Skip to content
BytePlusBytePlusBytePlus ModelArk for ComfyUI

Seed Prompt Writer

Template
Seed Prompt Writer, in the ComfyUI template browser under BytePlus
Workflow file
Seed Prompt Writer.json
BytePlus nodes
BytePlus LLM, Seedream 4.5 & 5.0
Keys needed
ModelArk API key
Also uses
Preview as Text, Save Image (built into ComfyUI)

What it makes

Section titled “What it makes”

One image, and the prompt that produced it. You type a few words, here “a cozy bookshop on a rainy evening”. The BytePlus LLM node expands them into one long prompt covering subject, setting, composition, lighting, color palette, lens and mood. That prompt is shown as text and also sent to Seedream, which draws the image.

The workflow

Section titled “The workflow”
The Seed Prompt Writer template: the BytePlus LLM node feeds a text preview and the prompt input of Seedream, which feeds Save Image
The LLM's text output goes to two places: a preview and the Seedream prompt.
  1. BytePlus LLM runs Seed 2.1 Turbo. Its prompt is the short idea. Its system_prompt tells it to write prompts for an image generator and to reply with the prompt only, in one paragraph, with no preamble.
  2. Prompt written by Seed (Preview as Text) shows what the LLM wrote.
  3. Seedream 4.5 & 5.0 receives the same text as its prompt and makes the picture with seedream 5.0 pro at (1.5K) 1536x1536 (1:1).
  4. Save Image keeps the result.

Run it

Section titled “Run it”
  1. Add your ModelArk API key in Settings → BytePlus if you have not yet (API keys). One key covers both nodes.
  2. Open the template browser and choose Seed Prompt Writer under BytePlus.
  3. Change the idea in the LLM node’s prompt box, or keep the bookshop.
  4. Press Run. The written prompt appears first in the preview and the picture follows in Save Image.

Settings worth changing

Section titled “Settings worth changing”
  • prompt (LLM): your short idea. A few words are enough.
  • system_prompt: change the style of the writing, for example ask for a shorter prompt or a particular art style.
  • size_preset (Seedream): pick another ratio or resolution.
  • model (LLM): the node also lists the other Seed models, DeepSeek V4.1 Flash and GLM 5.3 Flash.
  • reasoning_effort (LLM): the template uses medium. max thinks longest and costs the most, according to the tooltip.

Example result

Section titled “Example result”
The long bookshop prompt written by Seed and the finished picture of a cozy bookshop with an armchair, shelves and rain outside
Example result: the written prompt and the image it produced.

Time and cost

Section titled “Time and cost”

Each node bills once per run: one LLM request and one Seedream image. The seed controls are set to randomize, so every run writes a new prompt, draws a new picture and bills again. Set a control to fixed and ComfyUI reuses the cached result of an unchanged node without billing.

Section titled “Related”