One API for state-of-the-art audio and image AI models.
Noise removal, speech enhancement, vocal separation, filler-word removal and background removal today, with more models and providers to come. One key, one prepaid balance, billed per second from $0.020 per audio minute.
- $1 free trial credit on sign-up
- Free test keys
- Top up from $10
- Billed per second
- Failed jobs cost nothing
# Remove background noise from a recording
curl https://api.zilapi.com/v1/audio/denoise \
-H "Authorization: Bearer $ZILAPI_KEY" \
-H "Content-Type: application/json" \
-d '{"input_url": "https://example.com/episode-142.wav",
"webhook_url": "https://your-app.example.com/hooks/zilapi"}'
# A signed job.succeeded webhook arrives. Fetch the MP3:
curl https://api.zilapi.com/v1/jobs/job_7Hq2xV9mKp4R/output \
-H "Authorization: Bearer $ZILAPI_KEY"{
"id": "job_7Hq2xV9mKp4R",
"task": "audio.denoise",
"model": "zilapi/nr-4.1",
"status": "succeeded",
"input": { "duration_seconds": 191.4 },
"output": { "format": "mp3", "bitrate_kbps": 320 },
"cost": { "charged_micros": 105600, "display": "$0.11" }
}Audio and image tools, one API.
Audio
Noise Removal 4.1
zilapi/nr-4.1Removes background noise, hum and room tone from speech and music while keeping the voice natural.
Speech Enhancement 1.0
zilapi/se-1.0Makes recorded speech clearer and closer: for interviews, calls, lectures and voice-overs.
Vocal Separation 2.5
zilapi/vms-2.5Splits a song or mix into a vocal track and an instrumental track.
Filler Removal 1.0
zilapi/filler-1.0Finds filled pauses (um, uh, er) in speech and cuts them out. Returns the edited audio as MP3 and the removed segments as JSON.
Image
One request in, one webhook out.
- 01
Send one request
The audio's URL, or the file itself up to 50 MB, plus where to send the result. Any common format, up to 6 h or 5 GB by URL. No SDK required.
POST /v1/audio/denoise - 02
Get a signed webhook
We fetch the file, run the model on our GPUs and POST job.succeeded to your webhook_url. Or poll the job.
job.succeeded - 03
Download the result
One MP3, or vocals and instrumental for Vocal Separation. The job shows exactly what you paid.
GET /v1/jobs/{id}/output
Pay for the audio and image you process. Nothing else.
No subscription and no seats. Add credit from $10 with a card through Stripe. Audio is billed per second; images per image. Failed and canceled jobs cost nothing. New accounts get $1 of free trial credit once the email is verified, and test keys are always free.
| Your file | At $0.033/minNoise Removal, Vocal Separation | At $0.043/minSpeech Enhancement |
|---|---|---|
| Voicemail, 0 m 45 s | $0.0248 | $0.0323 |
| Podcast clip, 3 m 12 s | $0.1056 | $0.1376 |
| Lecture, 48 m 30 s | $1.6005 | $2.0855 |
| Longest file, 6 h | $11.8800 | $15.4800 |
Built for the people who pay the bill.
Keys shown once, stored hashed
We keep a hash of each key, never the key. Revoke one and it stops working on the next request.
Prepaid, with a hard stop
Jobs only run against credit you added. A job never takes your balance below zero, so there is no surprise bill.
A receipt for every job
Each job records its billable seconds, whether the model’s minimum charge applied, the unit price and the total.
Card payments by Stripe
Your card details go to Stripe, not to us. Unused credit can be refunded within 14 days of the top-up.
Your audio is deleted on a schedule
Uploads are deleted 3 days after they are validated, results 7 days after the job succeeds. DELETE /v1/files/{id} removes an upload sooner.
Live and test keys kept apart
zil_test_ keys are free and return a sample file, so CI never spends real credit or touches a GPU.
Get a key in two minutes.
Sign up with email, Google or GitHub. New accounts get $1 of free trial credit for live jobs; test keys are always free.