Newsletter issue as a 60-second avatar video and MP3

A newsletter can ship a spoken summary of each issue: generate a talking video of up to 60 seconds with Sume's avatar video, then pull the audio as an MP3 with audio detach for beehiiv's audio block. It is a summary, not a full read-aloud, because avatar scripts are limited to 4 to 60 seconds.

Sume's docs list no audio-only narration endpoint in the pages I read, so the voice here is the avatar video's. beehiiv's side is from its audio support article, updated August 27, 2026 per the page (read 2026-10-02).

What fits in 60 seconds?

Roughly the three headlines and one line of context each, plus a sign-off. Sume estimates the duration from the script and rejects anything outside 4 to 60 seconds, so write the summary first, submit it, and cut if it is refused. A full 2,000 word issue is far beyond the window and would have to be split into several jobs, which would sound like a list of fragments.

Pick the framing that suits the medium: the audio block is for listening, so avoid sentences that depend on a link or a chart. Say 'the table below' only if the reader can see it.

How do I produce the video and the MP3?

Submit the script to POST /v1/avatar-1.0/talking-video with an avatar_handle, then read the finished job's video_url, a media.sume.com artifact. Audio detach accepts that kind of URL directly, so no re-import is needed. Set format to mp3, which is 128 kbps; the default is wav.

Two calls, one for each step:

# 1. the video (async: poll /v1/jobs/:id/result for video_url)
curl -X POST https://api.sume.com/v1/avatar-1.0/talking-video \
  -H "Authorization: Bearer $SUME_API_KEY" -H "Content-Type: application/json" \
  -H "Idempotency-Key: issue-42-summary" \
  -d '{"avatar_handle":"sume_clawra","aspect_ratio":"16:9","script":"This week: three things worth your time."}'

# 2. the MP3 from the finished video
curl -X POST https://api.sume.com/v1/audio-detach \
  -H "Authorization: Bearer $SUME_API_KEY" -H "Content-Type: application/json" \
  -H "Idempotency-Key: issue-42-mp3" \
  -d '{"video_url":"https://media.sume.com/artifacts/artf_demo/issue-42.mp4","format":"mp3"}'

How does the audio block treat it?

beehiiv accepts mp3, wav, ogg, flac, aac and webm, inserted with /audio under Embeds. Email readers see a thumbnail, title and Play Online button that opens the web version, while the site shows a full player. You can limit the block by audience, such as paid tiers only.

That means the MP3 reaches inbox readers one click away, not inline. If the summary is the hook, put its key point in the text above the block too.

What should you disclose?

A spoken summary delivered by an avatar is synthetic. Tell subscribers in the issue text, in plain words near the audio block, and do not present the avatar as a real person on your staff. Rules on synthetic performers vary by place; the posts on California SB 1050 and audio detach pricing cover the two pieces this post touches. Track both jobs with Jobs and results.

What can go wrong?

The audio detach job fails with detach_source_has_no_audio if the video has no audio track, which should not happen for a talking video but is worth checking if you swap the source. A script outside 4 to 60 seconds is refused up front, and an audio file you export from the video keeps the avatar's voice, with no way to change that in the detach step.

Listen once before sending. Names, numbers and product terms are the usual trouble, so spell tricky words phonetically in the script if they sound wrong.

Keep the video too. The same job gives you a clip for the web version of the issue or social, so one script yields both the audio block and a short video.