> ## Documentation Index > Fetch the complete documentation index at: https://docs.wji.ai/llms.txt > Use this file to discover all available pages before exploring further. # Generating media > Use the generate panel to turn prompts and connected references into images, video, and audio. Selecting a generation node opens the **generate panel** beneath it. ## The prompt Generate panel below a node with the prompt box, model selector, parameters button, reference list, cost estimate, and send button The prompt box tells you what it accepts: describe what you want, press `@` to reference an asset, press `/` for commands. * **`@` mentions** address your connected references by name — `@image1`, `@video1`, `@audio1` * **`/`** opens the prompt preset picker, with tabs for workspace, personal, and marketplace presets. Choosing one creates a text node at the pointer ## Improving the prompt The panel has the same two-tab arrangement as the standalone generator — your prompt, and the AI-optimized final prompt. You can regenerate the optimization or accept it as written. **Save as prompt preset** stores a prompt you want to keep. ## The reference list Everything wired into this node appears here. * Remove a reference without deleting the connection * **Drag to reorder** — order matters to some models * A counter shows how many of the model's allowance you are using, for example "3 / 5" * Long text references have a **view full text** control For video models with frame support, first and last frame slots appear, with a swap control and individual remove buttons. ## Settings and cost Generation parameters popover with aspect ratio, batch size, and model specific settings The generation parameters popover holds the same settings as the standalone generator, filtered to what the selected model supports. The estimated cost updates as you change things. Press the send button to submit. ## Working in bulk Review dialog listing each selected node with its prompt, references, model, and parameters before submitting This is where the canvas pays off. Select several generation nodes and the multi-select toolbar gives you: * **Model settings** — apply one model and its settings to every selected node of that type * **Improve prompt** — optimize all the prompts at once * **Submit** — opens a review dialog listing every node's prompt, references, model, and parameters, so you can check the batch before committing, then submits them together The review dialog before bulk submit is the last checkpoint you get — generations cannot be cancelled. Read it, especially the reference counts, before pressing submit on twenty nodes. ## After it runs Results attach to the node. The node's history picker lets you look through every result it has produced and promote a different one to be the node's main output.