Why audio needs a video wrapper at all
No major social platform accepts a bare audio file. Instagram, TikTok, LinkedIn and X all take video, so the standard way to promote something that is fundamentally audio — a podcast episode, a track, an interview — is to wrap it in a video that gives the eye something to hold onto while the ear does the work.
That wrapper does two measurable jobs. It stops the scroll, because motion in a feed is what interrupts a thumb, and a waveform that reacts to speech reads as motion. And it survives silent autoplay: the title on screen tells someone what they would hear if they turned the sound on, so the clip is not wasted on the majority of viewers who never do.
This is why the format has a name at all. An audiogram is a specific, boring, repeatable artifact: still background, cover art, a line of text, a moving waveform, usually a progress bar. Its value is not artistic — it is that you can make one for every episode without thinking about it, and this tool exists to make that repetition take two minutes rather than twenty.
Choosing the clip, which matters more than the design
The single biggest determinant of whether a promo clip works is which 45 seconds you pick, and the most common mistake is picking the opening. Episode openings are introductions, throat-clearing and context — the least interesting audio you have. Take the moment somebody says something surprising, disagrees, gives a number, or lands a joke.
Practical constraints worth knowing before you cut:
- Length. 30 to 60 seconds suits every feed. Under 30 seconds tends to end before the point lands; over 90 seconds, completion rates collapse.
- Start on a sentence, not mid-word. Drag the left handle to just before an in-breath. A clip that starts mid-syllable sounds broken in the first half second, which is exactly when people decide to keep watching.
- End on a full stop. Ending on a trailing "and, uh…" undoes a good clip. Leave a beat of silence at the end rather than cutting hard.
- One idea only. If you find yourself extending the clip to include the setup, the setup belongs in the caption instead.
The waveform overview above the controls is there for exactly this: speech has a visible rhythm, and pauses between sentences show up as narrow gaps. Dropping the handles into a gap is usually enough to land a clean cut without listening through the whole file.
Format, and posting the same file in several places
| Format | Pixels | Best for | Notes |
|---|
| Square 1:1 | 1080×1080 | LinkedIn, Instagram feed, X | The safest single file — no platform crops it badly |
| Vertical 9:16 | 1080×1920 | Reels, Shorts, TikTok | Fills the screen; keep text away from the very bottom, where the caption UI sits |
| Portrait 4:5 | 1080×1350 | Instagram feed | The tallest the feed allows — takes more vertical space than square |
| Wide 16:9 | 1920×1080 | YouTube, X, embeds | Also the right shape for a website or an email newsletter |
If you only want to make one file, make it square. If you are willing to export twice, square plus vertical covers essentially everything, and the second export is just a dropdown change and another click — the trim, colours, artwork and text all stay as they are.
How the video is actually produced
Everything runs in the page. Your file is decoded with the Web Audio API into raw samples, which is what makes the waveform possible in the first place. As each frame is drawn, a 2048-point FFT analyses the short window of audio under the playhead and the result is grouped into log-spaced frequency bands — log-spaced because linear bins would crowd almost everything audible into the leftmost few bars and leave the right-hand side motionless.
The picture is a canvas, and the sound comes from the same audio graph that is playing, routed into a media stream. Those two are combined into one stream and handed to the browser's recorder, which is why the downloaded file has real audio rather than being a silent animation you would have to re-marry to the audio in an editor.
One consequence is worth stating plainly: recording happens in real time. A 45-second audiogram takes 45 seconds to export, because the recorder is capturing the canvas and the audio as they genuinely play. The frame clock is driven from the audio clock, so the waveform stays locked to the sound rather than drifting, and it uses timers rather than animation frames so that a background tab cannot silently freeze the export halfway through.
The trade-off of doing it locally is that output is WebM on Chrome, Edge and Firefox, and MP4 on Safari — the browser decides, not the tool. Every major platform accepts WebM on upload. Convert only if some desktop editor in your workflow insists on MP4.
Making the second one faster than the first
Your style, colours, format, subtitle and toggles are remembered in this browser, so the next clip starts already looking like your show. Nothing is stored anywhere else and nothing is tied to an account — clearing site data resets it.
A workflow that survives a weekly release schedule looks like this: set the accent to your brand colour once, put the show name in the subtitle once, load the artwork once, and thereafter the per-episode work is dropping the audio, dragging two handles and typing a hook line. That is the part designed to be repeatable, because a promo format only pays off if you actually keep doing it.
Two habits worth adding. Put the hook in the title rather than the episode number, because "Episode 42" tells a stranger nothing and the quote tells them everything. And keep a consistent accent colour across episodes — over months, that consistency is what makes clips recognisable in a feed before anyone reads the words.