TxT Sanitizer
TxT Sanitizer

Role
Scope and development
Team
Me
Platform
Web App
Timeline
2 weeks
project overview
Around 90 people on my team cleaned client text the same way every day: open Notepad, run the same find-and-replace pairs, hope nothing was missed. TxT Sanitizer is a free browser tool that applies saved replacement rules in one pass, and nothing leaves your browser. Roughly 50 of those colleagues use it regularly (my estimate), and I rebuilt it as v2 after their feedback.
Around 90 people on my team cleaned client text the same way every day: open Notepad, run the same find-and-replace pairs, hope nothing was missed. TxT Sanitizer is a free browser tool that applies saved replacement rules in one pass, and nothing leaves your browser. Roughly 50 of those colleagues use it regularly (my estimate), and I rebuilt it as v2 after their feedback.
What I built, and why it changed
V1 did one job: apply saved replacement rules and clean up basic formatting. I left out accounts and cloud sync on purpose. Keeping everything in the browser means client text never reaches a server, which made the tool easier to trust. The cost is no sync across devices, and I accepted that.
V2 came straight from colleague feedback: tabs for several texts, auto-clean with no button to press, and a choice of 14 themes. I also fixed the history, which didn't save reliably in v1, and added a restricted-words preset. The rebuild took 10 days.



Colleagues said V1 didn't look modern or professional, so V2 ships 14 themes. I checked contrast on some of them.

Results
30 s → 5 s per cleaning job (83% less time), a rough estimate from about 60 colleagues.
3.49k visits in the last 30 days (Cloudflare), 96% on desktop, 77% from one market (Bangladesh).
682 Google clicks in 3 months (Search Console), about 230 a month. Growth still comes mostly from word of mouth. I don't track unique or returning users yet.

Risks and ethics
The restricted-words preset helps freelancers spot words marketplace filters often flag, including false alarms. It could also be misused to dodge platform rules. The tool only finds and rewrites flagged words, makes no approval promises, and automates nothing on any platform.
If a client had brought me this brief
"We spend too long cleaning text. Can you make a tool?" My first questions:
How many texts a day, and who does it? (Sizes the usage.)
Does the word list change per client? (Decides shared vs personal rules.)
What does a missed word cost? (Sets how much testing it needs.)
Do people need accounts or sharing, or can client text never be stored? (The biggest scope jump.)
What I built, and why it changed
V1 did one job: apply saved replacement rules and clean up basic formatting. I left out accounts and cloud sync on purpose. Keeping everything in the browser means client text never reaches a server, which made the tool easier to trust. The cost is no sync across devices, and I accepted that.
V2 came straight from colleague feedback: tabs for several texts, auto-clean with no button to press, and a choice of 14 themes. I also fixed the history, which didn't save reliably in v1, and added a restricted-words preset. The rebuild took 10 days.



Colleagues said V1 didn't look modern or professional, so V2 ships 14 themes. I checked contrast on some of them.

Results
30 s → 5 s per cleaning job (83% less time), a rough estimate from about 60 colleagues.
3.49k visits in the last 30 days (Cloudflare), 96% on desktop, 77% from one market (Bangladesh).
682 Google clicks in 3 months (Search Console), about 230 a month. Growth still comes mostly from word of mouth. I don't track unique or returning users yet.

Risks and ethics
The restricted-words preset helps freelancers spot words marketplace filters often flag, including false alarms. It could also be misused to dodge platform rules. The tool only finds and rewrites flagged words, makes no approval promises, and automates nothing on any platform.
If a client had brought me this brief
"We spend too long cleaning text. Can you make a tool?" My first questions:
How many texts a day, and who does it? (Sizes the usage.)
Does the word list change per client? (Decides shared vs personal rules.)
What does a missed word cost? (Sets how much testing it needs.)
Do people need accounts or sharing, or can client text never be stored? (The biggest scope jump.)