools can read it there. On ChatGPT app v2 it attaches only a document of your own library, from list_documents with library: true, and it does not detach one. In Manan, detach removes a library document from the project. It spends no credits.
- Surfaces: ChatGPT app v2, Manan
- Parameters:
document_id(required),project_id(required),detach(Manan only)
export_document
Section titled “export_document”Export a document as a file. A reference document exports as md, txt, docx or pdf. A storyboard exports as pdf, png, docx, csv or xlsx in its page setup. A breakdown exports as xlsx, and kind: breakdown without document_id puts every breakdown of the project in one workbook. A whiteboard exports as png or pdf. On ChatGPT app v2 it starts a job, and get_job_status returns the download link when it finishes. The download link stays valid for 24 hours. It spends no credits.
- Surfaces: ChatGPT app v2, Manan
- Parameters:
format(required),project_id(required),columns,document_id,gap,kind,layout,margin,orientation,paper,style
Script and reference documents on Generic MCP
Section titled “Script and reference documents on Generic MCP”These tools read and change the project’s script and reference documents. The reference document tools work on text documents only and refuse a whiteboard, storyboard, breakdown or workflow. None of them spends credits.
get_script
Section titled “get_script”Read the project’s script: each block with its text and its scene link.
- Surfaces: Generic MCP
- Parameters:
project_id(required),max_blocks,start_block
edit_script
Section titled “edit_script”Rewrite the script. The blocks you send replace the whole document.
- Surfaces: Generic MCP
- Parameters:
blocks(required),project_id(required),base_version
suggest_script_edits
Section titled “suggest_script_edits”Park a proposed script as a suggestion. Nothing changes until you accept it.
- Surfaces: Generic MCP
- Parameters:
blocks(required),project_id(required),title(required),base_version
list_context_bits
Section titled “list_context_bits”List the project’s reference documents.
- Surfaces: Generic MCP
- Parameters:
project_id(required)
get_context_bit
Section titled “get_context_bit”Read one reference document: its text and its image bank.
- Surfaces: Generic MCP
- Parameters:
bit_id(required),project_id(required),max_blocks,start_block
create_context_bit
Section titled “create_context_bit”Create a reference document.
- Surfaces: Generic MCP
- Parameters:
blocks(required),project_id(required),title(required)
update_context_bit
Section titled “update_context_bit”Rewrite a reference document. The blocks you send replace its whole body.
- Surfaces: Generic MCP
- Parameters:
bit_id(required),project_id(required),blocks,title
suggest_context_bit_edits
Section titled “suggest_context_bit_edits”Park a proposed rewrite of a reference document as a suggestion.
- Surfaces: Generic MCP
- Parameters:
bit_id(required),blocks(required),project_id(required),title(required),proposed_title
add_context_bit_image
Section titled “add_context_bit_image”Add a picture to a reference document’s image bank from a public https url.
- Surfaces: Generic MCP
- Parameters:
bit_id(required),image_url(required),project_id(required),name
update_context_bit_image
Section titled “update_context_bit_image”Replace a picture in a reference document’s image bank from a public https url.
- Surfaces: Generic MCP
- Parameters:
bit_id(required),image_id(required),image_url(required),project_id(required),name
view_scene_image
Section titled “view_scene_image”Return a scene’s still as a picture the client can see.
- Surfaces: Generic MCP
- Parameters:
project_id(required),scene_id(required)
view_image
Section titled “view_image”Return a stored picture as a picture the client can see.
On ChatGPT app v2, it shows one picture per call. With only project_id, it shows every scene’s still on one numbered sheet, and page picks a page of a long sheet. scene_ids or scene_positions narrow the sheet. scene_id or scene_position shows one scene’s still, and clip: true shows that scene’s clip as a sheet of frames. render_id shows a finished render the same way. image_url takes only a picture Screel stores. It spends no credits.
- Surfaces: ChatGPT app v2, Generic MCP, Manan
- Parameters:
image_url(required on Generic MCP),clip,page,project_id,render_id,scene_id,scene_ids,scene_position,scene_positions
