Silence Remover API

POST /silence_process

About this tool

Long interviews and classroom captures ship faster when dead air disappears automatically. `/silence_process` ingests JSON with `video_url` plus `noise_threshold` parameters aligning with FFmpeg’s silence-detection idioms—energies below the threshold collapse after minimum gaps defined in docs. Outputs land in the ordinary async workflow: validate subscription keys, enqueue FFmpeg, poll task metadata, fetch trimmed media—all without multipart uploads bloating gateways. Archival librarians still annotate which passes were algorithmic versus hand-tuned metadata for FOIA responsiveness.

Threshold dialing is iterative: HVAC noise fights naive gates, while aggressive slicing can butcher comedic timing. Editors often preview locally using raw ffmpeg graphs, tune `noise_threshold` empirically, then lift those decimals into queued JSON batches that resume automatically when browsers close. Stitch `/cut` jobs ahead of silence detection when countdown leaders must disappear before metering speech energy.

Unlike consumer “jump cut bots,” this endpoint documents failure semantics when tracks lack audio entirely or when Dolby layouts confuse channel mapping. Logs cite channel indexes so engineers can remap before requeueing. Observability metrics quantify how many seconds each job removes so finance can compare storage savings versus manual editor hours. When speech overlaps room tone, raise thresholds slightly rather than chasing impossible zero-cross perfection.

Enterprise compliance still demands human review for legal depositions—auto trimming remains an editorial assistant rather than sworn testimony. Understand which retention tiers apply before betting on long-lived download mirrors; subscriptions define how aggressively cold storage rotates even when your JSON payloads stay identical.

Throughout, JSON plus HTTPS remains the contract: host files on durable TLS endpoints, never embed secrets in URLs, and throttle cron jobs with exponential backoff when queues show saturation signals. Finance podcasts adore `/silence_process` until deliberate dramatic pauses disappear—export A/B FFmpeg settings locally, annotate runbooks, then promote thresholds that preserve performer beats while shortening wall-clock listens. Remember silent B-roll cues still qualify as programmatic audio; annotate those chapters before automation runs.

Try it now

How it works

  1. Profile your room tone

    Measure typical noise floors, translate them into `noise_threshold` values documented alongside `/silence_process`, and ensure `video_url` is public HTTPS.

  2. POST JSON with API key

    Submit the payload through `POST /silence_process`. Receive `status_url` immediately while FFmpeg schedules work.

  3. Poll until FFmpeg finishes

    Use periodic GET requests to track progress; UI components can reflect percentage estimates if exposed.

  4. QC the tightened edit

    Download the processed file, listen for clipped consonants, and rerun with gentler thresholds if needed.

Frequently asked questions

Will music with rests get butchered?

Silence detection cannot distinguish artistic pauses from mistakes. Use higher thresholds or preprocess stems before calling the API.

Does it output the same container?

Consult docs for default mux settings; some jobs re-encode audio while copying video or vice versa depending on filter needs.

What if `noise_threshold` is invalid?

Validation errors return before compute starts. Adjust based on float ranges published in the reference.

Why must sources be HTTPS URLs?

Workers pull bytes without your cookies. HTTP or auth-gated links fail fast to protect queue integrity.

How do subscriptions affect batching?

Higher tiers raise concurrent silence jobs—critical when ingest pipelines fire thousands of lecture captures nightly.

Is multipart uploads supported?

No. The Droid Apps FFmpeg suite standardizes JSON with remote HTTPS references for operational consistency.