Video rendering: what happens between “render” and the MP4
Video rendering is the step that turns a project (scenes, footage, text, audio and timing decisions) into a finished video file by drawing every frame and encoding the result. In a desktop editor, video rendering happens on your own computer when you export; in PageToVid it happens on a cloud machine reserved for your film, and it ends with an MP4, a poster image and a subtitle file.
This guide is for the marketer watching a progress bar who wants to know what the machine is doing, why it takes minutes, what is in the file and what doing it again costs. Every figure comes from our render code or a file we measured.
AEK Hub (aekhub.gr), a news and match-centre site for AEK Athens fans, in a 45-second 16:9 film with 91 spoken words; we measured its MP4 for the worked example.
What rendering actually does
Until it is rendered, a video is a set of instructions: this clip from second 3 to second 7, this title sliding in over it, this voice underneath, this caption at 4.2 seconds. A renderer reads them and produces the frames, one still image after another, at a fixed rate. An encoder then compresses the frames and the mixed audio into a file a player can stream.
That is why rendering is slow in a desktop editor: your processor computes every effect on every frame, and a minute of 30 fps video is 1,800 frames. It is also why “export” and “render” share one button.
PageToVid has no timeline for you to manage, but the stages are the same. Your storyboard is the instruction list, our compositor draws the frames, the encoder makes the MP4.
What our render job does, step by step
In PageToVid, “render” names the whole job that turns a storyboard into a film. The site is filmed and the voice recorded first; the steps below are the compositing pass that ends it.
Start a job of its own
The web server only queues the render. Each film then runs as a separate job on a machine of its own, so a site update or another customer's film cannot interrupt it. If the job crashes, it is started again automatically.
Check every file the film points at
Scene images, the music bed and the brand logo are fetched before the compositor starts. A missing or broken file is dropped and noted in the render's repairs, instead of aborting a film whose recording and voice are already done.
Compose the scenes
The storyboard becomes one composition: the canvas for the chosen shape, 30 frames per second, each scene's length taken from its slice of the voice track, captions, transitions, brand colours and logo.
Draw the frames
Remotion draws each frame in a headless Chrome browser. If the browser fails mid-render, the film is drawn again more slowly rather than failed.
Encode and master
Frames are encoded to H.264 at CRF 18 (a high-quality setting) in the yuv420p colour format every player reads. A final audio-only pass sets the mix to −14 LUFS and −1.5 dBTP and moves the index to the front of the file so playback starts before the download ends.
Publish the outputs
The MP4 is uploaded with a designed poster (hook text and a browser-framed screenshot in your brand colours) and a WebVTT subtitle file timed from the recorded voice. If you set a webhook through the API, a signed completion event goes out.
A render cannot run forever either. A task has a 30-minute ceiling; shortly before it, a watchdog ends the render cleanly and refunds it, instead of leaving you in front of a progress bar that will never finish.
Worked example: the AEK Hub file, measured
AEK Hub is a fan site for AEK Athens that gathers news and fixtures across football, basketball, handball and futsal. The film opens on a question, “Missing crucial updates for your favorite AEK team?”, shows the news feed and the team categories, and ends on the Match Center countdown with “Visit the Match Center now”. It speaks 91 words in 45 seconds.
We ran ffprobe on the published MP4. This is what a PageToVid render actually delivers:
| Property | Value | What it means for you |
|---|---|---|
Duration | 45.16 s (about 1,355 frames) | Every one of those frames was drawn, then encoded |
Video | H.264, 1920×1080, 30 fps | Plays everywhere: browsers, phones, LinkedIn, YouTube |
Video bitrate | about 3.5 Mb/s | Sharp text on a screencast without a heavy file |
Audio | AAC, 48 kHz stereo, about 192 kb/s | The voice and music mix, mastered for web players |
File size | 20.9 MB | Small enough to upload anywhere or send by link |
Two things are not in the MP4: the poster image and the captions file. Captions are drawn into the frames and also delivered as a separate WebVTT file, so a player that wants a caption track has one. Subtitles and captions covers when to use which.
Shapes, canvas sizes and what gets filmed
The shape is decided before the render, because it changes what is filmed, not just how the frame is cropped. A 9:16 film records the site's own mobile layout in a phone viewport; 16:9 and 1:1 record the desktop layout. The compositor then draws on the canvas for that shape: 1920×1080 for 16:9, 1080×1920 for 9:16, 1080×1080 for 1:1, and 1080×1350 for 4:5 through the API or an assistant. Aspect ratio explains how to choose.
When one storyboard is ordered in several shapes, each extra shape is rendered from the same storyboard and the same voice track; only the recordings are made again for that shape.
How long rendering takes, and what it costs
From a pasted URL to a finished film usually takes about 10 minutes, and most of that is filming the site, writing and recording the voice. The compositing pass is measured against the film's own length: our target is to draw a film in no more time than it lasts. AI clips, if your plan uses them, are generated before the render and add their own wait.
A finished render costs 40 credits whatever its length or shape (pricing has the plans, and credits and what a video costs works through examples). The rules that matter when something goes wrong:
- A render that fails for a system reason is refunded automatically, once, with the reason written on the film.
- A re-render costs one render. Change a scene's words, visual, voice or music and render again.
- Captions only? Re-timing the captions from the existing voice track records nothing new and is free once per paid render.
- Want to see the cut first? A draft render is free: a third of the resolution, a DRAFT mark on every frame, no AI images or clips, limited to five an hour.
Fix a video without remaking it shows which change needs which kind of render.
Mistakes we see around rendering
- Re-rendering to fix caption timing. Use the captions re-time; it reuses the voice and is free once.
- Expecting a longer film from the same script. Length comes from the words, not a render setting; see narration is the length.
- Rendering 16:9 and cropping it for Reels. You lose the mobile layout the 9:16 render would have filmed.
- Re-downloading and re-encoding the MP4 before upload. Every extra encode costs quality; upload the file as delivered.
- Rendering before the page is ready. We film the live site (website to video), so a pricing table that changes tomorrow means a new render tomorrow.
When you need a different renderer
Our renders are built for web video: H.264 at 30 fps, 1080p-class canvases. If you need a 4K or 60 fps master, a ProRes or DNxHR file for broadcast, colour grading of camera footage or frame-accurate cuts on a timeline, use a desktop editor such as DaVinci Resolve, Premiere Pro or Final Cut Pro and export there. If you just need to show a colleague a bug, a screen recorder like Loom gives you a shareable link without any render step. All comparisons lists where other tools fit better.
Turn your website into a video — free
Paste the page you want filmed. PageToVid writes the script, films the real site, records the voice and renders the MP4.
Frequently asked questions
What does it mean to render a video?
Rendering means computing every frame of a video from its project (clips, text, effects, audio, timing) and encoding the frames into a playable file such as an MP4. Until it is rendered, a project is only a set of instructions.
Why is my video taking so long to render?
On a desktop editor, length, resolution and effects all multiply the work your computer must do per frame. On PageToVid, most of the wait is filming the site and recording the voice; the compositing pass aims to take no longer than the film itself.
What format are PageToVid videos rendered in?
An H.264 MP4 at 30 frames per second with AAC stereo audio, at 1920 by 1080, 1080 by 1920 or 1080 by 1080 depending on the shape, plus a JPEG poster and a WebVTT subtitle file.
Do I pay again if a render fails?
No. A render that fails for a system reason is refunded automatically. A re-render you ask for, after editing the film, costs one render.
Can I render in 4K or at 60 frames per second?
No. PageToVid renders at 30 frames per second on 1080p-class canvases, which is what web and social players show. For a 4K or 60 fps master, export from a desktop editor.