Developer API: Free Audio Analysis and Mastering

Endpoints you can call without a key

Base address: https://echoteka.ai. Every endpoint below is public and anonymous.

Endpoint	What it does	Cost
`POST /api/v2/analyze`	Measures a file: integrated LUFS, peak, duration, sample rate and stereo correlation. With `extended=true` it also returns the spectrum, a loudness timeline and vectorscope points.	Free
`POST /api/v2/clean`	Removes the marks AI generators leave in the audio, without mastering. Loudness and tone stay as they are. Returns a job, like mastering.	Free within a small quota
`POST /api/v2/multipreview`	Masters a 24-second fragment from the middle of the track in 2 to 6 styles, so you can compare them by ear. Returns mp3 previews.	Free, 6 per day
`POST /api/v2/master`	Masters one track and returns a `job_id`.	One free full master a day, then credits
`POST /api/v2/batch`	Masters up to 10 files in one request with the same settings. `album_mode=1` keeps the loudness balance between tracks.	One free run, then one credit per file
`GET /api/master/status/{job_id}`	Progress from 0 to 100 and, once done, the loudness and true-peak figures.	Free
`GET /api/master/result/{job_id}`	Downloads the finished file after the status reports done.	Free
`GET /api/v2/limits`	What this caller can do right now: free masters left and when they reset, credit balance, upload cap.	Free
`GET /api/presets`	Target loudness in LUFS for each platform preset, as a map from name to value.	Free

Paid credit packs are on sale. The endpoints above are the only public ones.

Quick start with curl

Measure a track

The analysis answers at once, with no job to poll.

curl -X POST https://echoteka.ai/api/v2/analyze \
  -F "file=@track.wav"

Master a track

Mastering starts a job. Poll its status until it reports done, then download the file.

curl -X POST https://echoteka.ai/api/v2/master \
  -F "file=@track.wav" \
  -F "style=standard" \
  -F "target_lufs=-14" \
  -F "out_format=wav"
# -> {"job_id": "..."}

curl https://echoteka.ai/api/master/status/JOB_ID
curl -o master.wav https://echoteka.ai/api/master/result/JOB_ID

Python

import time
import requests

API = "https://echoteka.ai"

with open("track.wav", "rb") as f:
    r = requests.post(f"{API}/api/v2/master", files={"file": f},
                      data={"style": "standard", "target_lufs": -14, "out_format": "wav"})
r.raise_for_status()
job_id = r.json()["job_id"]

while True:
    s = requests.get(f"{API}/api/master/status/{job_id}").json()
    if s.get("error"):
        raise RuntimeError(s["error"])
    if s.get("status") == "done":
        break
    time.sleep(3)

audio = requests.get(f"{API}/api/master/result/{job_id}")
with open("master.wav", "wb") as out:
    out.write(audio.content)

print(s.get("after_lufs"), s.get("true_peak_db"))

Parameters

style: one of 24 names: standard, edm, hiphop, house_basic, classical, podcast, lofi, dry_vocal, trap, pop, rock, metal, rnb, country, afrobeats, kpop, jazz, voice, audiobook, warm, bright, bass, unlimiter, big_room. An unknown name is replaced by standard without an error, so check the spelling.
target_lufs: integrated loudness target in LUFS, for example -14 for streaming or -9 for club. The platform values are listed at /api/presets.
out_format: wav (default), wav24, mp3, flac, opus or aac.
files (batch): send each track as a repeated files field. The free tier takes up to 10 files per request.
album_mode (batch): 1 keeps the relative loudness of the tracks.
styles (multipreview): a JSON array of 2 to 6 style names.
extended (analyze): true adds the spectrum, the loudness timeline and the vectorscope points.

Send WAV or MP3 for the safest result. The server checks the contents of the file, not only its name.

What the status returns

While a job runs, GET /api/master/status/{job_id} returns progress (0 to 100) and a message. When it is done, the response also carries before_lufs, after_lufs, true_peak_db, lra, target_lufs and duration_sec. An error field explains a failed job.

Limits

Call GET /api/v2/limits before you send a job. It tells you how many free masters are left, when the count resets and what the upload cap is, so you do not hit a limit after the upload. The terms are on the terms page.

Results are not kept for long. Download the file as soon as the job is done.

Use it from an AI assistant

The same engine is wrapped as nine MCP tools: get_service_info, list_presets, check_limits, analyze_track, clean_ai_trace, preview_styles, master_track, batch_master and get_job. The connection steps for Claude Code, Claude Desktop, Cursor and Codex are on the MCP page.

Questions

Do I need an API key?

No, not for the endpoints above. The X-API-Key header is not available yet, and we have no date for it.

Is my audio used to train models?

No. Echoteka downloads the file only to process it. See the privacy policy for the details.

Where do I get the facts an assistant needs?

The short fact sheet for language models is at /llms.txt.

Who do I contact?

Write to support@echoteka.ai.

Developer API: analyze and master audio over HTTP