In August 2018, musical.ly folded into TikTok and about 100 million accounts changed platform overnight. A lot of people discovered that week that videos they had made in one app were now sitting in another one, in a slightly different form, with no obvious way to get a copy onto their own phone.
That problem has never gone away. It has changed shape roughly every eighteen months, and each shift left behind a layer of advice that is now wrong. Which is why searching for how to save a TikTok returns instructions from four different eras, most of them describing buttons that no longer exist.
- 2016 to 2018, the era with no watermark
- 2019, the watermark becomes strategy
- 2020, bans, mirrors, and a mess of duplicate sites
- 2021, the in app toggle and its limits
- 2022, editing moves in and quality starts to matter
- 2023, platforms start penalising the watermark
- 2024 to 2025, audio rights and photo mode
- Ranked for where things stand now
- Where that ranking is not fair
- The rights question, which the timeline keeps making sharper
- What the next shift probably looks like
Here is the actual sequence, and where it leaves things in 2026.
2016 to 2018, the era with no watermark
Early musical.ly saves came out clean. There was a download option, it produced a file, and the file had nothing stamped on it.
That was not generosity, it was an app that had not yet worked out that redistribution was free marketing. Clips spread to other platforms with no branding, which built audiences elsewhere and gave the original app nothing.
2019, the watermark becomes strategy
TikTok introduced the moving watermark: the handle and the logo, drifting between corners so a static crop cannot remove them.
The bouncing behaviour is the clever part. A fixed corner mark can be cropped or covered. One that migrates means any crop tight enough to lose it destroys the framing. The effect was immediate. Reposted TikToks became instantly recognisable on Instagram and YouTube, which was exactly the intention.
This is also the moment the third party save tool became a category, because a watermark that cannot be cropped creates demand for a version that never had one.
2020, bans, mirrors, and a mess of duplicate sites
The Indian ban in June 2020 removed the app from its largest market and pushed enormous volumes of traffic to browser based alternatives. Dozens of near identical sites appeared, most of them the same underlying service behind different names, all of them monetised aggressively.
That year produced the pattern people still complain about: pages where the actual link is grey and small, and three green button shaped adverts surround it. It was an arbitrage moment and the sites built for it were built for arbitrage rather than for use.
2021, the in app toggle and its limits
TikTok added a per video download switch, off by default for some accounts, on for others, and controllable by the creator.
It was a real improvement and it solved less than it appeared to. The file it produces carries the watermark, the creator can disable it entirely, and it does nothing for a video you want to reference in a presentation rather than repost. Advice written in this period usually says just use the save button, which fails silently the moment a creator has switched it off.
2022, editing moves in and quality starts to matter
CapCut’s rise changed what people were saving videos for. No longer just archiving, but recutting, remixing, and using clips as source material inside other edits.
The moment a file becomes an edit input rather than a keepsake, compression stops being cosmetic. A watermarked, re-encoded copy dropped into an editing timeline and exported again is two generations down before anyone has made a creative decision. This is the period where the difference between a tool that passes the file through and one that recompresses it became the whole story.
2023, platforms start penalising the watermark
Instagram and YouTube both began deprioritising content carrying a competitor’s watermark in their short form feeds.
Suddenly a watermarked repost was visually obvious and measurably worse performing at the same time. Cross posting creators went from mildly preferring clean files to needing them, and the demand shifted from casual users to people doing this professionally, several times a day, at volume.
2024 to 2025, audio rights and photo mode
Two smaller shifts with practical consequences.
Music licensing tightened, so a saved video’s audio track may be licensed for playback inside the app and not for redistribution. The file saves fine. Whether the sound on it can legally appear in your own upload is a separate question, and the answer is often no.
Then photo mode arrived: multi image posts with a soundtrack, which are not videos at all. Plenty of tools handled them badly for months, returning either one image or a broken file, because everything had been built assuming a single video track.
Ranked for where things stand now
Judged on 2026 conditions rather than 2020 ones: clean output, no re-encode, handles photo mode, and holds up across a batch.
- savett, returns the file without a watermark or a second encode, and behaves the same on the tenth save as the first
- ssstik, reliable and long running, with more page furniture between the paste and the file than is comfortable at volume
- snaptik, does the job, occasionally normalises longer clips to a smaller size
- musicaldown, the oldest name on this list and the one whose interface most clearly belongs to an earlier era
For a straightforward clean save with nothing installed, the savett route is the shortest version of this in 2026, and it handles photo mode posts rather than choking on them.
Where that ranking is not fair
This list measures one narrow thing: getting an intact, unbranded file quickly and repeatedly.
musicaldown sits at the bottom of it while being the tool with the longest continuous track record here, which counts for something in a category where services vanish without warning. snaptik’s page loads faster than anything else in the group, and for a single clip in a hurry that is the metric that matters.
None of them is a bad product. They are tuned for different moments in the history above, and some of them are still tuned for 2021.
The rights question, which the timeline keeps making sharper
Every technical step in this decade made saving easier, and none of them changed who owns the video.
A creator’s clip is theirs. A copy for reference, research or private viewing is unremarkable. Reuploading it to your own account, even with the handle in the caption, is republishing somebody else’s work, and the fact that the watermark is gone makes it less obvious rather than more permitted. If the audio is licensed music, that is a second layer of exposure on top of the first.
The practical rule has not moved in ten years: use your own material, or ask.
What the next shift probably looks like
The direction of travel is towards provenance metadata rather than visible marks. Invisible signatures embedded in the file, surviving crops and re-encodes, letting a platform identify the origin of a clip without a logo bouncing around the frame.
If that lands, the watermark question becomes irrelevant and the attribution question becomes unavoidable. Which, given that attribution was always the actual issue, would be a reasonable place for the decade to end up.
