Add audio sample support to AppStream metadata (e.g. for TTS voice previews)

[yeager]

Summary

AppStream currently supports images and videos in <screenshots>, but has no way to include audio samples. This would be valuable for components where audio is the primary "preview" — most notably text-to-speech (TTS) voices.

Use Case

The Spiel project provides a TTS framework for Linux with installable voice packages distributed via Flatpak. The spiel-installer app lets users browse and install voices, but currently there is no way to preview how a voice sounds before downloading it.

See: project-spiel/spiel-installer#2

Audio previews would let users hear a voice sample before committing to a download — similar to how screenshots let users see an app before installing.

Proposed Solution

Add an <audio> element as a sibling to <image> and <video> within <screenshot>, or introduce a new top-level <samples> element. For example:

Option A: Extend <screenshot> with audio media kind

<screenshots>
  <screenshot type="default">
    <audio type="vorbis" xml:lang="sv">https://example.org/voices/alma_sample.ogg</audio>
    <caption>Swedish female voice sample — "Hej, jag heter Alma"</caption>
  </screenshot>
</screenshots>

This follows the existing pattern where <screenshot> already supports both <image> and <video> elements, and would add AS_SCREENSHOT_MEDIA_KIND_AUDIO alongside the existing image/video kinds.

Option B: New <audio-samples> element

<audio-samples>
  <sample xml:lang="sv" type="vorbis">https://example.org/voices/alma_sample.ogg</sample>
  <sample xml:lang="en" type="vorbis">https://example.org/voices/alma_en_sample.ogg</sample>
</audio-samples>

Benefits

• TTS voices: Users can preview voice quality, accent, and naturalness
• Music apps/plugins: Sound fonts, instruments, audio effects could include samples
• Accessibility: Audio previews complement visual screenshots for users who rely on hearing
• Codec/media components: Could demonstrate supported formats

Audio Format Considerations

Similar to how videos support multiple containers (WebM, Matroska), audio samples could support:

• Ogg Vorbis (.ogg) — open, widely supported
• Opus (.opus) — modern, efficient
• FLAC (.flac) — lossless option

Short samples (5-15 seconds) would be typical, keeping file sizes small.

Prior Art

• Video support was added via issue Video as additional metadata #175, extending the screenshot element — a similar approach could work for audio
• Android and iOS app stores support video previews; some voice assistant marketplaces include audio samples

I would be happy to help with a specification draft or implementation if this is of interest.

[ximion]

Why couldn't you just submit a video file that has the audio embedded? Supporting a new, pure-audio option for basically just one application seems rather annoying, permitting Opus audio in video though is something we could do without breaking everything else.

[yeager]

That's a pragmatic approach — embedding Opus audio in a video container (e.g. WebM with audio-only) would work for TTS voice previews without requiring a new media type.

Would you accept a PR that extends the existing video/screenshot handling to also permit audio-only WebM/Opus files? Or would you prefer a simpler approach like just linking to an external audio sample URL in the metadata without any new elements?

Happy to implement whichever fits best.

[vixalien]

Why couldn't you just submit a video file that has the audio embedded?

This will only work for Opus, not MP3, flac, etc... which are much more common.

Using videos also will make app stores display them as videos instead of native audios