The latest articles on DEV Community by creasac (@creasac). https://dev.to/creasac https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3797996%2F48723c78-a039-46fb-a91b-871023fb70d0.png https://dev.to/creasac en creasac Mon, 02 Mar 2026 07:52:50 +0000 https://dev.to/creasac/built-shruti-a-minimal-system-wide-speech-to-text-tool-for-linux-47f2 https://dev.to/creasac/built-shruti-a-minimal-system-wide-speech-to-text-tool-for-linux-47f2 <p><em>This is a submission for the <a href="https://dev.to/challenges/weekend-2026-02-28">DEV Weekend Challenge: Community</a></em></p> <h2> The Community </h2> <p>I built this for Linux users who do a lot of typing and want fast transcription directly in their existing workflow.</p> <p>That includes:</p> <ul> <li>developers writing docs, comments, and messages</li> <li>students taking quick notes</li> <li>writers drafting ideas</li> <li>multilingual users who need reliable transcription across languages</li> </ul> <p>The goal was simple: no heavy app, no complicated flow, just speak and continue typing.</p> <h2> What I Built </h2> <p>I built <strong>Shruti</strong>, a minimal system-wide speech-to-text utility for Linux X11.</p> <p>Workflow:</p> <ol> <li>Place cursor in any text field</li> <li>Press hotkey once to start recording</li> <li>Press the same hotkey again to stop</li> <li>Shruti transcribes and types text at the cursor</li> <li>Press <code>Esc</code> anytime to cancel current recording</li> </ol> <h3> Why it’s useful </h3> <ul> <li>Works across apps where typing is possible</li> <li>Uses the user’s own Gemini API key</li> <li>High-quality transcription across languages</li> <li>Minimal visual HUD while recording/transcribing</li> <li>No always-on background process while idle.</li> </ul> <h2> Demo </h2> <ul> <li>GitHub: <a href="https://github.com/creasac/shruti" rel="noopener noreferrer">https://github.com/creasac/shruti</a> </li> <li>Demo video: <a href="https://youtu.be/UYwDzhuUQPQ" rel="noopener noreferrer">https://youtu.be/UYwDzhuUQPQ</a> </li> </ul> <h2> Code </h2> <p> </p> <div class="ltag-github-readme-tag"> <div class="readme-overview"> <h2> <img src="https://assets.dev.to/assets/github-logo-5a155e1f9a670af7944dd5e12375bc76ed542ea80224905ecaf878b9157cdefc.svg" alt="GitHub logo"> <a href="https://github.com/creasac" rel="noopener noreferrer"> creasac </a> / <a href="https://github.com/creasac/shruti" rel="noopener noreferrer"> shruti </a> </h2> <h3> minimal linux (x11) speech-to-text utility </h3> </div> <div class="ltag-github-body"> <div id="readme" class="md"> <div class="markdown-heading"> <h1 class="heading-element">shruti</h1> </div> <p>Minimal desktop speech-to-text using Gemini.</p> <div class="markdown-heading"> <h2 class="heading-element">Install</h2> </div> <div class="highlight highlight-source-shell notranslate position-relative overflow-auto js-code-highlight"> <pre>curl -fsSL https://raw.githubusercontent.com/creasac/shruti/main/bootstrap.sh <span class="pl-k">|</span> bash</pre> </div> <p>What setup asks:</p> <ul> <li>Gemini API key (hidden input)</li> <li>Preferred hotkey (default <code>Ctrl+Space</code>)</li> </ul> <p>Hotkey behavior after setup:</p> <ul> <li>Press hotkey once: start recording</li> <li>Press hotkey again: stop and transcribe</li> <li>Press <code>Esc</code>: cancel current recording</li> </ul> <p>Nothing runs in background while idle.</p> <div class="markdown-heading"> <h2 class="heading-element">Configuration</h2> </div> <p>Files:</p> <ul> <li><code>~/.config/shruti/config.toml</code></li> <li><code>~/.config/shruti/credentials.toml</code></li> </ul> <p>API key location:</p> <ul> <li>Stored only in <code>~/.config/shruti/credentials.toml</code> </li> </ul> <p>To remove your key:</p> <div class="highlight highlight-source-shell notranslate position-relative overflow-auto js-code-highlight"> <pre>rm -f <span class="pl-k">~</span>/.config/shruti/credentials.toml</pre> </div> <p>To remove all Shruti config data:</p> <div class="highlight highlight-source-shell notranslate position-relative overflow-auto js-code-highlight"> <pre>rm -rf <span class="pl-k">~</span>/.config/shruti</pre> </div> <p>Editable config fields:</p> <ul> <li><code>model</code></li> <li><code>hotkey</code></li> <li><code>max_record_seconds</code></li> <li><code>sample_rate</code></li> <li><code>channels</code></li> <li><code>prompt</code></li> </ul> <div class="markdown-heading"> <h2 class="heading-element">Commands</h2> </div> <div class="highlight highlight-source-shell notranslate position-relative overflow-auto js-code-highlight"> <pre>shruti setup shruti doctor --verbose shruti oneshot</pre> </div> <div class="markdown-heading"> <h2 class="heading-element">Limitations</h2> </div> <ul> <li>Linux X11 only (Wayland blocks unrestricted global hotkey/input injection for security)</li> </ul> <div class="markdown-heading"> <h2 class="heading-element">License</h2> </div> <p>MIT. See <a href="https://github.com/creasac/shruti/LICENSE" rel="noopener noreferrer">LICENSE</a>.</p> </div> </div> <br> <div class="gh-btn-container"><a class="gh-btn" href="https://github.com/creasac/shruti" rel="noopener noreferrer">View on GitHub</a></div> <br> </div> <h2> How I Built It </h2> <ul> <li> <strong>Python 3.11+</strong> for fast iteration and portability</li> <li> <strong>Gemini API</strong> (<code>gemini-2.5-flash-lite</code>) for transcription</li> <li> <strong>sounddevice + numpy</strong> for microphone capture and waveform data</li> <li> <strong>xdotool</strong> for text insertion at the active cursor location (X11)</li> <li> <strong>Tkinter</strong> for a compact overlay HUD</li> <li> <strong>TOML config</strong> in <code>~/.config/shruti/</code> for user-controlled settings and API key storage</li> </ul> <h3> Design choices </h3> <ul> <li>Keep the product intentionally narrow: one job, done well</li> <li>Hotkey-first interaction for speed</li> <li>Minimal UI and minimal config surface</li> <li>Users bring their own Gemini API key, so usage and billing stay fully under their control.</li> <li>Cloud-based transcription gives wider language coverage, keeps local compute light, and allows easy model upgrades.</li> </ul> <h2> What’s Next </h2> <ul> <li>Better packaging/distribution options</li> <li>More desktop-environment setup helpers</li> <li>Explore Wayland-compatible path where feasible</li> </ul> <p><code>Built by @creasac.</code></p> devchallenge weekendchallenge showdev