Use NAVER's lightweight Korean-first vision-language model for text, images, video, visual questions, charts, diagrams, documents, and OCR-aware analysis. Use the playground to give HyperCLOVA X SEED Vision Instruct 3B a representative input and review its output. Adjust the available settings to shape the result for your task.
Model
naver-hyperclovax/HyperCLOVAX-SEED-Vision-Instruct-3B
Documentation
Use this reference to understand how every field in the playground changes the request sent to naver-hyperclovax/HyperCLOVAX-SEED-Vision-Instruct-3B.
Required
Type: JSON array
Provides a multimodal conversation as a JSON array whose content items may contain text, images, or video. Include media sources and their related instruction in the same user turn; optional OCR, transcript, and Lens metadata can ground the model with information extracted upstream.
Optional
Type: Integer
Default: 64
Range: 1–16384
Caps how many tokens the model can add after the input. The model may stop earlier naturally, but a limit that is too low can cut off an answer; input tokens are not counted in this value.
Optional
Type: Boolean
Default: On
Chooses whether the next token is sampled from a probability distribution or selected deterministically. Enable it for more varied responses that are affected by temperature and Top P; disable it when repeatability matters or when using beam search.
Optional
Type: Number
Default: 0.6
Range: 0–1
Limits sampling to the smallest set of likely tokens whose probabilities add up to this value. Lower settings make output more focused, while a value near 1 keeps more alternatives available; it only affects sampled generation.
Optional
Type: Number
Default: 0.5
Range: 0–2
Adjusts how strongly the model favors its most likely next token. Lower values produce more focused and repeatable output, while higher values allow less-likely choices and increase variety; it only has an effect when sampling is enabled.
Optional
Type: Number
Default: 1
Range: 0.01–2
Changes the scores of tokens that already appeared in the generated text. Values above 1 discourage repeated words and loops, 1 leaves probabilities unchanged, and values below 1 make repetition more likely.
Optional
Type: Integer
Default: None
Range: 1–256
Sets how many frames the processor samples from the source video. More frames preserve additional events but consume more visual tokens and memory; leave it empty to let adaptive sampling choose a count that fits the media and context.
Optional
Type: Number
Default: None
Range: 0.01–Unbounded
Sets the rate at which frames are sampled from the source video for model input. A higher rate preserves faster motion but creates more visual tokens and processing work; leave it automatic unless the task requires a known temporal sampling rate.
Optional
Type: String
Default: opencv
Values: OpenCV, Decord, PyAV, Torchvision
Selects the library used to decode and sample source video. OpenCV, Decord, PyAV, and Torchvision can differ in codec support, performance, and frame seeking, but should not change the intended model task.