How to Download a Video From a URL to MP4, Properly

“Download this URL to MP4” looks like a trivial instruction and hides four decisions: which stream, which codec, which container, and what you will do with the file afterwards. Get those wrong and you end up with a 720p clip that will not import into your editor, or a 2GB download for a shot you only wanted to look at. Here is the version where you decide instead of accepting.

What “to MP4” actually means

An MP4 is a container — a box holding a video stream, an audio stream and timing. What is inside matters more than the extension: usually H.264 or H.265 video with AAC audio. Platforms serve several renditions of the same clip, so “download the MP4” is really a set of choices:

• Resolution — capped by what was uploaded and what the platform still serves; often 1080p maximum on social sources.
• Codec — H.264 imports everywhere; HEVC/H.265 is smaller and fussier.
• Separate versus merged streams — many services deliver video-only and audio-only strands that must be muxed, which is why “my download has no sound” is the most common complaint in this niche.
• Frame rate and scanning — variable frame rate from phones breaks timeline editors even when the file “plays fine”.

Three ways to get the file

1. Native download — the platform’s own button, where it exists (Reddit for hosted clips, TikTok and YouTube for your own uploads or where the app offers it). Simplest and most compliant.
2. A web tool — paste the URL, pick a rendition, get the file. SocialToPrompt does this across 27 named platforms plus anything yt-dlp resolves, at 1 credit per download on its metered model (3 free credits with a new account, valid 24 hours), and returns the source-quality stream the platform serves rather than a re-encoded copy.
3. Command line — yt-dlp "URL" locally: free, scriptable, and the only option at real volume. Its downside is that it needs maintaining whenever a platform changes something.

The web tool is right for one or two clips and for people who do not want a terminal; the command line is right for batches and for footage you would rather not send anywhere.

Choosing resolution and codec

Three situations, three answers:

Purpose	Pick	Why
Reference / study	720p, H.264	Small, opens everywhere, plenty for reading a shot
Editing into a project	Highest you are permitted, then transcode to an editing codec	Delivery codecs decode badly on timelines
Uploading to another platform	Native resolution, matching aspect	Every recompression costs detail
Analysing into a prompt	720–1080p, short segment	The analysis reads structure, not pixel count

To inspect what a file actually holds, ffprobe -v error -show_streams -of default=nw=1 file.mp4 tells you codec, resolution, frame rate and whether audio is present.

Post-processing for use

# constant frame rate + 1080p, the combination editors tolerate
ffmpeg -i in.mp4 -vf "scale=-2:1080,fps=30" -c:v libx264 -crf 20 -preset slow -c:a aac -b:a 160k out.mp4

# make a clean MP4 from a webm
ffmpeg -i in.webm -c:v libx264 -crf 20 -pix_fmt yuv420p out.mp4

# extract one 8-second shot for prompt analysis
ffmpeg -ss 4 -t 8 -i in.mp4 -c:v libx264 -crf 20 shot.mp4

The -pix_fmt yuv420p flag is the classic fix for “plays in QuickTime, black in Premiere”, and re-encoding also strips camera GPS and device metadata, which matters if the file is going to leave your machine. Full flag reference is in the ffmpeg documentation, and the prompt-side preparation is converting an MP4 to a prompt.

The two reasons you wanted a prompt instead

First: the file is not the information. If you saved a clip to work out why it worked — the reveal at second four, the hip-height camera, the hard rim light — a structured prompt captures that in 60 words you can act on, and it is portable across every generator you try afterwards. Second: the file carries obligations. Someone else’s clip in your drive is a copyright and platform-terms object (see the rights discussion), while a description of its camera technique is not the same kind of thing at all.

Neither is a rule: keep files when you need pixels — offline editing, frame inspection, archiving your own uploads. But most of the time the deliverable is a sentence about a shot, and that is what an extraction pass gives you in under a minute. Related reading: shot lists from long videos, and turning viral videos into prompts.

Frequently asked questions

How do I download a video from a URL to MP4?

Three routes: the platform's own download button where it exists, a web tool that accepts the URL and lets you choose a rendition, or yt-dlp locally. All three give you a container rather than a guarantee — check the codec and frame rate with ffprobe before assuming it will import into an editor.

Why does my downloaded video have no audio?

Most services deliver video-only and audio-only streams, and a tool that saves only the video rendition gives you exactly that symptom. Re-download with both streams merged, or mux them yourself with ffmpeg.

What resolution should I download?

Match the purpose: 720p for study, highest you're permitted for editing, native for reposting to another platform, and 720 to 1080p for prompt analysis, which reads structure rather than pixel count. Remember social platforms usually cap what they serve at all.

When is a prompt a better deliverable than a file?

Whenever you wanted the technique rather than the footage. A structured description captures the shot idea in a form you can reuse on any generator, takes 60 words instead of 60MB, and doesn't carry the copyright and platform-terms obligations that somebody else's video file does.