Video Upscaling - Kandinsky 6 SR ×2 (Precision)

Video Upscaling - Kandinsky 6 SR ×2 (Precision)

🛠

Take 1080p Video to 4K — Faithful, Low-Flicker 2× Upscaling

Kandinsky 6 SR ×2 doubles a video's resolution, so a 1920×1080 clip becomes 3840×2160. It is a Precision upscaler from Kandinsky Lab: it stays faithful to the source, keeps flicker low and invents little texture. Diffusion-based, distilled to just 2 steps per tile. MIT license.

📹 See it in action

Breaking news, now in Full HD. Our evening anchor, a pigeon, filmed at 960×540 and upscaled ×2 to 1920×1080. Default settings, seed 7.

Side-by-side zoom of a pigeon news anchor's face: 540p input on the left, 1080p Kandinsky 6 SR ×2 output on the right

Same frame, same crop. Left: the 540p input, zoomed. Right: Kandinsky 6 SR ×2. Eyes, beak and neck feathers sharpen; the anchor keeps the same serious face.

The full 1080p output. The source clip was generated with LTX-2.5 in CNAPS Studio and downscaled to 540p before upscaling.

What it does

Kandinsky 6 SR ×2 is the video super-resolution model from the Kandinsky 6.0 Video family. Give it a clip up to 1920×1080 and it returns the same clip at exactly 2×, up to 4K.

It is the only upscaler in CNAPS Studio built for 1080p sources — the 4× models accept at most 960×540. If your source is already Full HD and you want 4K without the image being redrawn, this is the node.

For sources at or below 960×540, use its sibling Kandinsky 6 SR ×4 instead.

Problem it solves

Use case
How it helps
1080p to 4K delivery
Meet 4K delivery specs from Full HD masters
Brand and product video
Faithful output — packaging, text and faces stay as they were
Low flicker
Steady frame-to-frame detail, so textures don't shimmer
AI-generated video
Lift 1080p generated clips to 4K without changing the look

Input/Output

Input

  • Video (required) — up to 1920×1080 in either orientation, 9–121 frames (about 5 s at 24 fps)

A larger clip such as 2560×1440 or 3840×2160 is rejected before it is queued (VIDEO_RESOLUTION_EXCEEDS_LIMIT). Clips shorter than 9 or longer than 121 frames are rejected too. In a Batch Run, one clip outside these limits stops the whole node. The output has no audio — add the original sound back with Audio Replace / Mix.

Parameters

Parameter
Details
scale_factor
Fixed at 2× (read-only). 1920×1080 → 3840×2160
seed
-1 (default) picks a new random seed each run. Set a fixed number to reproduce an earlier result (advanced)

Output

  • Video — exactly 2× the input size, at most 3840×2160 (4K)

Speed

Slow and GPU-heavy: about 13 minutes at the maximum size and length (1080p, 121 frames). For a quick upscale, a preview or a longer clip, use FlashVSR-v1.1 or SeedVR2 3B.

Choosing a video upscaler

What you need
Use
A 1080p source taken to 4K, faithfully
Kandinsky 6 SR ×2 (Precision)
A 540p source taken to 4K, faithfully
Kandinsky 6 SR ×4 or RelayVSR (Precision)
The sharpest-looking 4K, extra texture is OK
FlashVSR+ Turbo (Creative)
A fast upscale or a preview
FlashVSR-v1.1, SeedVR2 3B

Technical Details

Pipeline
Distilled super-resolution diffusion transformer + KVAE video VAE + latent upscaler bank
Parameters
SR DiT 1.41B · KVAE 1.74B · latent upscaler bank 3.65B
Sampling
Distilled with π-Flow to 2 steps per tile (the teacher uses 4)
Tiling
Tiled diffusion in KVAE latent space
Scale
2× (fixed in this node)
Audio
Video only — source audio is not carried through
Runtime
GPU, cloud — no local setup
Developer
Kandinsky Lab

Compliance & Provenance

Provider
Kandinsky Lab (open weights)
Provider type
Specialized
License
MIT
EU AI Act risk class
Minimal Risk
Art. 50 transparency
Not applicable
Region availability
Available globally
Training data summary
Pending — provider has not yet published per Art. 53(d)

For more on how we classify models and mark outputs, see our AI Policy.

Model Source

How Kandinsky 6 SR ×2 compares

Model
Profile
Scale
Max input
Notes
Kandinsky 6 SR ×2
Precision
2×
1920×1080, 121 frames
Only option for 1080p sources; ~13 min at max size
Kandinsky 6 SR ×4
Precision
4×
960×540, 121 frames
Same model family, for 540p sources
RelayVSR
Precision
4×
960×540, 241 frames
~7 min per full clip
FlashVSR+ Turbo
Creative
4×
960×540, 241 frames
Sharpest; may add texture
FlashVSR-v1.1
—
up to 4×
—
Fast streaming upscale
SeedVR2 3B / 7B
—
up to 4×
—
Upscale + restoration

Scope notes come from each node's own documentation; the models have not been benchmarked against each other inside CNAPS Studio.

All of these run inside CNAPS Studio — compare them side-by-side with the Video Compare tool.

Run it now

Click to try Kandinsky 6 SR ×2 in CNAPS Studio