Skip to content
Back to home

Changelog

What's new

Features, improvements, and fixes — shipped regularly.

Feature

Viral Clip Suggester — 9:16 export with auto-captions

The clip suggester now exports short-form clips directly as 9:16 vertical video with burnt-in captions derived from the live transcript. Clips are scored by engagement signals and ranked. Export to MP4 directly from the clip panel.

  • One-click 9:16 export with transcript-derived captions
  • Engagement scoring based on speech density and transcript sentiment
  • Clip library for managing and downloading generated clips
Improvement

Podcast 2.0 chapters and transcript RSS tags

RSS feeds now include Podcast 2.0 namespace tags for chapters, transcripts, and per-episode images. Chapter timestamps are automatically derived from the AI-generated show notes outline. Transcript files are hosted and linked per episode.

  • <podcast:chapters> auto-generated from show notes outline
  • <podcast:transcript> linked per episode (VTT + SRT)
  • <podcast:images> per-episode art supported
Feature

Guest recording — invite link flow

Hosts can now invite guests to a recording session via a shareable link. The guest joins in their browser with no account required. Audio tracks are recorded separately per participant and merged server-side after the session.

  • No-account guest join via shareable link
  • Separate per-participant audio tracks
  • Server-side audio mix with configurable volume levels
Improvement

Filler word detection accuracy improvement

The filler word detector (targeting "um", "uh", "like", "you know", "right") was re-tuned against a larger corpus. False positive rate dropped by approximately 35% in internal testing. The confidence threshold is now configurable per workspace.

  • ~35% reduction in false positive detections
  • Configurable confidence threshold (0.6–0.95)
  • Preview mode to review detections before committing removal
Fix

Audio export crash on files over 2 GB

Long-form recordings (90+ minutes at 48kHz stereo) triggered an out-of-memory error during WAV export. The export pipeline now streams audio chunks through the audio engine in segments and reassembles on disk, supporting files up to ~8 GB without buffering the full file in memory.

Older changelog entries available on request. Contact us