---
name: reels-room-sound
description: "Make a generated talking head sound recorded in a real room — H3 added no ambience in the source test, so mix your own or licensed room/street ambience under the voice with a short room reverb in ffmpeg. Distinguishes the author's later sound-source preference from the historical ElevenLabs Sound Effects test. Part of the reels reference set."
---

These creative references describe source experiments with external tools. Use your own supported account and tools; this download installs no automation. Settings and timings are experimental observations, not current FlashBoards availability. Use only media and voices you have permission to use.

# reels-room-sound · room tone, street, reverb

> **Source author's sound preference (9.10.2026), not FlashBoards policy:** avoid ElevenLabs Sound Effects (`/v1/sound-generation`, `fal-ai/elevenlabs/sound-effects`) for effects, room tone, ambience and jingles; the author found them strange and louder than the voices and preferred ElevenLabs for voices (TTS v4, Speech-to-Speech, Audio Isolation). Historical demonstrations below and in related guides still include ElevenLabs Music beds and an earlier Sound Effects ambience test; those are separate experiments, not a universal voices-only rule. For your own mix, use permissioned film sound (line-free gaps, Demucs without vocals), the video model's native sound or supported video-to-audio tools; keep effects under the voice.

Measured 9.10.2026: asking H3 Max for "the reverb of the room, a car passing, voices from the street" left the silence level unchanged (10th-percentile RMS 0.0014 vs 0.0015 for the clean line) — nothing was added.

## Steps

1. Obtain ambience of {place} from your own recording, licensed footage, the video model, or a supported video-to-audio tool, following the source preference above. Match its length to the clip in your editor; use only settings supported by the chosen tool. A reusable sound brief is:
   `"<the sound of the place, heard from where the character sits>: <2–3 soft background sounds>, <one event>, a small room tone. No music."`
   Example value (a bedroom with a window on a street): `"Quiet evening street heard through a closed bedroom window: soft distant traffic hum, one car passing by slowly, faint voices of people far away, a small room tone. No music."`
2. Mix (light ffmpeg, video copied, not re-encoded):

```bash
ffmpeg -i clip.mp4 -i ambience.mp3 -filter_complex "[0:a]aecho=0.8:0.6:28|47:0.18|0.12[v];[1:a]volume=0.22,apad[s];[v][s]amix=inputs=2:duration=first:normalize=0[a]" -map 0:v -map "[a]" -c:v copy -c:a aac -b:a 160k -movflags +faststart out.mp4
```

The source author approved this mix level in that experiment: ambience ≈ −13 dB under the voice, a short two-tap room echo. Listen to your own mix and adjust it for intelligible speech; no source-author approval is needed.

## Historical ElevenLabs ambience test (not the recommendation above)

The source's earlier test used ElevenLabs `POST /v1/sound-generation` (or `fal-ai/elevenlabs/sound-effects/v2`) with the same prompt structure, `duration_seconds` = clip length and `prompt_influence` = 0.5; generation took **3.4 s**. Those parameter names and that timing belong only to this ElevenLabs observation. They do not apply to recordings, licensed footage or other tools, and the later source preference does not erase that measured test.

## Example run (reference only — never reuse its files, prompts, voices or images for a person)

- Demo character Mara Vale, «Mara's Rules» ep. 1: the historical ElevenLabs street prompt above → mixed into `h3max_room_light_street.mp4`. No example ambience or clip is delivered or required; `clip.mp4` and `ambience.mp3` in the command mean your own inputs, and `out.mp4` is your output.