Changelog

Platform updates and improvements. Latest: Engine v7.0 · 2026-10-25

2026-10-25

Your before/after video now shows up right after you post

On a computer, the before/after video for Shorts and Reels now starts making itself the moment your song lands on Mastered Charts, and plays at the top of the window.
One large button saves it as a 1080×1920 mp4 with sound. On phones it is one large button too — tap it and the video is ready in a few seconds.

2026-10-22

Turn your chart song into a before/after Shorts video

After you add a song to Mastered Charts, a new “Make before/after video” button turns it into a vertical video: your raw upload first, then the same bars mastered.
The video is made in your browser in a few seconds and saves as a 1080×1920 mp4 with sound, ready for Shorts, Reels and TikTok.
It shows your song title and artist name, and how much louder and fuller the master is.

2026-10-15

The AI-probability score now reads other generators too

Tracks made with generators other than Suno — Udio among them — now get a proper AI-probability reading. They could read low before.
Turning on Lower AI fingerprint moves the score further down than it used to, so you can see what the toggle did.
Songs made by people still read low, as before.

2026-10-15

Lower AI fingerprint works on more songs — and the score shows it

Lower AI fingerprint now reaches songs it used to leave untouched. It still depends on the song, as it always has, and it runs on every path: fast commit, premium, stems and our API.
The score on screen now comes from a detector that follows how outside services read a track. Before, it sat near the top whether the toggle was on or off, so you could not see what it did. Turning it on now moves the number.
Premium no longer stalls on some songs — the file arrives as it should.

2026-10-13

Stem Remaster: steady loudness on quieter mixes

On some quieter mixes, the final level step backed off more than it needed to, so the song came out about 0.8 dB quieter than intended. It now backs off only as much as the song needs.
Most songs are unchanged.

2026-10-12

Stem Remaster: cleaner highs on songs that turn rough easily

Before the full-mix texture pass, we now check how each song reacts. Songs whose highs would start to turn rough get a lighter pass, so the top end stays clean.
Other songs keep the current setting.

2026-10-11

Stem Remaster: a little louder, with more drive

Stem Remaster masters now come out about 0.5 dB louder, a touch above Fast Commit.
Songs that would start to flatten at the top are measured one by one and pushed a little less, so nothing gets squashed.

2026-10-11

Stem Remaster: one more texture pass on the full mix · Classical Fast Commit retuned

After filling in each instrument's texture, Stem Remaster now gives the combined mix one more pass, for a fuller sound.
Songs that already have a dense texture get less, so nothing turns harsh.
Fast Commit and Premium with the Classical preset now have less low-end boom and a little more presence, closer to Stem Remaster.

2026-10-11

Fast Commit & Premium: matched to Stem Remaster loudness, tighter low end

Fast Commit and Premium masters now come out at about the same loudness as Stem Remaster, about 1.5 dB lower than before.
Less squeezing means more room for drums and vocals to breathe.
The extra low-end boost in the Pop preset is gone, so the bottom end sounds tighter and less boomy.
The AI texture correction from v26.0 stays exactly as it was.

2026-10-11

Stem Remaster: AI artifact correction per instrument

Stem Remaster now fills in the thin, hollow AI texture on each separated part — vocals, drums, bass and the rest — rather than on the finished mix.
Each instrument keeps its own character, for a more natural, analog-like sound.
It still only runs on tracks identified as AI-made. Songs you recorded and mixed yourself are left as they were.
Loudness stays the same. Fast Commit and Premium keep the v26.0 correction.

★ Major update

2026-10-10

Enhanced AI artifact correction

Fast Commit, Premium and Stem Mastering now fill in the thin, hollow texture typical of AI tracks, for a fuller and more natural sound.
It only runs on tracks identified as AI-made. Songs you recorded and mixed yourself are left exactly as they were.
The amount is set per song. Tracks that already sound dense get less, so nothing turns harsh.
Loudness stays the same. You hear more body, not a louder file.

2026-10-10

Charts that keep score — 30-day ranking and a Hall of Fame

The live chart now ranks songs on the last 30 days of plays, likes and shares. There's no Monday reset anymore, so a song's score keeps building.
New Hall of Fame: when a song passes 30 days, it moves there with its score locked in and its best rank shown.
Every row now shows its score, and when two songs tie, the one with more plays goes first. A play, like or share moves the song up the list right away.
Share-link visits only count when they come from a listener's share.
Songs you chose to keep public stay on the chart without a gap, and the reminder email simply tells you the song stays up.
After you post a song to the chart from the studio, you get the same share-card panel as on the home page.

2026-10-02

Know how long a Stem Remaster will take — before you start

Stem Remaster runs on your own computer, so the time depends mostly on your graphics card. The studio now checks this computer and shows the estimate before you start.
Optimal: NVIDIA GTX 1660 / RTX 2060 or newer, or an Apple M1+ Mac · 16 GB or more memory · 8+ threads · latest Chrome or Safari — about 2–3 min for a 3:30 song.
Recommended: Intel / AMD integrated graphics · 8 GB or more memory · more than 4 cores · latest Chrome, Edge or Safari — about 7–8 min for a 3:30 song.
Not recommended: older cards without graphics acceleration (e.g. GT 710) — over 30 min for a 3:30 song. Under 8 GB of memory or 4 cores or fewer also makes it slow. The result sounds the same either way, and the full table is in the manual.
On slower computers, a Stem Remaster step that stalls partway is now caught and re-run automatically. You'll never get a file with a silent gap — if the result still isn't complete, the render stops and no credits are used.

2026-10-01

Louder masters — as loud as they go without breaking

Masters are louder again. Every song is pushed as loud as it can go cleanly: songs that already arrive loud stay where they are, quieter dynamic songs get a little more, and only songs that would start to distort are held back, just slightly.
Stem remasters bring back the top-end air the stem separation used to lose, up to your original's level.
Turning on AI fingerprint reduction no longer makes the master quieter or darker.

2026-09-30

Cleaner stems, and masters that keep your song's punch

In some songs, a faint buzz could sit behind the vocals in the quiet intro or ending of a stem remaster. It's gone.
Stem remasters now keep your original track's air and sense of space in the top end.
Quiet, dynamic songs are no longer pushed into the ceiling. When more loudness would flatten the peaks, mastering stops where the song stays close to its original punch.
Those songs may come out 1–3 dB quieter, with more punch. Songs that aren't being squashed are left exactly as before.

2026-09-29

Share links that open cleanly, and share buttons that respond

Chart share links are now short (antiaimaster.com/t/…), and “Copy link” copies just the link. Paste it into a browser or a chat and it opens the track card directly. Links you already shared keep working.
On a phone, the Instagram and TikTok buttons open the share sheet right away, so you can send the card straight to the app or save it. Inside in-app browsers, the card opens full screen so you can press and hold to save it.
Layout fixes on phones and smaller laptops: the home page no longer slides sideways, headers fit on one line, and the in-app browser notice is now in English too and no longer covers pop-ups.
A chart play now counts after 10 seconds of listening instead of 20. Clips are 30 seconds long, so this matches how streaming services count a play on a full song. Still once per listener per track per day.

2026-09-28

Masters get a little louder and keep their punch

Masters now come out about 2 dB louder, closer to how professionally mastered releases sit. The loudest peaks are shaped more gently before the final limiter, so the extra level doesn't flatten drums or make the mix pump. Classical and acoustic presets keep their current level and only get the gentler peak handling

2026-09-27

Tracks that arrive already squashed stay clean

Tracks that arrive already heavily limited — many recent Suno exports — now go through the mastering chain at the same working level as a regular track, so the master no longer adds distortion on top of what was already there. The difference is largest on dense, loud uploads; regular tracks sound the same

2026-09-25

Loud tracks keep their punch, and a credit for charting

Masters no longer push an already-loud track into the limiter a second time. Recent Suno exports come out much louder than before, and that second push was flattening them — masters now keep their punch and come out a little quieter. Streaming services turn loud masters down anyway, so nothing is lost there
Publish a track to the chart and get 1 free credit — once per track, up to 3 tracks a day. Reward credits can be used for 90 days
On phones, you now sign in before mastering. Your uploaded track is kept for 30 minutes and comes back after you sign in, so you don't have to pick it again
FLAC downloads now carry complete header information (frame sizes and checksum), which some car stereos and hardware players rely on
The AI scanner no longer guesses on tracks with almost no high-frequency content — it tells you it can't judge instead of showing a number
Charts: plays and likes you add no longer drop back after a refresh, and the share sheet shows the card before you post it

★ Major update

2026-09-11

Charts — share cards, comments, and tracks that can stay

Every chart track now has its own link. Paste it anywhere and it shows the cover, the title and the artist instead of a generic site card — and the person who opens it lands on that track with the RAW / MASTER buttons already open
Comments. Signed-in listeners can ask how a track was mastered or say what they liked, reply to each other, and react with emoji. The chart is a place to ask about settings, not a message board
If someone comments on your track, or replies to you, we email you — and the chart shows a badge so you can spot it while you are there
You decide what happens at 30 days. We email you before the deadline; the track stays up unless you ask us to take it down, so the questions and answers under it are not lost
Ranking now counts reach, not intent: tapping share is worth one play, and each person who actually arrives from your link is worth five
A play now counts once per listener per day instead of once ever, so people who come back to a track are counted
Fixed: the chart list was blank for a few hours on 11 September after a bad character slipped into the page script. We now block that class of mistake before a release goes out

2026-09-04

Reliability pass — nothing gets lost between render and download

If a download ever stalls, we now rebuild the file on our servers and send it automatically — a finished master can no longer be lost because your browser dropped it
Fixed occasional distortion on dynamic and ballad-style tracks; masters now sit right at the ceiling without that edge
WAV files now carry a fully standard header — some distributors were rejecting our float WAVs at upload
The in-app AI scanner no longer gets stuck on "analyzing", and it handles long or large files properly
Uploads now show the supported formats and size limit up front, and a file we can't read is caught immediately with a plain explanation instead of failing later
Premium downloads follow the format and bit depth you picked, and Stem Remaster recovers on its own on slower machines instead of giving up
Free and event credits now show their expiry date (valid 90 days from the day they land); purchased credits still never expire

★ Major update · ENGINE V6.0

2026-08-13

Engine v6.0 — Stem Remaster

Stem Remaster is now a full feature — your song is separated into vocals, drums, bass and instruments, rebalanced, then mastered. Mastering alone can't move the balance between instruments; this can
You get one finished master, not a folder of stems — the download set is the same as a normal master (WAV, lossless FLAC, MP3 320)
Charged when the render starts, and refunded in full if it fails or you cancel — you're never charged for something you didn't receive
Included in every subscription: Light gets 2 on signup, Standard 2 per cycle, Pro unlimited
Honest trade-off: the top end opens up and the stereo image widens, but the low-mids sit back a little — on sparse, already balanced songs the normal master is still cleaner. Run both and keep the one you like
New three-tier pricing in Korea — a ₩5,900 Light plan below the existing ₩12,900 tier, plus the ability to move down a tier instead of cancelling

★ Major update · ENGINE V5.5

2026-08-09

Engine v5.5 — reverse-engineered from professional masters

The mastering engine was rebuilt by reverse-engineering professional masters — we measured commercial releases against the mixes they came from, and matched the engine to what human engineers actually do: press the moment-to-moment dynamics hard, but leave the song's arc alone
Every track is treated on its own terms now — a song with heavy low end gets it cleaned up, and a song with a wide dynamic arc is pushed less so its chorus doesn't flatten out
The live preview finally sounds like the download — presence, air and level all line up, so the tone and width you set by ear come out the way you heard them
Anti-AI sounds like the rest of the presets — it used to arrive darker and quieter than the others; it now carries the same tone and the same level
Detection scores are higher than they used to be, though — the AI detectors themselves have gotten stronger recently, and codec-based evasion no longer moves them the way it did. We chose sound quality for this release, and we're researching a different approach to protection separately. Treat Anti-AI as "lowers the odds", not as a guarantee
Premium comes out at the standard level — it used to land about 3 dB quieter than a Fast Commit of the same song
One sound, whichever way your file is processed — the in-browser render, the server fallback and the API now produce the same master

2026-07-26

Anti-AI preset: highs restored

Anti-AI masters keep their top end — the airy 16 kHz-and-up range that used to get thrown away now survives, so cymbals, air and reverb tails stay open
More dynamics kept — the same loudness, but about 1 dB more crest, so transients breathe instead of flattening
Protection held up or improved — on our test set the detection score came out the same or lower than before (it varies by track and by detector, as always)
Nothing changes on older browsers — if your browser can't do the new processing, you get exactly the previous Anti-AI result, not a worse one
AI detector is more accurate — it now listens to a longer stretch of the track and is no longer fooled by tiny speed/pitch changes, so AI tracks that used to slip through are caught (12/12 on our test set, with no false alarms on human-made music)
AI tracks are recognised even when the file has no AI tag — re-encoded or re-exported AI songs get the AI-tuned processing instead of the general one

2026-07-11

Cleaner, more open Anti-AI masters

Clearer Anti-AI sound — the Anti-AI preset now opens up more, with a cleaner low end instead of the previous boxy feel
Per-song tonal cleanup — each track gets a gentle tonal-balance pass that tames muddy or boomy areas while keeping its character
Same protection — the detection-score reduction is still a side effect of the mastering (varies by track & detector); that part is unchanged

2026-06-21

Smarter genre recommendation

Smarter "Better sound" — the easy wizard now listens to your track and recommends the preset that best matches its genre, far more accurately than before
Handles in-between tracks — melodic dance, K-pop and other hard-to-place styles now get the right vibe instead of a mismatched preset
Still your call — the suggestion is just a starting point; switch to any other style in one tap

2026-06-17

Consistent loudness · clearer Premium files

Consistent loudness — every preset now lands at the same level, so switching presets no longer surprises you with a louder or quieter master
Preview matches the download — the level you hear in the live preview now lines up with the file you get
Cleaner on loud inputs — hot or squashed tracks no longer pump or get over-squeezed
Clearer Premium downloads — Premium files now carry "Premium" in the filename, with the 96 kHz master included

2026-06-13

Smoother finish · more guides

Even smoother finish — we tightened the final result step so your master comes through cleanly on a wider range of setups
More mastering guides — fresh how-tos for AI music creators are now live on the blog

2026-06-13

Smooth on more machines

Finishes every time — if mastering used to stall at the final step on some computer setups, our server now takes over automatically and completes it reliably
Clear status — when stable mode kicks in, you always see what's happening

2026-06-10

Steadier sessions · stronger Match-EQ consistency

Rock-steady back-to-back mastering — fixed the cases where the page could jump back or drop out while processing many tracks in a row
Stronger Match-EQ consistency — the same track with the same settings always delivers the same master, every time
Instant commit — switch presets and master right away, with no waiting

2026-06-04

See it in action — demo videos on the home page

New demo section on the home page — watch a full AI track go from raw to mastered, plus the one-click Wizard, before/after A/B, AI-detection score drop, and automatic genre matching
Tap any clip (or its speaker icon) to hear the sound — only one plays at a time, so nothing overlaps
Clips follow your language — Korean or English narration switches with the site toggle
Lightweight by design — videos only play while on screen, so they don't drain mobile data or battery

2026-06-03

Sample try-out, clearer results & refund terms

Try before you commit — master a bundled sample track for free, hear before/after, and see the AI-detection score (no charge, no download needed)
After the sample, one tap takes you straight to mastering your own track
Anti-AI goal now leads with the score change (before → after, e.g. 99% → 45%) right at the top of your result
Clearer refund terms across the site — full refund within 7 days as long as you haven't downloaded a master yet (downloading uses the credit)

2026-06-01

Easy Wizard mode — guided one-click mastering

New Wizard mode (now the default) — master in four simple steps: upload → pick a goal → master → see your result
Pick your goal in plain language: 'Better sound' or 'Lower AI-detection score' — we set everything up for you, no knobs to learn
The result shows the AI-detection score before and after, side by side, so you can see the change at a glance
Full Pro mode (all knobs + presets) is always one tap away via the top-right toggle — and your choice is remembered
Cleaner downloads — files from Wizard mode are named simply, without internal preset codes

2026-05-26

Self-serve account · free re-mastering · landing auth widget

New /account page — credits, subscription, and full payment / redeem history in one place (KO + EN)
Self-serve subscription cancel — from the studio dropdown or /account; access kept until period end, no immediate revoke
Free re-mastering — the same file with the same preset never charges twice (privacy-safe; we don't store your audio on the server)
Landing top nav matches the studio: Google sign-in pill ↔ avatar dropdown with redeem code, account, pricing, help, support, logout
Smarter checkout — sign in mid-flow and you're sent right back to the payment page (no plan re-selection)
Quality fixes: Pro Delivery re-download is now free, mastering count shows the exact total (e.g. 237), logos always link home, mobile admin layout fixed, and more

2026-05-21

Blog launch · multi-method sign-in · in-app browser support

Launched /blog — original mastering insights for AI music creators, bilingual (KO + EN), new posts weekly
Email magic-link sign-in — works without a Google account, and reliably inside KakaoTalk / Instagram in-app browsers
In-app browser banner — one-click 'Open in Chrome / Safari' when you arrive from KakaoTalk or Instagram (Google login is blocked in those webviews)
Korean pricing simplified — Light ₩12,900 / month · Pro ₩29,000 / month (clearer than the previous credit model)
AI-search optimized — discoverable by Gemini, ChatGPT, and other AI tools (Organization + publisher metadata)

2026-05-17

Bilingual site · loudness tuning · sample-track demo

Full Korean / English bilingual support across the entire site
Loudness engine tuned — resolved limiter pumping in loud pop sections
Replaced the Phase control with Tone Balance (tilt EQ) for more intuitive shaping
Credit plans and a redeem-code system introduced (Try Pack and more)
New: 'Try with a sample track' — experience a full master instantly, no upload needed
Smoother path from applying a redeem code to your first master
Email notifications fully redesigned to match the studio's look
Korean payment method coming soon (KakaoPay)

2026-05-08

Accurate scores · smooth detection · UX improvements

AI Detector upgraded from v17b to v18 — Korean vocal false-positive rate reduced from 70% to 0%
Anti-AI processing now produces a smooth score drop (previously swung between 0% and 99%)
Score display capped at 5–99% — no detector can be 100% certain; we reflect that honestly
Detection speed improved 16× (40 s → 2.5 s per track)
Before/After comparison now shows an Anti-AI effect badge (e.g. ↓ −15 pt)
Preset label shown dynamically in the After panel (e.g. 'After · Hip-Hop')
Fixed a cache bug where re-uploading a different track could download the previous master
Eliminated digital click at the start and end of mastered files (stronger fade)
Temp files cleaned up automatically — your audio never persists on disk after processing
Added a re-select file button — switch tracks without a full page reload
Landing page copy made more honest — changed '96%→14%' single-case claim to '15–50% range'
Changelog page launched (/changelog) — cumulative update history