<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Smart Online Tools</title>
    <description>The latest articles on DEV Community by Smart Online Tools (@smartonlinetools).</description>
    <link>https://dev.to/smartonlinetools</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4252692%2Ff2097eae-49f7-4b95-b922-687a730251aa.png</url>
      <title>DEV Community: Smart Online Tools</title>
      <link>https://dev.to/smartonlinetools</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/smartonlinetools"/>
    <language>en</language>
    <item>
      <title>How I built 80+ free browser tools, including AI that never uploads your files</title>
      <dc:creator>Smart Online Tools</dc:creator>
      <pubDate>Wed, 18 Nov 2026 23:26:47 +0000</pubDate>
      <link>https://dev.to/smartonlinetools/how-i-built-80-free-browser-tools-including-ai-that-never-uploads-your-files-1f8h</link>
      <guid>https://dev.to/smartonlinetools/how-i-built-80-free-browser-tools-including-ai-that-never-uploads-your-files-1f8h</guid>
      <description>&lt;p&gt;I wanted a set of everyday tools — merge a PDF, compress an image, read text from a screenshot — that I could use without uploading my files to someone else's server. So I built &lt;a href="https://smartonline.tools" rel="noopener noreferrer"&gt;Smart Online Tools&lt;/a&gt;: 80+ free tools where all the processing happens in the browser.&lt;/p&gt;

&lt;p&gt;The PDF and image tools were the easy part. The interesting part was the AI tools: OCR, speech-to-text, translation, background removal, face blurring and alt-text writing, all running &lt;strong&gt;on the visitor's device&lt;/strong&gt;. Here's how that works and what I learned.&lt;/p&gt;

&lt;h2&gt;
  
  
  The stack
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Next.js (App Router) as a static export.&lt;/strong&gt; No backend at all. The whole site is HTML, JS and WebAssembly served from ordinary shared hosting.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;&lt;a href="https://huggingface.co/docs/transformers.js" rel="noopener noreferrer"&gt;Transformers.js&lt;/a&gt;&lt;/strong&gt; to run ONNX models in the browser, using WebAssembly, or WebGPU where the browser supports it.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;&lt;a href="https://github.com/naptha/tesseract.js" rel="noopener noreferrer"&gt;Tesseract.js&lt;/a&gt;&lt;/strong&gt; for OCR.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;&lt;a href="https://ai.google.dev/edge/mediapipe/solutions/vision/face_detector" rel="noopener noreferrer"&gt;MediaPipe Tasks Vision&lt;/a&gt;&lt;/strong&gt; for face detection.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Web Workers&lt;/strong&gt;, so the page never freezes while a model is working.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Which model does what
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Tool&lt;/th&gt;
&lt;th&gt;Model&lt;/th&gt;
&lt;th&gt;First download&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Image to text (OCR)&lt;/td&gt;
&lt;td&gt;Tesseract&lt;/td&gt;
&lt;td&gt;2–15 MB per language&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Speech to text&lt;/td&gt;
&lt;td&gt;Whisper base&lt;/td&gt;
&lt;td&gt;≈ 80 MB&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Translator&lt;/td&gt;
&lt;td&gt;OPUS-MT (Helsinki-NLP)&lt;/td&gt;
&lt;td&gt;≈ 80–110 MB per direction&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Background remover&lt;/td&gt;
&lt;td&gt;MODNet&lt;/td&gt;
&lt;td&gt;≈ 26 MB&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Blur faces&lt;/td&gt;
&lt;td&gt;MediaPipe BlazeFace&lt;/td&gt;
&lt;td&gt;&amp;lt; 1 MB&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Alt-text generator&lt;/td&gt;
&lt;td&gt;Florence-2 base&lt;/td&gt;
&lt;td&gt;≈ 250–450 MB&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Every model has a permissive licence (MIT, Apache-2.0 or CC BY 4.0). I deliberately skipped a few good models because of their licences. For example, one popular text-to-speech model depends on a GPL phonemizer. For text to speech I use the browser's built-in device voices instead.&lt;/p&gt;

&lt;h2&gt;
  
  
  One worker, loaded on demand
&lt;/h2&gt;

&lt;p&gt;Each model downloads only when someone actually clicks the tool. A single worker keeps loaded models in a cache, so the second run is instant:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight typescript"&gt;&lt;code&gt;&lt;span class="k"&gt;import&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;env&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;pipeline&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt; &lt;span class="k"&gt;from&lt;/span&gt; &lt;span class="s2"&gt;"@huggingface/transformers"&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;

&lt;span class="n"&gt;env&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;allowLocalModels&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="kc"&gt;false&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;span class="n"&gt;env&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;useBrowserCache&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="c1"&gt;// the model is downloaded once, then reused&lt;/span&gt;

&lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;cache&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="k"&gt;new&lt;/span&gt; &lt;span class="nb"&gt;Map&lt;/span&gt;&lt;span class="o"&gt;&amp;lt;&lt;/span&gt;&lt;span class="kt"&gt;string&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nb"&gt;Promise&lt;/span&gt;&lt;span class="o"&gt;&amp;lt;&lt;/span&gt;&lt;span class="kt"&gt;any&lt;/span&gt;&lt;span class="o"&gt;&amp;gt;&amp;gt;&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;

&lt;span class="kd"&gt;function&lt;/span&gt; &lt;span class="nf"&gt;loadTranslator&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;pair&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kt"&gt;string&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
  &lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;key&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="sb"&gt;`translate:${pair}`&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
  &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;!&lt;/span&gt;&lt;span class="n"&gt;cache&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;has&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;key&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;p&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nf"&gt;pipeline&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s2"&gt;"translation"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="sb"&gt;`Xenova/opus-mt-${pair}`&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;dtype&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"q8"&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
      &lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="k"&gt;catch&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt; &lt;span class="nf"&gt;pipeline&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s2"&gt;"translation"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="sb"&gt;`Xenova/opus-mt-${pair}`&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;dtype&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"fp32"&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
    &lt;span class="n"&gt;p&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="k"&gt;catch&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt; &lt;span class="n"&gt;cache&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="k"&gt;delete&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;key&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="c1"&gt;// allow a retry after a network error&lt;/span&gt;
    &lt;span class="n"&gt;cache&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;set&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;key&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;p&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
  &lt;span class="p"&gt;}&lt;/span&gt;
  &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="n"&gt;cache&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;get&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;key&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="o"&gt;!&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;span class="p"&gt;}&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;I try a quantized version first (smaller and faster), and fall back to full precision if a model doesn't ship one.&lt;/p&gt;

&lt;h2&gt;
  
  
  Translating 13 languages with small models
&lt;/h2&gt;

&lt;p&gt;OPUS-MT models translate one direction each, for example &lt;code&gt;en-de&lt;/code&gt; or &lt;code&gt;de-en&lt;/code&gt;. Not every pair has a direct model, so the translator routes through English when it needs to:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight typescript"&gt;&lt;code&gt;&lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="no"&gt;DIRECT&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="k"&gt;new&lt;/span&gt; &lt;span class="nb"&gt;Set&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"de-es"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"es-fr"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s2"&gt;"fr-de"&lt;/span&gt; &lt;span class="cm"&gt;/* … */&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;

&lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;route&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="k"&gt;from&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kt"&gt;string&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;to&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kt"&gt;string&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt;
  &lt;span class="k"&gt;from&lt;/span&gt; &lt;span class="o"&gt;===&lt;/span&gt; &lt;span class="s2"&gt;"en"&lt;/span&gt; &lt;span class="o"&gt;||&lt;/span&gt; &lt;span class="n"&gt;to&lt;/span&gt; &lt;span class="o"&gt;===&lt;/span&gt; &lt;span class="s2"&gt;"en"&lt;/span&gt; &lt;span class="o"&gt;||&lt;/span&gt; &lt;span class="no"&gt;DIRECT&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;has&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="sb"&gt;`${from}-${to}`&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
    &lt;span class="o"&gt;?&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="sb"&gt;`${from}-${to}`&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;
    &lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="sb"&gt;`${from}-en`&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="sb"&gt;`en-${to}`&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Long text is split into sentence groups of about 400 characters, so each call stays inside the model's limits and the progress bar actually moves.&lt;/p&gt;

&lt;h2&gt;
  
  
  Lessons learned
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Tell people about the download.&lt;/strong&gt; A 100 MB model with no warning feels broken. Each tool says up front how big the first download is and that the browser will cache it.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Workers are not optional.&lt;/strong&gt; Running a model on the main thread freezes scrolling and buttons. Moving everything into a worker fixed that.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Small models need honest expectations.&lt;/strong&gt; Browser-sized models are good at focused jobs but less accurate than big cloud services. The UI says so, and the result can always be edited.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Check licences before you fall in love with a model.&lt;/strong&gt; It saved me from shipping something I'd have had to remove later.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Static hosting is enough.&lt;/strong&gt; No GPU servers and no API bills. The heavy lifting happens on the visitor's own hardware, which is also what keeps their files private.&lt;/li&gt;
&lt;/ol&gt;

&lt;h2&gt;
  
  
  Try it
&lt;/h2&gt;

&lt;p&gt;All the AI tools are at &lt;strong&gt;&lt;a href="https://smartonline.tools/ai-tools/" rel="noopener noreferrer"&gt;smartonline.tools/ai-tools&lt;/a&gt;&lt;/strong&gt;. They're free, need no account, and your files stay on your device.&lt;/p&gt;

&lt;p&gt;I'd love feedback, especially on speed on your hardware, accuracy, and which tool you'd want next.&lt;/p&gt;

</description>
      <category>ai</category>
      <category>javascript</category>
      <category>nextjs</category>
      <category>webdev</category>
    </item>
  </channel>
</rss>
