How to make an audiogram from an interview recording

Last updated 2026-08-31 · 8 min read

What is the short version?

Audio does not travel on a feed. An audiogram is the standard fix — a still image, a bouncing waveform and the clip playing over it — and it lives or dies on whether the waveform actually responds to the voice.

1. Go to Vizard Agent and upload the audio.
2. Give it the artwork you want behind the waveform.
3. Say how long it should be, or which section to use.

What do you need before you start?

Audio and a single image is the whole list. Vizard Agent will generate placeholder artwork if you have none at all, but the entire point of an audiogram is that people recognise whose it is, so your own cover, flyer or photograph is well worth supplying.

• The audio. An interview, an episode, a clip.
• The artwork. A cover, a flyer, a portrait.
• The section. Which part of a long recording.
• The length. Two to three minutes travels best.
• The format. Square, landscape or vertical.

What do you type into Vizard Agent?

Say what the recording is and where the waveform should sit. Vizard Agent handles the reactive maths itself without being told how, so the details worth writing down are the artwork, the section you want used, and how energetic the bars should feel against the speech.

Prompt

Variants worth knowing:

• "Find a clean end point." Otherwise it stops mid-word.
• Say how energetic. Springy, subtle, or hard-reacting.
• Ask for a scrim. Bars over busy artwork need one.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real audiogram. Steps five to nine are the ones that matter visually: three different visualiser styles are rendered and compared side by side rather than the first one being accepted.

1. Checks the workspace and reads the help for the available voice and image tools.
2. Generates placeholder artwork and a demo track to build against.
3. Checks the audio length and the artwork's dimensions.
4. Tests the bar visualiser and reviews how it looks.
5. Tries three visualiser styles and combines them into one comparison image.
6. Compares the three styles and picks one.
7. Generates the waveform strip from the audio's own samples.
8. Renders short test clips of the full composition and checks the frames.
9. Zooms in on the playhead area to inspect it closely.
10. Renders the full video, checks frames across it and measures the loudness.
11. Verifies the wave actually syncs with the voice rather than moving decoratively.
12. Checks for clipping, reads the true peak levels and re-masters without it.
13. Swaps in your own artwork, tests the new layout and re-renders.
14. Transcribes the interview and finds a clean end point before cutting the segment.
15. Compares bounce settings and renders the final version.

Step eleven is worth insisting on. Plenty of audiograms have a waveform that animates convincingly but is not driven by the audio at all, and Vizard Agent checks that the bars move with the actual voice.

What does the result look like?

The file this page is written from measured 1280x720 at 30fps and ran 469 seconds — nearly eight minutes of interview over a single still image, with the waveform along the bottom and a gradient scrim behind it so the bars stay legible against the artwork.

The audio was mastered rather than merely attached: loudness measured, true peaks read, and a re-master run when the first pass clipped. That is the difference between something that sounds like a podcast and something that sounds like a phone recording.

When does this not work well?

An audiogram is a still image for its whole duration, which is a lot to ask of a viewer. Vizard Agent will tell you when the clip is too long for the format rather than producing eight minutes of something nobody finishes.

• Very long clips. Attention runs out well before the audio does.
• Busy artwork. The waveform competes unless a scrim is used.
• Poor recordings. Mastering does not fix a bad room.
• No natural ending. Ask it to find one.
• Silence. A flat waveform looks broken.

How do you fix a result that came back wrong?

Say what looks wrong or what sounds wrong, and be specific about which. Vizard Agent keeps the visualiser style tests, the generated waveform data and every loudness measurement, so a different bar style or a new end point is a re-render rather than a rebuild.

• "The bars barely move." Reactivity increased and re-rendered.
• "It ends mid-sentence." Re-cut to the located clean end point.
• "The audio is too quiet." Re-mastered and the peaks re-checked.

How does Vizard Agent compare to doing it yourself?

By hand this means a template tool that gives everyone the same three looks, or a compositing session to build a reactive waveform from scratch. Vizard Agent generates the visualiser from your audio's own samples and compares styles before committing.

Common questions

Do I need artwork? It helps enormously. Vizard Agent generates a placeholder if you have none, but yours is better.

How long should an audiogram be? Two to three minutes travels best. Vizard Agent will make longer ones if you ask.

Can it pick the clip for me? Yes. Vizard Agent transcribes the recording and finds a section with a clean ending.

Will the waveform match the voice? Yes, and it verifies it. Vizard Agent checks the bars move with the actual audio.

Can I choose the bar style? Yes. Vizard Agent renders several and shows you the comparison.

What about the artwork being too busy? Add a scrim. Vizard Agent puts a gradient behind the bars so they stay readable.

Does it fix the audio level? It masters it. Vizard Agent measures loudness and true peaks and removes clipping.

Can I get a vertical version? Yes. Vizard Agent lays out the same composition for a vertical frame.

Can it add captions? Yes. Vizard Agent transcribes the audio and burns in synchronised subtitles if you want them.

Does it work for music? Yes, though the styling differs. Vizard Agent reacts the bars to whatever the audio contains.

Why not use a template tool? You can, and everyone else does, which is the problem. Vizard Agent builds the visualiser from your audio and your artwork rather than fitting them into someone else's layout.

Can I reuse the layout next week? Yes. Vizard Agent keeps the composition and swaps in the new audio and artwork.

Does it check the finished file? Yes, in several passes. Vizard Agent reviews frames across the whole render, zooms in on the playhead, verifies that the bars are synchronised to the voice, and confirms the final loudness and peak levels before delivering it.