---
name: "asmr_content_production"
description: "Produce ASMR content end to end: choose and place the mic, record quietly, edit out clicks and mouth noise, and package metadata (titles, chapters, tags) that surfaces the video. Triggers on 'make an ASMR video', 'ASMR recording', 'ASMR editing', or planning ASMR metadata."
---

# ASMR Content Production

## Purpose

Take an ASMR concept from trigger plan to published video: mic setup, quiet recording technique, click-free editing, and discoverable packaging. The listener's whole experience lives in the audio, so every step optimizes for clean, close, immersive sound.

## Workflow

### 1. Plan the session

- Pick one primary trigger (tapping, whispering, crinkling, brushing, scratching) and at most two secondary ones. One trigger done well beats five done thinly.
- Note the goal: sleep, relaxation, or tingles. It shapes the pacing and the title.
- Write the trigger timeline as timestamped chapters before recording. Chapters need at least 3 entries and the first must be `0:00` for YouTube chapters to activate.
- See `references/trigger-library.md` for the search vocabulary viewers actually type.

### 2. Set up the mic and room

- **Self-noise matters more than brand.** Use a mic with self-noise below 15 dB(A); ASMR levels are quiet enough that the mic's own floor is the limit.
- **Binaural for immersion.** Two capsules at ear spacing (dummy head, or a binaural rig like 3Dio Free Space / Sennheiser AMBEO) give headphone listeners left/right/near/far placement. High-quality stereo condensers work too; beginners get good results from a decent USB condenser.
- **Placement.** Whisper distance: 6 to 12 inches from the mic as the rule, 6 to 8 for extra intimacy; get closer only as a deliberate effect. Aim the mic diagonally or slightly downward so breath lands off-axis; use a pop filter when you have one. Cardioid patterns reject room noise; omni/stereo suits immersive pieces where the room is part of the experience.
- **Kill noise at the source.** Turn off fans, AC, refrigerators, computer fans; close windows and doors; record at quiet hours (late night); hang blankets or use acoustic foam on reflective surfaces. Denoising in post degrades the recording, so prevent first.
- **Monitor on closed-back headphones while recording** to catch clipping and intrusions in real time. Record 48 kHz or 96 kHz, 24-bit.
- Details and a comparison table: `references/mic-and-room-setup.md`.

### 3. Record quietly

- Move slowly and deliberately. Subtle changes in pressure and speed change the trigger's feel; consistency is the skill.
- Control breath and vocal delivery on whispers to avoid plosives and sibilance. Drink water; a bite of green apple before the session cuts mouth noise at the source.
- Leave 2 to 3 seconds of room tone at the head and tail of every take for edit fill.

### 4. Edit without clicks

Process in this order; verify each step by listening, not by assumption:

1. **De-click first.** iZotope RX de-click (or equivalent): start at low sensitivity, run multiple passes, and use the clicks-only output mode to verify you are removing clicks and not trigger transients. Never de-click a passage and assume the tingle survived.
2. **Spectral touch-ups.** For stubborn clicks, paint them out in a spectral view (Adobe Audition spot healing brush) or paste a few milliseconds of room tone from beside the click over it with short crossfades. Room-tone paste keeps the room's character; plugins color the sound.
3. **Gentle noise reduction only.** High-pass rumble below 100-150 Hz if needed; light denoise, never heavy. If you hear artifacts, you went too far; re-record rather than force a noisy take.
4. **Breaths and gaps.** Attenuate breaths to ~50% instead of deleting them; fill gaps with extracted room tone; 40 ms qsin crossfades at every edit so nothing chops.
5. **Loudness.** Normalize to the platform's target and check the whole piece on cheap earbuds too, because that is how most viewers listen.
- Full procedure with tool specifics: `references/editing-playbook.md`.

### 5. Package for discovery

Generate the package with the bundled CLI:

```bash
python3 bin/asmr_metadata.py session.json
```

The session JSON holds the primary trigger, secondary triggers, format (binaural, stereo), duration hint, goal, channel, and the chapter list. Output is a publish package: 3 title options (each within YouTube's 70-character truncation limit), a description with a hook in the first 150 characters, chapter timestamps, 3 to 5 hashtags, and a tag set held inside YouTube's 500-character tag limit.

Rules the CLI enforces and you should know:
- Title: max 70 characters including spaces; primary trigger keyword first.
- Description: the first 150 characters are the hook (visible in search); body 150-300 words, keywords placed naturally, never stuffed.
- Chapters: minimum 3 entries, first at `0:00`, each chapter title describing what the listener gets, not "part 1".
- Tags: 12-15 specific plus a few broad; tags matter less than title/description/thumbnail; misspellings of key terms capture typo traffic.
- Do not invent view counts, milestones, or engagement claims in the description.
- Title formulas and the full template: `references/metadata-playbook.md`.

## Output Contract

A finished session delivers:
1. The edited audio/video with no audible clicks, plosives, or digital-silence chops.
2. `package.md` from the CLI: chosen title, description with chapters and hashtags, tag list.
3. A chapter list that matches the actual trigger timeline.

## Operating Rules

1. Never invent specs, prices, review claims, or engagement stats in copy or metadata.
2. One trigger done well per session; do not stack five triggers thinly.
3. Prevention beats repair: quiet room and good gain staging before any plugin.
4. De-click is verified on the clicks-only output; if a trigger transient dies, redo the pass gentler or by hand.
5. No digital silence in the final piece; gaps are room tone with crossfades.
6. Every factual claim about gear or platform rules traces to a source; unverifiable claims get cut or framed as opinion.