Riverside just added AI that turns podcast recordings into newsletters. It's one instance of the biggest workflow shift in content: the repurposing pipeline. Toggle the outputs below and watch one recording fan out.
Drag to rotate. Toggle formats to add or remove output nodes from the pipeline.
2Assets per episode
5.0 hManual repurposing
1.1 hWith AI + human edit
78%Time saved
How the AI pipeline actually works
Transcribe. Speech-to-text converts a 60-min episode into ~9,000 words with speaker labels and timestamps.
Segment. The transcript is chunked by topic shifts, so the model works with coherent sections, not a wall of text.
Extract. An LLM pulls key claims, quotes and stats — the raw material every format shares.
Reformat. Each output gets its own prompt: a 500-word newsletter, 3 clip suggestions with timecodes, a thread hook.
Human pass. The step tweets skip: someone still checks facts, tone and names before hitting send.
Worked example: the time math
Task (60-min episode)
Manual
AI-assisted
Transcription
60 min
3 min
Newsletter draft
90 min
15 min edit
Clip selection
75 min
20 min review
Show notes
45 min
10 min edit
Total
4.5 h
~48 min
The honest caveat: AI drafts are 80% products. Newsletters that go out unedited read generic and can hallucinate quotes. The winning workflow is AI for structure, humans for judgment — which is why "time saved" here is 75–85%, never 100%.