Free API and REST API

Free API

No key and no account. Send a file to POST https://audo.ai/api/try/<tool> and get the result back. Each tool processes the first part of the file for free, a few times a day.

curl -F file=@interview.wav https://audo.ai/api/try/remove-noise -D response.headers -o response.body

The file-expiry periods below apply to Audo's copies. AI processing providers may retain data under their own terms; see the privacy policy. Audo expiry does not establish deletion timing at a downstream provider.

Inputs

• file, or url for a public HTTPS link, for tools that take one file. Send the body as multipart/form-data, or as JSON when every input is a url.
• Join takes file (or url) 2 to 20 times, in order.
• Swap audio takes video and audio (or video_url and audio_url).
• Align a script takes file, plus a text field or a script file (TXT, SRT, or VTT, up to 2 MB).
• Options are the tool's inputs, with the same names and defaults as the MCP tools. Send numbers and booleans as text, and arrays, such as cut's ranges, as JSON. IDs and idempotency_key aren't accepted, and free and max_credits need an API key.
• Transcription returns json, srt, vtt, or txt. Alignment returns json, srt, or vtt (no TXT output).
• A request takes up to 100 MB of files, with or without a key; without one, what its links fetch counts too. Links are fetched one at a time. Please send a User-Agent that names your agent.

Responses

The curl examples save response.headers and response.body, which are not yet media. Read the HTTP status and Content-Type before naming the result. A 200 JSON response can be a transcript, media report, or a job manifest with multiple output links. A 202 response is unfinished work. A 4xx or 5xx response is an error, even when a file was written.

For a complete single-file recipe, download and inspect the Python client. It uses Python 3 and ffprobe, handles synchronous media and text/JSON, polls only the returned HTTPS Audo job link, detects terminal failure, validates every downloaded output, and saves to a new directory. It refuses redirects and unrelated origins, and never executes returned commands or prints capability links. The in-memory example handles files below 100 MB; join, mix, replacement and script uploads need their documented multi-input layouts.

python3 free-api.py remove-noise interview.wav noise-results
# Keep only seconds 1 through 4 (mode=remove would delete them):
python3 free-api.py cut interview.wav cut-results '{"mode":"keep","ranges":[[1,4]]}'

Anonymous filler previews are review-only: to make automatic cuts, send the same original to remove-fillers with dry_run=false, without transcript_file_id or cut_ids. This transcribes again and uses another free use. To select preview cuts or reuse a transcript, sign in, upload the original on that account, and create a new preview through REST /api/v1/jobs or MCP remove_fillers. Anonymous IDs cannot be reused in that account workflow.

• The request waits up to about 30 seconds and returns the result itself: the audio or video file, the transcript, or the media info JSON.
• Send Prefer: respond-async to get 202 as soon as the upload is checked.
• A job that's still running returns 202 with a link to check on it: /api/try/jobs/<job_id>?t=<token>. The link works for 1 hour.
• RateLimit-Policy and RateLimit say how many free uses are left. Audo-Processed-Seconds says how much of the file was processed, and Audo-Full-Price-Credits gives the price of the whole file.
• When the free uses run out, the API returns 429 with JSON that says which limit and when it resets, and still takes the file: it waits for payment, and details has finishUrl (the person pays there, from $5, no sign-up, and the whole file runs), finishPriceCredits, jobId, and statusUrl, whose finishedBy names the whole-file job once paid. One address can have 3 waiting at once in an hour.
• A result cut short has a finish link. JSON results carry finishUrl and finishPriceCredits with a nextStep, and a file returned directly carries Audo-Finish-Url and Audo-Finish-Price-Credits. Give the link to the person: they pay there (from $5, no sign-up), and the whole file runs. Keep checking the job's status link: once paid, finishedBy names the whole-file job with its own status link, which gives the whole result. The link works while the free run's files are kept (an hour).

Whole files with an API key

Send an API key from the API page, and the same request runs on your account: omitting free processes the whole file with credits when the available balance covers it, otherwise the free portion if a free use is available. Credits are charged only if the job succeeds. Add an Idempotency-Key header (a new UUID for each request) to retry safely: the same key returns the same job. Uploads are still up to 100 MB; send a larger file with the REST API, which uploads in parts.

curl -H "Authorization: Bearer $AUDO_API_KEY" -F file=@interview.wav -F free=false -F max_credits="$APPROVED_MAX_CREDITS" https://audo.ai/api/try/remove-noise -D response.headers -o response.body

Set APPROVED_MAX_CREDITS to a positive integer explicitly approved by the person before using this paid example.

A disabled free tier is also unavailable. A positive max_credits caps paid cost (402 if exceeded); it does not prohibit spending. Use free=true to prohibit spending. Media info always runs free and counts against its daily quota. Report processedSeconds and free from the job, or Audo-Processed-Seconds for synchronous results, so the person knows if only a portion was processed.

Free limits

Uses are counted per tool, per UTC day. The website and the API share one count for each visitor.

Without a key, one address can send 2 uploads at once (an account, 4). Another upload while those are still arriving gets 429 with Retry-After; send it again once one finishes.

Every tool

Remove noise

curl -F file=@interview.wav https://audo.ai/api/try/remove-noise -D response.headers -o response.body

Enhance voice

curl -F file=@voice-note.m4a https://audo.ai/api/try/enhance-voice -D response.headers -o response.body

Transcribe

curl -F file=@interview.mp3 -F format=srt https://audo.ai/api/try/transcribe -D response.headers -o response.body

Align a script

curl -F file=@talk.mp3 -F script=@script.txt -F format=srt https://audo.ai/api/try/align-script -D response.headers -o response.body

Cut

curl -F file=@episode.mp3 -F mode=remove -F 'ranges=[[0,12.5]]' https://audo.ai/api/try/cut -D response.headers -o response.body

Remove silence

curl -F file=@episode.wav https://audo.ai/api/try/remove-silence -D response.headers -o response.body

Set loudness

curl -F file=@episode.wav -F target=podcast https://audo.ai/api/try/loudness -D response.headers -o response.body

Convert

curl -F file=@episode.wav -F format=mp3 https://audo.ai/api/try/convert -D response.headers -o response.body

Join files

curl -F file=@intro.mp3 -F file=@episode.mp3 -F file=@outro.mp3 https://audo.ai/api/try/join -D response.headers -o response.body

Swap a video's audio

curl -F video=@talk.mp4 -F audio=@talk-clean.wav https://audo.ai/api/try/swap-audio -D response.headers -o response.body

Media info

curl -F file=@interview.wav https://audo.ai/api/try/media-info -D response.headers -o response.body

Remove filler words

curl -F file=@interview.mp3 https://audo.ai/api/try/remove-fillers -D response.headers -o response.body

Level speakers

curl -F file=@podcast.wav https://audo.ai/api/try/level-speakers -D response.headers -o response.body

Mix tracks

curl -F file=@voice.wav -F file=@music.mp3 -F 'tracks=[{"role":"voice"},{"role":"music","volume_db":-6}]' -F duck=true https://audo.ai/api/try/mix -D response.headers -o response.body

REST API

The REST API uses the same paid/free choice matrix above: omitted free prefers paid whole files, then falls back to an available free portion. Explicit free=false requires credits for the whole file; use an approved positive max_credits. Its base address is https://audo.ai/api/v1. Send an API key from the API page as a bearer token:

curl https://audo.ai/api/v1/account -H "Authorization: Bearer $AUDO_API_KEY"

Run a tool

1. POST /api/v1/files with {"filename": "interview.wav", "sizeBytes": 48213000} returns the file and how to upload it: one link in upload (up to 100 MB), or, for files over 16 MB, multipart with the part size, a part link, and a complete link, all under one token.
2. Upload the bytes with PUT to upload.url. For parts, PUT part n (counting from 1) to multipart.partUrl with {n} replaced by n, then POST to multipart.completeUrl, which joins them or lists the parts to send again.
3. GET /api/v1/files/<file_id>?wait=30 until its status is ready. Its prices gives each tool's price for the file, in credits.
4. POST /api/v1/jobs with the tool and its inputs, in snake_case: {"tool": "remove_noise", "file_id": "file_8Kc2QmP4", "idempotency_key": "<a new UUID>", "free": true}. Add max_credits to cap the price, or "free": true for the free part.
5. GET /api/v1/jobs/<job_id> until its status is succeeded, failed, or cancelled. A job is charged only if it succeeds.
6. GET /api/v1/files/<output_file_id>/download returns a link that works for 15 minutes.

Endpoints

Errors

Errors are JSON: {"error": {"code": "...", "message": "...", "retryable": false, "nextAction": "..."}}. The message says what went wrong, and nextAction says what to do. Codes: bad_request, unauthorized, forbidden, not_found, conflict, idempotency_key_reused, insufficient_balance, over_spending_cap, free_limit_reached, free_budget_exhausted, free_disabled, file_not_ready, file_invalid, file_expired, file_too_large, file_too_long, unsupported_format, wrong_file_type, text_too_long, no_audio, ranges_outside, fetch_failed, limit_exceeded, upload_limit_reached, rate_limited, job_not_cancellable, tool_unavailable, build_mismatch, internal_error.

OpenAPI

The full description of the free API and the REST API is at https://audo.ai/api/openapi.json.