--- title: "SeedVR2 AI Video Upscaler | Made" description: "Upscale and restore video 2x or 4x with ByteDance SeedVR2 in Made. See original before/after examples, per-megapixel pricing, and limits. On Made: 2.5 credits / video." url: https://www.made.app/models/seedvr-video-upscale --- [Models](https://www.made.app/models) / SeedVR2 Video Upscale ByteDance · video model # SeedVR2 Video Upscale Upscale video 2.5 credits / video · every plan · [price breakdown](https://www.made.app/price-sheet/seedvr-video-upscale) · [API playground](https://www.made.app/playground/seedvr-video-upscale) [Upload a video](https://www.made.app/tools/ai-video-generator?model=seedvr-video-upscale) Add the video to upscale and pick 2× or 4× — no prompt needed. Made with SeedVR2 Video Upscale · Product clip at 2× SeedVR2 is ByteDance’s video restoration model, upscaling any clip 2× or 4× with temporal consistency — frames stay coherent instead of flickering — from a single video input, no prompt required. Original results ## SeedVR2 Video Upscale examples made in Made These examples use media produced for Made. Prompts are included when the production record allows them. ### Product clip at 2× ### The 360p source clip ### Creator clip at 4× ### The 320p source clip Input ![Frame from the 360p source clip](https://www.made.app/images/models/seedvr-video-upscale/product-frame-before.jpg) Output ![The same frame after SeedVR2 upscaled the clip 2× in Made](https://www.made.app/images/models/seedvr-video-upscale/product-frame-after.jpg) ### Same frame at 2× Input ![Frame from the 320p source clip](https://www.made.app/images/models/seedvr-video-upscale/creator-frame-before.jpg) Output ![The same frame after SeedVR2 upscaled the clip 4× in Made](https://www.made.app/images/models/seedvr-video-upscale/creator-frame-after.jpg) ### Same frame at 4× ## Where SeedVR2 Video Upscale fits best ### Best uses - Upscaling AI-generated clips to delivery resolution - Restoring low-resolution or compressed footage - Temporally consistent detail without frame flicker - Preparing 360p–540p sources for 1080p masters ### Limits to know - It is billed by output data — 0.01 credits per megapixel of output (width × height × frames) — so long or high-resolution outputs cost more; Made quotes the exact price before generation. - It restores what the footage implies; it cannot re-frame, extend, or edit the clip’s content. ## Specifications in Made These are the controls exposed by Made, which may be narrower than a provider’s full API. **Aspect ratios** Matches the input video **Resolution** 2× or 4× the input, billed per megapixel of output **Duration** Matches the input video **Output formats** MP4 **Reference images** 1 **Processing** Scales with clip length and output size; progress appears in the generation view. ## How to use SeedVR2 Video Upscale in Made 1. **Add the video** Upload the clip to upscale — there is no prompt to write. 2. **Pick the factor** Choose 2× for finished clips or 4× for small, degraded sources. 3. **Check the quote** Cost follows output megapixels — resolution times frame count — and Made shows it before you run. 4. **Compare frames** Scrub the output against the source; detail should hold steadily across frames, not shimmer. Capabilities reviewed 2026-09-03 · Provider: [ByteDance](https://fal.ai/models/fal-ai/seedvr/upscale/video) Made shows the price in credits before generation. [View Made pricing](https://www.made.app/pricing) ## Go deeper on SeedVR2 Video Upscale ### [SeedVR2 Video Upscale pricing](https://www.made.app/price-sheet/seedvr-video-upscale) 2.5 credits / video — the full breakdown, verified. ## Compare relevant models image ### [Topaz Image Upscale](https://www.made.app/models/topaz-upscale-image) Topaz Image Upscale runs Topaz Labs’ precision upscaling models — the professional standard for enlargement — recovering real detail at 2× or 4× from a single input image, no prompt required. video ### [Kling LipSync](https://www.made.app/models/kling-lipsync) Kling LipSync takes a video of a person and a new audio track and regenerates realistic lip movement to match — swap the script, the language, or the voice without reshooting a frame. video ### [WAN 3.0](https://www.made.app/models/wan-3-0) WAN 3.0 is Alibaba’s WAN 3.0 video model at its standard speed and price, generating 2 to 30 second clips at up to 1080p with audio included. WAN 3.0 Prime runs the same inputs faster for about 40% more per second. ## Continue with a related tool ### [AI Video Generator](https://www.made.app/ai-video-generator) Turn any image into a captivating video. Upload a photo, describe the motion you want, and watch AI bring it to life with realistic movement, cinematic camera work, and professional quality. ### [Image Upscaler](https://www.made.app/image-upscaler) Enlarge photos 2× or 4× with Topaz and get the detail back, or upscale video with SeedVR2 or FLUX 3. Drop a file to see the exact price before you pay. ### [Image to Video Generator](https://www.made.app/image-to-video-generator) Upload a still photo, describe the motion or pick a suggestion, and MiniMax H3 Max animates it into a five-second MP4 clip with camera movement and subject motion. Questions, answered ## SeedVR2 Video Upscale FAQ Need a specific answer? [Contact support](mailto:support@made.app) ### What is SeedVR2? It is ByteDance’s diffusion-based video restoration model, upscaling clips 2× or 4× while keeping detail consistent from frame to frame. ### How is it priced? By output data: 0.01 credits per megapixel, counted as width × height × frames of the output video. Upscaling a clip to 1920×1080 across 121 frames costs 3 credits (each job rounds up to a whole credit), and Made quotes the exact figure before generation. ### Does it need a prompt? No. The video is the only required input — just pick the upscale factor. ### What makes it different from upscaling frame by frame? Per-frame upscalers make each frame sharp but inconsistent, which reads as flicker. SeedVR2 is temporally aware, so recovered detail stays stable across the clip. ### How do I keep the cost down? Trim the clip first and pick the smallest factor that reaches your target resolution — billing scales linearly with frames and quadratically with the factor. ### When should I choose another model? Choose Topaz Image Upscale for stills, or regenerate with a video model like WAN 3.0 when the clip needs new content rather than more resolution. --- Markdown version of https://www.made.app/models/seedvr-video-upscale. Every public Made page has one at its URL plus .md (https://www.made.app/models/seedvr-video-upscale.md). Site summary for AI assistants: https://www.made.app/llms.txt