Quick answer

FFmpeg: ffmpeg -i input.mp4 -vf scale=1920:1080:flags=lanczos -c:v libx264 -crf 18 -pix_fmt yuv420p output_1080p.mp4. DaVinci Resolve (free): right-click the clip in the Media Pool → Clip Attributes → enable Super Scale → 2x → render. Both beat naive resizing; neither invents detail the sensor never captured.

Honest expectations first

Upscaling never adds real detail — an 848×560 frame contains exactly 474,880 pixels of information, and no free algorithm can know what the missing pixels “should” have been. What good upscaling does do: produce clean, sharp-edged larger frames without the blur, ringing, or blockiness of naive resizing, and qualify your video for YouTube’s higher-bitrate 1080p encoding tier. Paid AI tools go one step further with models trained to hallucinate plausible detail — that is the real gap between free and paid, and no free scaler closes it. If you want the best possible result regardless of cost, see our Topaz upscaling guide.

Before you upscale: convert first

Upscale from the cleanest source you have. If your footage is still in eos-movrec’s original MJPEG AVI, convert to H.264 MP4 at CRF 18 first — upscaling a heavily compressed file just magnifies compression artifacts. And upscale last: edit at native resolution, then upscale the finished timeline, so every intermediate step stays fast and clean.

Method 1 — FFmpeg with the lanczos scaler

FFmpeg’s default scaler is bicubic; lanczos is sharper and worth the few extra seconds. To take 848×560 footage to 1080p:

ffmpeg -i input.mp4 -vf scale=1920:1080:flags=lanczos -c:v libx264 -crf 18 -pix_fmt yuv420p output_1080p.mp4

Notes:

• Target size: 1920×1080 works, but be realistic — from 848×560, 1280×720 (scale=1280:720:flags=lanczos) is the honest sweet spot. 1080p is a 2.3× upscale of every dimension; 720p is gentler and usually looks cleaner.
• CRF 18 keeps the encode near-transparent so the scaler’s work is not undone by compression.
• Alternative scaler: flags=spline is slightly softer than lanczos with fewer ringing artifacts — try both on a still frame and pick what you prefer.
• Upscaling will not fix a shaky or badly exposed shot. Garbage in, larger garbage out.

Method 2 — DaVinci Resolve Super Scale (free version)

Resolve’s Super Scale — available in the free version — is a step above lanczos: it analyzes the image and applies edge-aware sharpening during the upscale. Workflow:

1. Finish your edit at native resolution on the timeline as usual.
2. In the Media Pool, right-click each source clip → Clip Attributes → check Super Scale and choose 2x.
3. Set your timeline resolution to 1920×1080 (Timeline → Timeline Settings → uncheck “Use Project Settings”).
4. On the Deliver page, export with the YouTube preset at 1920×1080.

Super Scale 2x on 848×560 gives 1696×1120, which Resolve fits cleanly into a 1080p timeline. Rendering takes noticeably longer than a normal export — that is the algorithm doing its work.

Free vs. paid: the honest comparison

Method	Cost	What you get
FFmpeg lanczos	Free	Clean, sharp upscale; no invented detail. Best for quick jobs.
Resolve Super Scale	Free	A notch above lanczos — smarter edges, slightly more natural textures.
Topaz Video AI	Paid	AI models trained to reconstruct plausible detail — visibly better on faces and textures, at a price.

Our recommendation: try Super Scale first — it is free and surprisingly capable. If the result still is not clean enough for your project, the paid Topaz route is the documented next step, and Topaz Video AI is the tool built exactly for this kind of old-footage rescue.

Does upscaling help on YouTube?

Modestly, yes — for a technical reason: YouTube assigns higher bitrates to 1080p uploads than to 560p ones. An upscaled-to-1080p upload gets more bits per second from YouTube’s encoder, which can slightly reduce the mushiness YouTube would otherwise inflict on a tiny native-resolution upload. It will not look like native 1080p footage, but it usually looks better than uploading 848×560 directly.

Frequently asked questions

Will free upscaling ever look as good as Topaz Video AI?

No — and anyone who tells you otherwise is selling something. Free scalers (lanczos, Super Scale) rearrange the pixels you already have as intelligently as possible; Topaz uses neural networks trained on millions of images to reconstruct plausible detail like skin texture and hair. The gap is real and visible on close inspection. Free methods are still well worth doing — they are a large improvement over naive resizing — but they have a ceiling.

What’s the best FFmpeg scaler for this footage?

lanczos is the standard recommendation: sharper than the default bicubic without going overboard. If you see ringing (halos along high-contrast edges), try spline instead — slightly softer, fewer artifacts. Both are one word in the scale filter: scale=1280:720:flags=lanczos.

Should I upscale to 1080p or 4K?

1080p at most — and honestly, 720p is the sweet spot from 848×560. Going to 4K (3840×2160) means inventing roughly 17 pixels for every real one; no free algorithm does that gracefully, and the file balloons for zero visible gain. Even the paid route in our Topaz guide treats 1080p as the sensible target for this footage.

Does upscaling improve how my video looks on YouTube?

A bit, for a boring technical reason: YouTube gives 1080p uploads a higher bitrate than 560p uploads. So an upscaled 1080p file survives YouTube’s re-encoding with less additional damage than a native 848×560 upload. It will not look like true HD, but side-by-side the upscaled upload usually wins.

Is Super Scale really in the free version of DaVinci Resolve?

Yes. Super Scale (Clip Attributes → Super Scale → 2x/3x/4x) is included in free DaVinci Resolve — it is not a Studio-only feature. The free version also handles the entire eos-movrec editing workflow, so you can edit and upscale without spending anything.

Should I upscale before or after editing?

After. Edit at native resolution with lightweight files, then upscale the finished timeline as the final step. Upscaling first means every cut, effect, and export operates on bloated files for no benefit — slower editing, bigger intermediates, identical final quality.