Product video sound: when to add it, and how with AI
Browsers generally block autoplay with sound until a visitor interacts, so a looping product clip must work muted. Save sound for clips people play.

A product video that plays on its own should be muted: browsers let muted video and video with no audio track autoplay, and generally block autoplay with sound until the visitor has interacted with the site. So a looping product-page clip should show the product, the action, and any key words in the picture. Add sound, a voiceover or music, to clips the shopper chooses to play. With AI you can generate the clip with its own sound, generate it silent, or lay a voiceover and a music bed over it afterwards.
The playback rules come from MDN's autoplay guide, Chrome's autoplay policy, and Shopify's video_tag reference, read on 2026-09-28. The Sume steps come from the Video generation and Timeline 1.0 docs and current Sume code.
Do product videos autoplay with sound?
Usually not before the visitor interacts with the page. MDN says autoplay blocking does not apply to a video with no audio track or a muted one, while media with an active audio track is subject to it. Chrome's policy is short: muted autoplay is always allowed, and autoplay with sound is allowed only in cases such as the user having already clicked or tapped on the site.
Shopify's video_tag filter, for example, takes autoplay, loop, muted, and controls attributes, and its example for product media turns all four on, muted included:
{{ media | video_tag: autoplay: true, loop: true, muted: true, controls: true }}Should my product video have sound?
- An autoplaying loop on a product page or banner: plan for silence. Put the benefit in the picture or in on-screen text, and generate the clip without audio if it will only ever loop muted.
- A clip the shopper starts, with controls: sound can play once they press play, so a voiceover can explain what the picture can't.
- An ad or a social post: follow that platform's own spec page for audio, and check how its player starts.
How do I make an AI product video with or without sound?
On Sume, sound is set per request with generate_audio on POST /v1/videos, and each model's catalog entry says whether it can make audio. Which video models generate sound lists every model; this table is the short version for product clips.
| You want | How | Note |
|---|---|---|
| A clip with its own sound | generate_audio: true, or omit it on a model with audio | Omitted, it defaults to the model's audio capability |
| A silent clip | generate_audio: false on the Seedance models, wan-3.0, or kling-3 | minimax-h3, minimax-h3-max, and gemini-omni-flash-1.1 always make sound and refuse false |
| Sound removed from a finished clip | Video trim with audio: "drop" | Reads this workspace's media.sume.com files |
| Voiceover and music over the clip | Timeline 1.0: audio spine plus optional soundtrack | Each clip's own audio is dropped (current code) |
How do I add a voiceover and music to a product video?
Render the clip in Timeline 1.0: the voiceover goes on the audio spine (audio.url, one Sume-hosted file) and the music bed in soundtrack, which can duck under the voice. In current code a Timeline render takes sound only from the spine and the soundtrack, so any sound the clip was generated with is not in the output; if the clip will sit under a voiceover, it can be generated silent.
How to make an AI voiceover covers the voice file, and Add background music to a video covers the soundtrack fields and ducking.
Can I caption the video for people watching muted?
Yes, if the clip has a voice. Sume's captions API burns captions into the picture, so they show with the sound off. In current code the caption job refuses a source with no audio stream and a source over 60 seconds, so caption the voiced version, not the silent loop. How to burn captions onto a video shows the call.
Sources
Related posts
More in Media tools
- Roku ad specs: video and audio rules for Roku Ads Manager
Roku ads must be 16:9, 6–92 seconds, .mov or .mp4 up to 1 GB, 23.98, 25 or 29.97 fps, 6 Mbps or more, with 48 kHz 2-channel stereo audio.
- Social media video sizes: 8 platforms in one table
Vertical 9:16 is the video shape most platforms list. The sizes TikTok, Instagram, Facebook, YouTube, Snapchat, X, LinkedIn, and Pinterest give.
- How to split a video by file size without re-encoding
Split a video by file size: scale its length by your limit ÷ its size, cut parts that long without re-encoding, then check each part's size.
- SRT vs VTT: the difference between the two subtitle formats
SRT and VTT are both plain-text subtitle files. VTT adds a WEBVTT header, dot milliseconds, positioning and styling, and is what HTML5 players read.
Written by Sume