Skip to content
BytePlusBytePlusBytePlus ModelArk for ComfyUI

BytePlus LLM

Node name
BytePlus LLM
Node ID
BytePlusSeed
Find it
Double-click the canvas and search for “BytePlus LLM”, or browse BytePlus ModelArk
Keys needed
ModelArk API key
Runs when
Its output is used by another node, such as Save Image or Save Video. It is not an output node.

What it does

Section titled “What it does”

Sends a prompt to a language model on BytePlus ModelArk and returns the answer as text. Pick Seed 2.0 Pro, Lite or Mini, Seed 2.1 Turbo, DeepSeek V4.1 Flash or GLM 5.3 Flash, and add images and videos as context (Seed 2.0 Lite and Mini also take audio clips).

Use it to describe pictures, write or rewrite prompts for other nodes, or answer questions about a video. The node bills for the tokens it uses and runs only when another node uses one of its outputs.

On the canvas

Section titled “On the canvas”
The BytePlus LLM node with the model set to Seed 2.0 Pro, one image slot, one video slot, a prompt box and the advanced inputs such as reasoning_mode, turns and stream
Seed 2.0 Pro selected, so the node has no audio slots. Seed 2.0 Lite and Mini add them.

Inputs

Section titled “Inputs”
InputTypeDefaultWhat it does
prompttextText input to the model.
seedinteger0Seed controls whether the node should re-run; results are non-deterministic regardless of seed.Range: 0–2147483647

model is required and decides which of the inputs below appear. The model used to generate the response.

InputTypeDefaultWhat it does
imagesgroup of inputs of IMAGEOptional image(s) to use as context for the model. Up to 20 images.Slots: image_1 to image_20 (20)
videosgroup of inputs of VIDEOOptional video(s) to use as context for the model. Up to 4 videos.Slots: video_1 to video_4 (4)
Advanced inputs for Seed 2.0 Pro
InputTypeDefaultWhat it does
temperaturenumber1Controls randomness. 0.0 is deterministic, higher values are more random.Range: 0–2
Advanced inputs (8)
InputTypeDefaultWhat it does
system_promptoptionaltextFoundational instructions that dictate the model's behavior.
detailoptionalchoicehighImage detail level sent with each image.Choices: low, high
fpsoptionalnumber1Frames per second sampled from each video.Range: 0.2–5
reasoning_modeoptionalchoiceautoDeep thinking (thinking.type). auto sends nothing: the model default (enabled) applies.Choices: auto, enabled, disabled
reasoning_effortoptionalchoicemediumChain-of-thought length (reasoning.effort). max thinks the longest (and costs the most); models that do not tell two levels apart treat them alike. Not sent when reasoning_mode is disabled.Choices: minimal, low, medium, high, max
turnsoptionalinteger11: every run starts a new conversation (the response is not stored). 2 or more: the length of a conversation. Each run continues this node's last conversation (with the same model) with the new prompt until it has this many turns; the next run starts a new one. Images, videos and audio from the first turn stay in the conversation.Range: 1–10
streamoptionalon/offoffStream the answer to the console while it is generated.
file_expire_secondsoptionalinteger604800How long uploaded images, videos and audio are kept in Ark Files (1 to 30 days).Range: 86400–2592000

Notes on inputs

Section titled “Notes on inputs”
  • Media slots: image_1, video_1 and so on appear as you connect them, up to 20 images and 4 videos.
  • Audio: only Seed 2.0 Lite and Mini show audio_1 to audio_4. The other models cannot hear audio, so the node does not offer the input for them.
  • Thinking: reasoning_mode set to auto sends nothing, and the model uses its own default. reasoning_effort goes from minimal to max; max thinks the longest and costs the most. It is not sent when reasoning_mode is disabled.
  • Several turns: with turns at 1, every run starts a new conversation and nothing is stored. With 2 or more, turns is the length of a conversation: each run continues this node’s last conversation with the new prompt until it has that many turns, and the next run starts a new one. Only the first turn sends media, which then stays in the conversation: media you connect later is not sent. Changing the model, the key or the region starts a new conversation.
  • System prompt: system_prompt is sent again on every turn.
  • Streaming: turn on stream to watch the answer appear in the console. The STRING output still arrives complete at the end.

Outputs

Section titled “Outputs”
OutputTypeWhat it holds
STRINGtext
raw_jsontext

Notes on outputs

Section titled “Notes on outputs”
  • STRING is the answer text.
  • raw_json is the full response from ModelArk, as JSON. Use it to debug a result or to read fields the text leaves out.

Limits and tips

Section titled “Limits and tips”
  • Up to 20 images and 4 videos per request. Every image of a connected batch counts toward the 20.
  • Audio is limited to 4 clips and 120 minutes in total, and only on Seed 2.0 Lite and Mini.
  • Images, videos and audio are uploaded to Ark Files and kept for file_expire_seconds (1 to 30 days). Uploads are reused when the content has not changed. This node needs no Comfy.org login.
  • fps sets how many frames per second the model samples from each video.
  • seed only decides whether ComfyUI runs the node again. The answers are not deterministic, even with the same seed.
  • Activate each model in the ModelArk console for the region you use.
  • Interrupting a run stops the node’s wait. The node sends no cancel request, so a request that was already sent may still be billed.

Common errors

Section titled “Common errors”
Message What to do
[BytePlus] No BytePlus API key is set. Add it in Settings > BytePlus … Add the key in Settings → BytePlus. See API keys.
[BytePlus] prompt is empty. Write a prompt.
[BytePlus] Up to 4 videos are supported per request; … are connected. Disconnect videos until four or fewer remain. The same limit applies to images (20) and audio clips (4).
[BytePlus] The audio clips total … minutes; ModelArk accepts at most 120 minutes of audio per request. Shorten or remove audio clips.
[BytePlus] The model refused to respond: … Rephrase the prompt or change the media. The message includes the reason the model gave.

Used in templates

Section titled “Used in templates”