SSyncDev

SkyReels-V3-V2V-14B

SyncDev model directory · Profile updated Nov 19, 2026

Compare with another model

SkyReels-V3-V2V-14B is Skywork's 14B video-extension model that continues a clip for 5-30 seconds or cuts to a new shot, at 720p from a text instruction.

Key specifications

Specification score
13.4
License
Skywork License

The specification score summarizes recorded features and limits. It is not a benchmark result.

Pricing conditions: Open weights under the Skywork License; no first-party price listed.

Inputs and outputs

Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.

Accepts

  • Text input
  • Video input
  • Image input

Produces

  • Video output

Capabilities

VisionOpen weights

What is SkyReels-V3-V2V-14B?

You have eight seconds of footage and you need twenty. SkyReels-V3-V2V-14B extends it. Skywork's video-extension model takes an existing clip plus a text instruction and continues the scene, either as one unbroken shot or by cutting to a new angle. It shipped with the rest of SkyReels V3 on 29 January 2026.

Two ways to extend

Single-shot continuation. The camera keeps rolling. Output length is adjustable from 5 to 30 seconds, and the model is meant to preserve motion, scene layout and subject identity throughout.

Shot switching. The model cuts, using one of five cinematic transitions: Cut-In, Cut-Out, Shot/Reverse Shot, Multi-Angle and Cut Away. These extensions are limited to 5 seconds. Skywork suggests using a language model to write precise prompts that name the transition you want.

Both modes output 720p and support 1:1, 3:4, 4:3, 16:9 and 9:16 framing.

Command-line essentials

Flag Purpose
--task_type single_shot_extension continuous extension (5-30 s)
--task_type shot_switching_extension cut to a new shot (max 5 s)
--input_video source clip, path or URL
--duration seconds to add
--offload CPU offload
--low_vram FP8 weight-only quantisation and block offload for GPUs under 24 GB
--use_usp multi-GPU via xDiT USP

What Skywork says is new

The extension engine uses unified multi-segment positional encoding and hierarchical hybrid training to predict motion across segments, and a history-enhancement mechanism that Skywork says stretches coherent generation to minute-level durations. It is pitched for short-series production, live commerce, product demos and game cinematics. Skywork claims state-of-the-art single-shot and multi-shot extension, but unlike the reference-to-video model it does not publish a comparison table for this task, so there are no numbers to check.

Practical caveats

  • Chaining several 30-second extensions to reach minutes will accumulate drift. Check identity and lighting at every join.
  • Shot switching is short by design; build longer cut sequences step by step.
  • Licence is the Skywork License; read it before commercial distribution.

A quick way to judge it

Take three clips with different motion (a static interview, a walking subject, a fast pan) and extend each by 15 seconds, then cut each to a reverse shot. Look for warping hands, background changes and colour shifts at the boundary. Its family members are SkyReels-V3-R2V-14B, which builds clips from reference images, and SkyReels-V3-A2V-19B for audio-driven avatars. Skywork's world model Matrix-Game 4.0 is a different kind of video tool. More options are listed under video-to-video.

How to evaluate SkyReels-V3-V2V-14B

Compare the model on the work you need to complete.

Tasks and inputs

SkyReels-V3-V2V-14B is listed for Text-to-video, Video-to-video. It accepts text, video, image and produces video. Use the publisher’s input format and preprocessing for your chosen deployment.

Capacity and cost

Choose duration, resolution and frame rate together. Compare the cost of an accepted clip, including repeat generations.

Choosing a deployment

The weights are downloadable under Skywork License. Check runtime support and size the complete model and working context for your expected concurrency.

Frequently asked questions

5 to 30 seconds for single-shot continuation; shot-switching extensions are limited to 5 seconds.

These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.

NVIDIA

Cosmos3-Nano-Sim-Auto

Cosmos3-Nano-Sim-Auto is NVIDIA's open causal driving simulator: real-time multi-camera video and LiDAR that react to an AV policy's planned path.

— context

NVIDIA

Cosmos3-Nano-Sim-Transfer

Cosmos3-Nano-Sim-Transfer is NVIDIA's open 16B model that renders simulator depth, segmentation or edge maps into photorealistic robot video in real time.

— context

NVIDIA

Cosmos3-Nano-Transfer-Auto

Cosmos3-Nano-Transfer-Auto is NVIDIA's open 16B model that generates synchronised multi-camera driving video and LiDAR from map controls and a prompt.

— context

skywork

Matrix-Game 4.0

Matrix-Game 4.0 is Skywork's open set of 5B real-time video renderers that turn a running game's frames into clean styled video, per scene.

— context

Tencent

WeVisCut

WeVisCut is Tencent's 9B tool-calling agent that edits narration scripts into multi-minute videos through 12 editing tools and an editable Studio timeline.

— context

AInVFX

AInVFX/SeedVR2_comfyUI

AInVFX/SeedVR2_comfyUI is a source-linked Hugging Face repository assigned to Video-to-video. Its recorded task interface accepts video and produces video.

— context