nectr

AI content director — for podcasts

Episode in.
Clips out.

Give nectr a full episode. It comes back as vertical clips — cut on word boundaries, framed on the person speaking, captioned, normalised to −14 LUFS. A 48:41 episode runs end to end in 48 minutes.

Request access See the evidence ↓
MEASURED RUN INPUT 48:41 OUTPUT 3 × 1080×1920 TURNAROUND 48 MIN ≈1.0× REALTIME

01 — Evidence

One source, drawn.

This is the actual edit of a 12:00 excerpt of a Linus Torvalds interview (CC BY) — every kept span, with the in and out points nectr chose, exactly as delivered.

SOURCE — TORVALDS INTERVIEW (CC BY) · 12:00 · 720 s KEPT SPANS — 3 0:00 2:00 4:00 6:00 8:00 10:00 12:00 01 · 47.10 → 102.72 02 · 290.88 → 342.86 03 · 611.74 → 657.86 DETAIL A
Spans drawn to scale at their measured positions on the 720 s source. Clip 01 enlarged and dimensioned in Detail A.
DETAIL A — CLIP 01 SCALE 42:1 SPAN 55.62 s IN 47.10 s OUT 102.72 s CLIP 01 1080 × 1920 63.97 s — FRAME CUTS TO THE ACTIVE SPEAKER 0:40 0:50 1:00 1:10 1:20 1:30 1:40 1:50
In and out points sit on word boundaries. At 63.97 s the speaker changes — the frame follows. Span on the source: 55.62 s; the rendered file is 55.2 s after the crossfade trim.
DOCEDIT DRAWING — TORVALDS EXCERPT
SOURCE12:00 · 720 s · CC BY
OUTPUT3 CLIPS · 1080×1920
CUTSSNAPPED TO WORD BOUNDARIES
FRAMINGHOLDS THE ACTIVE SPEAKER
CAPTIONSWORD-SYNCED, BAKED IN
LOUDNESS−14 LUFS
SHEET1 / 1

The clips it sent back.

SOURCE: 12:00 EXCERPT — LINUS TORVALDS INTERVIEW (CC BY) · PLAYS WITH SOUND
CLIP 01RENDERED 55.2 s
IN 47.10 → OUT 102.7200:00.0
CLIP 02RENDERED 51.5 s
CAPTIONS WORD-SYNCED00:00.0
CLIP 03RENDERED 45.7 s
−14 LUFS00:00.0

02 — How it works

The cut, step by step.

01 Listen

Episode-level diarization separates every voice, and each speaker is bound to their face on screen.

02 Cut

In and out points snap to word boundaries. Mid-clip, when the voice changes, the frame cuts to the person talking.

03 Finish

Captions are word-synced and baked in. Audio is normalised to −14 LUFS. Files arrive vertical, 1080×1920, ready to post.

You stay the editor. nectr drafts the clips; you decide what ships.

03 — Beyond podcasts

Same machine, pointed at a kitchen.

Podcasts are where nectr works today. To show the range: 41:07 of raw cooking footage, watched end to end — 14 shots kept, 97.2% discarded.

A FULL RUN, FILMED — 41:07 IN, 1:08 KEPT

Job card

FOOTAGE41:07
KEPT1:08
DISCARDED97.2%
SHOTS KEPT14
TAKES DETECTED33
FRAMES LABELLED1,234
TRANSCRIPT SEGMENTS518

The 14 kept moments — frames 00–13

Kept moment frame 0000
Kept moment frame 0101
Kept moment frame 0202
Kept moment frame 0303
Kept moment frame 0404
Kept moment frame 0505
Kept moment frame 0606
Kept moment frame 0707
Kept moment frame 0808
Kept moment frame 0909
Kept moment frame 1010
Kept moment frame 1111
Kept moment frame 1212
Kept moment frame 1313

04 — Access

Request access.

Early access is small on purpose. Send an episode; we run it and return your clips — then you tell us where the cut is wrong.

We store your email, your answer, and — with the request itself — your IP address and browser string. Nothing more: no list sharing, no tracking pixels, no analytics. Delete any time: 112@nectr.video. Privacy policy.