Compress audio

Create an asynchronous audio compression job from an uploaded .mp3 or .wav file. Audio always runs queued: the endpoint responds with HTTP 202 Accepted and a job object { id, status }.

Request

string

required

Bearer sk_live_...

string

Optional, 1–255 characters. Returns the existing job if a job with this key was already created.

string

required

Id of an upload, must be an audio file (.mp3 or .wav).

string

One of chill, medium, aggressive, podcast, lecture. Either preset or operations is required.

array

1–10 custom operations, applied in order. See below.

string

URL to receive job.completed / job.failed events for this job. Requires a plan with webhooks.

Presets

Preset	Pipeline	Output
`chill`	trim-silence (0.2) → normalize (-16)	Opus 32k: near-transparent
`medium`	trim-silence (0.5) → normalize (-14) → compress (3:1)	Opus 24k: balanced
`aggressive`	trim-silence (0.8) → speedup (1.75x) → normalize (-10) → compress (8:1)	Opus 12k: max size reduction
`podcast`	trim-silence (0.6) → normalize (-16) → compress (4:1)	Opus 24k: voice-optimized
`lecture`	trim-silence (0.4) → speedup (1.5x) → normalize (-14)	Opus 16k

Operations

Type	Params (defaults)	Description
`trim-silence`	`aggressiveness` 0–1 (0.5), `minSilenceDuration` 100–5000 ms (500)	Remove silent parts
`normalize`	`targetLevel` -70–0 dB (-14), `peakNormalize` (false)	Normalize volume levels
`compress`	`ratio` 1–20 (4), `threshold` -60–0 (-20), `attack` 0–100 (10), `release` 0–1000 (100)	Dynamic range compression
`speedup`	`rate` 0.5–5.0 (1.25)	Change speed without changing pitch
`encode`	`format` `opus`\|`mp3` (opus), `bitrate` 12–128 kbps (24), `channels` 1–2 (1)	Output encoding

Encoding rules:

If your chain has no encode operation, one is appended with defaults (Opus, 24 kbps, mono).
encode always runs last, regardless of where you place it. If several are sent, the last one wins.
Output is .ogg (Opus) by default, the best quality per bitrate for voice. Pass {"type": "encode", "params": {"format": "mp3"}} for .mp3 when you need broad compatibility.

GET /v1/audio/presets and GET /v1/audio/operations return these definitions as JSON.

Examples

Preset

curl -s https://api.robinzip.app/v1/audio \
  -H "Authorization: Bearer $ROBIN_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: episode-42" \
  -d '{"audioId": "665f1c2ab8d4e91a2c3d4e5f", "preset": "podcast"}'

Custom operations, MP3 output

curl -s https://api.robinzip.app/v1/audio \
  -H "Authorization: Bearer $ROBIN_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "audioId": "665f1c2ab8d4e91a2c3d4e5f",
    "operations": [
      {"type": "trim-silence", "params": {"aggressiveness": 0.7}},
      {"type": "normalize"},
      {"type": "encode", "params": {"format": "mp3", "bitrate": 64, "channels": 2}}
    ]
  }'

{
  "success": true,
  "data": {
    "id": "665f1d00b8d4e91a2c3d4e60",
    "status": "created"
  }
}

Results

Poll GET /v1/jobs/:id (ideally with ?wait=25 to long-poll) until status is completed, then download result.outputUrl (signed, 1-hour validity, regenerated on every read). Results are retained for 7 days.

Credits

1 credit per started 5 MB of input, reserved when the job is created and refunded if it ultimately fails. See Credits.

Getting Started

Core Concepts

API Reference

Compress audio

Request

Presets

Operations

Examples

Results

Credits

​Request

​Presets

​Operations

​Examples

​Results

​Credits

Request

Presets

Operations

Examples

Results

Credits