AI Upscaler — DynamicVibe Productions

UPSCALE.AI

Upscaling for AI-generated video and images, 2× or 4×. Images run on SeedVR2 and Topaz, video on Topaz Precision or Starlight, and frame rate goes to 48 or 60fps with Topaz Interpolate. It all runs from your browser through your own fal.ai key.

2× / 4× Upscale Creative Upscaling 60fps Frame Interpolation Pipeline Mode Finishing Presets Delivery Manifest Face Restore Before / After Slider Image Quick Presets Batch Queue Kling · Runway · Midjourney BYOK · fal.ai No Install Required

Guide

How to Use the Upscaler

Quick Setup — 3 Steps

Get a fal.ai API key

Free account at fal.ai/dashboard/keys. You pay fal.ai directly for compute — DVP never touches your billing.

Paste your key above

Stored only in your browser localStorage. Never sent to DVP servers. Click the key panel to expand it.

Drop a file, click Upscale

Choose your tab, drop or click to upload, select settings, and start. Images are fast — videos take 2–6 min.

◈ Image Tab ~15–45s

Upscales still images up to 4×. Standard runs SeedVR2. Use the Quick Preset buttons (Portrait/Face, Anime, Quick 2×, Max 4×) to set everything at once, or choose by hand.

Settings

2× vs 4× — 2× for web sharing; 4× for print, compositing, or archiving

Standard (photos) — Best for Kling, Runway photorealistic outputs. Sharpens edges and recovers fine texture.

Anime / Illustrated — Optimized for Midjourney illustration style. Handles flat tones without over-sharpening.

Standard + Face Restore — Runs Topaz Precision with face enhancement on. Best for portraits.

Formats: JPG · PNG · WebP · HEIC — max 20MB

▶ Video Tab 2–6 min

Upscales video 2× or 4×. Precision (Topaz) stays close to the source and is the default. AI footage (Topaz Starlight) rebuilds detail on soft or artifacted clips and costs about four times as much. Draft (Real-ESRGAN) is cheaper, takes 1000 frames at a time, and suits previews.

What each scale achieves

2× — 720p→1440p · 1080p→2160p (4K). Recommended for most Kling and Runway outputs.

4× — 540p→2160p · 720p→2880p. Maximum quality, takes longer. Best for hero shots.

          ⚠ Keep the tab open while processing. The result auto-saves to History so you can switch tabs, but don't close the browser.
        

Formats: MP4 · MOV · WebM · M4V

⊞ Batch Queue Tab Multiple images

Queue multiple images and upscale them sequentially with a single click. Each item is processed one at a time and saved to History automatically so you can keep working while jobs run.

How to use

1. Add files — drag and drop or click to select multiple images at once

2. Set scale and model — all items in the queue use the current Image tab settings

3. Run Batch — items process sequentially; status icons (○ pending / ◌ running / ✓ done / ✕ failed) update live

⟳ Enhance Tab — 60fps 2–4 min

Uses Topaz Interpolate (Apollo) to synthesize the in-between frames, converting 24fps or 30fps video to 48fps or 60fps.

Pro workflow (max quality)

1Upscale your clip in the Video tab (2× or 4×)

2Download the upscaled result

3Drop that file into the Enhance tab

4Select 60fps → Boost

History & Cost Reference

Jobs save to your browser's local History automatically. The last 25 results are kept — download links expire in ~1 hour after fal.ai purges the output CDN.

Typical cost per job (fal.ai)

Image 2× (1080p frame)
Standard 1¢ · Face 8¢ · Creative 33¢

Image 4× (1080p frame)
Standard 3¢ · Face 16¢ · Creative $1.33

Video 2× (10 s, 1080p → 4K)
Precision $0.60 · Starlight $2.60

Video 4× (10 s, 720p → 4K)
Precision $0.60 · Starlight $2.60

Frame rate to 60fps (10 s)
$0.30 at 1080p · $0.60 at 4K

Costs vary by input resolution and clip length. fal.ai list prices as of 2026-10-14. Billed by fal.ai directly — DVP has no billing relationship.

When a production packet is active, completed outputs also create delivery receipts. Use the delivery panel to download a manifest with preset, scale, fps, filename, and output URL metadata.

Tips & Gotchas

Download results immediately. fal.ai output URLs expire in ~1 hour. Don't leave them in History and assume they'll be there tomorrow.

Save the manifest. For Core Workflow packets, the delivery manifest is the clean record of which preset, model, scale, fps, and output URL finished the shot.

Source model matters for images. Kling outputs are photorealistic — use Standard. Midjourney illustrations use Anime mode for cleaner upscaling without over-sharpening.

Large videos upload slowly. Your file goes browser → fal.ai CDN before processing starts. A 100MB clip may take 30–60s just to upload before the GPU begins.

Test with images first. Before running a long video job, test your key and pipeline with the Image tab on a still frame — it's 10× faster and costs pennies.