Video Enhancement
Super Resolution API — AI upscaling that recovers detail
Deep-learning upscaling trained to reconstruct real detail — not just sharpen edges. No duration caps, native batch support, and artifact-safe defaults.
Before & after
See the difference
Drag the handle to compare the low-resolution source with the AI-upscaled 4K output.
What it does
Upscaling is not sharpening. A deep network learns where texture belongs and reconstructs faces, foliage and fabric instead of painting over them. No duration caps means hours-long content is a single job, and native batch support queues thousands of files with per-file callbacks and retries. Denoise strength and detail gain are per-job, with anti-ringing and anti-halo defaults.
Full API referenceimport requests
url = "https://api.mlslabs.io/v1/super-resolution/jobs"
resp = requests.post(
url,
headers={"X-API-Key": "YOUR_API_KEY"},
json={
"input": "s3://bucket/archive-sd.mov",
"output": "s3://bucket/upscaled/",
"scale": 4,
"target_resolution": "2160p",
"denoise": 0.4,
"detail": 0.6,
},
)
job = resp.json()
print(job["job_id"])Key features
Why teams pick this API
Detail-recovery architecture
A deep CNN learns where texture belongs, reconstructing faces, foliage and fabric instead of painting over them.
No duration caps
Hours-long content, one job. Built for libraries, not just clips.
Native batch
Queue thousands of files with per-file callbacks and retries.
Artifact-safe controls
Denoise strength and detail gain per job, with anti-ringing and anti-halo defaults.
Use cases
Built for real workflows
Upscale legacy content
Bring SD and HD libraries to 1080p/4K for modern platforms.
Archive restoration
Recover detail from degraded footage with controlled denoise.
VOD catalog refresh
Batch-upscale entire catalogs in one queue instead of per-title renders.
Pricing
Credit-based pricing
Every request draws from your credit pool. This API bills 35 credits per second of output video.
| Billing unit | Cost in credits | Notes |
|---|---|---|
| Processing time | 35 credits | Per second of output video |
| Failed jobs | 0 credits | Credits auto-refunded |
| Batch processing | Same rate | One queue, per-file callbacks |
Learn
Guides, tutorials & comparisons
Batch Upscale Videos
Batch upscale a video library to 4K: folder-prefix jobs, per-file status, and a plan that stages the...
Read TutorialReal-ESRGAN API Alternative: Migrate Your Upscaling Pipeline
Moving from self-hosted Real-ESRGAN to an API? A migration table for VRAM, tiling, FFmpeg stitching ...
Read TutorialRemove JPEG Artifacts and Upscale
Remove JPEG blocking artifacts and upscale in one pass: clean compressed video before upscaling so t...
Read TutorialUpscale 480p to 4K
Upscale 480p to 4K: AI super resolution rebuilds real detail in old DVD and VHS-era footage, with de...
Read TutorialUpscale Anime Video
Upscale anime video with anime-tuned super resolution: clean line art, no ringing artifacts, and fai...
Read TutorialUpscale Old Videos to 4K
Upscale old home videos and archive footage to 4K: handle grain, motion blur and low light, then res...
ReadDoes AI Upscaling Really Work?
Does AI upscaling actually add detail or just smooth pixels? We explain what super-resolution models...
Read GuideGAN vs Diffusion for Super-Resolution
GAN-based and diffusion-based super-resolution compared: sharpness vs naturalness, artifacts (GAN no...
Read GuideReal-ESRGAN API vs Self-Hosted
Real-ESRGAN is free and excellent — on a GPU you own. Compare the true cost of self-hosting vs a sup...
Read GuideReal-ESRGAN vs Topaz Video AI
Real-ESRGAN (free, open source) vs Topaz Video AI (commercial) for video upscaling: quality, ease of...
Read GuideSuper Resolution vs Upscaling: What's Actually Different?
Upscaling is interpolation; super resolution is semantic reconstruction. Learn the difference, when ...
ReadFAQ
Common questions
How is this different from simple sharpening?
Sharpening amplifies existing edges; we reconstruct missing texture. The difference shows in faces, foliage and gradients — see the concept page.
What is the maximum output resolution?
4K today; higher on request for approved workflows.
How fast is batch processing?
Batch jobs run in parallel across a shared queue with per-file callbacks.
Can I control artifacts on noisy source?
Yes — denoise and detail gain are per-job parameters, with safe defaults.
Related APIs