Room Tone Maker
Find stable non-speech ambience inside dialogue or location audio, choose the best region on a ranked waveform, and generate an exact-length, stereo-preserving repair fill with controlled seams.
RESTORATION MAP
Room Tone Patch
Separate the damaged event from healthy context. A repair is safest when the affected span is narrow, the surrounding signal is trustworthy and the difference audition contains mostly the defect.
Open the full signal breakdownBest source, processing logic, limits and three-pass listening check
- Dialogue ambience
Start with the right material
Filling an edit gap, smoothing a removed breath or word, building handles around dialogue cuts and preserving the recorded location instead of inserting generic synthetic noise.
- Find stable texture
Understand the transformation
Short-window level, zero-crossing and spectral-change features rank candidate regions; the chosen stereo segment is rearranged deterministically as overlapping micro-slices with crossfades and bounded output gain.
- Seamless room fill
Verify the usable result
Candidate-region audition, generated stereo room tone and a finite exact-duration WAV or MP3 fill that retains measured level and channel character from the source region.
Where this workflow stops
A region containing speech, traffic or other foreground events will reproduce their character, and very short candidates can sound repetitive. This does not remove noise, reconstruct missing words or create a calibrated acoustic model.
Make the final decision by ear
- Review every detected event near drums, consonants and other sharp attacks.
- Use Difference audition to hear what the repair removes or invents.
- Prefer several small repairs over one aggressive global pass.
- Candidate scan
- Ranks regions using short-window RMS, level variation, zero-crossing consistency, high-frequency difference ratio and spectral change.
- Natural fill
- Seeded micro-slice mosaics or crossfaded whole-region repeats create reproducible fills at an exact sample length.
- Stereo aware
- Shared slice positions and fades preserve left-right timing and mid-side relationships instead of randomizing channels independently.
How to create room tone for dialogue repair
- Load a dialogue or location recording, or start with the synthetic stereo production demo.
- Scan locally and listen through the ranked candidate regions highlighted on the waveform.
- Select or type a clean region, choose target duration, deterministic seed, mosaic or repeat mode and crossfade length.
- Generate, compare candidate, result and seam preview, then download exact WAV or MP3 audio.
How candidate ranking works
The scanner does not claim speech recognition. It measures short windows and favors sustained, low-variation ambience. Loud regions relative to the recording’s lower RMS distribution receive a speech-risk penalty. Regions with unstable zero crossings or rapidly changing high-frequency difference energy also rank lower. The candidate rail is evidence for listening, not an automatic editorial decision.
Mosaic versus seamless repeat
Mosaic mode uses one deterministic seed to choose many micro-slices inside the selected room-tone region. Every channel uses the same slice positions and overlap envelope. Seamless repeat mode uses the entire selected region in order. Both modes crossfade internal joins and report a seam preview, while the exact requested output frame count is never shortened by overlaps.
Different from noise generation and looping
Room Tone Maker derives ambience from a real production recording and helps locate non-speech source material. A noise generator synthesizes a new spectrum. Audio Loop Maker optimizes a user-chosen musical loop. Silence Remover and Breath Remover detect material to cut; this tool creates a replacement bed for an editorial gap.
Limits worth knowing
- Always audition candidates: quiet speech, distant voices, traffic changes or production sounds can fool simple stability metrics.
- The lightweight spectral profile uses time-domain difference energy rather than a certified forensic or dialogue classifier.
- Crossfades reduce boundary discontinuities but cannot guarantee an inaudible fill in every scene.
- MP3 can add codec delay; WAV preserves the exact generated sample count.
- Long recordings and fills are refused before rendering if the decoded-memory estimate exceeds the device safety budget.
Privacy: source audio is decoded, scanned, generated and auditioned locally. Audio payloads and the optional compact project file are not uploaded. Normal page assets may still be requested at page load.
Room Tone Maker: quick answer and technical limits
Quick answer: A dialogue-repair workspace that finds stable low-variation ambience in a recording and extends the selected texture into an exact-duration room-tone fill.
- Best for
- Filling an edit gap, smoothing a removed breath or word, building handles around dialogue cuts and preserving the recorded location instead of inserting generic synthetic noise.
- How it works
- Short-window level, zero-crossing and spectral-change features rank candidate regions; the chosen stereo segment is rearranged deterministically as overlapping micro-slices with crossfades and bounded output gain.
- What you get
- Candidate-region audition, generated stereo room tone and a finite exact-duration WAV or MP3 fill that retains measured level and channel character from the source region.
Know before you use it: A region containing speech, traffic or other foreground events will reproduce their character, and very short candidates can sound repetitive. This does not remove noise, reconstruct missing words or create a calibrated acoustic model.
Privacy: Selected audio is processed in this browser and is not uploaded to AudioWrench. Normal page assets can still be requested as described in the privacy notice. Privacy details →