What does implementing word-level audio precision actually look like in a React Native app? We did it for Perch using our own React Native Audio API — a library that brings the power of Web Audio API to mobile, giving developers full, low-level control over audio playback, synthesis, and streaming. Using it, we were able to give Perch features that off-the-shelf libraries simply couldn't handle: tap any paragraph to start listening, tap a specific word to jump right to it, highlight sections to save for later. On top of that, we cut audio preloading from ~10 seconds to ~1.5 seconds. ⏱️ We spend ~30% of our engineering time on open source — react-native-audio-api is one of those projects. Read more about the results we achieved with Perch: https://lnkd.in/dGmxsFru Check the docs: https://lnkd.in/d264i5ht
Software Mansion
Tworzenie oprogramowania
Kraków, małopolskie 8343 obserwujących
We are a developer experience lab & software agency from Kraków, contributors to React Native and Membrane Framework.
Informacje
We are a team of engineers with strong technical know-how, and we enjoy complicated, unconventional projects. We are based in Kraków, Poland, but our clients come mostly from New York and San Francisco Bay Area. Work culture We draw from the experience of working with global startups and lead our projects using agile software development methodologies. We are quick to react to the changes in product development, and we are always ready for the plot twists. Team & Technology As the challenges we like to take on are exceptional, we focus on building an exceptionally talented team. We hire developers who are flexible, willing to change projects and learn new technologies. We choose our technologies based on whether it is the best fit for the project. While we have always been happy to try out new, cutting-edge technologies, we choose them by reason, not by fleeting trends. In most of our projects, we use React, React Native, Node.js, Python, Ruby on Rails and Elixir — but we do not like to limit ourselves. Open source We build our own tools. We actively contribute to open source, working, among others, on Expo, Membrane, React Native Reanimated and Gesture Handler. At the same time, we are an important part of the React Native community - we organize community meetups and a React Native focused conference called App.js.
- Witryna
-
http://swmansion.com/
Link zewnętrzny organizacji Software Mansion
- Branża
- Tworzenie oprogramowania
- Wielkość firmy
- 201-500 pracowników
- Siedziba główna
- Kraków, małopolskie
- Rodzaj
- Spółka akcyjna
- Data założenia
- 2012
- Specjalizacje
- Web development, Android, iOS, Startups, Mobile applications, Software development, Ruby on Rails, Clojure, React.JS, React Native, Elixir, Node.js, WebRTC, Streaming i GStreamer
Lokalizacje
-
Główna
Otrzymaj wskazówki o trasie dojazdu
ul. Zabłocie 43B
Kraków, małopolskie 30-701, PL
Pracownicy Software Mansion
Aktualizacje
-
The Software Mansion × Google DeepMind Gemini Hackathon is a wrap – and the numbers still surprise us: 🔥 250+ applications received 🔥 100+ last-minute applicants for fewer than 20 remaining spots – we had to close registration early 🔥 149 participants checked in at our office 🔥 66 teams formed 🔥 49 working projects submitted in just 7 hours The scale of this event went far beyond what we initially planned – and seeing our office packed wall-to-wall with builders determined to create something great made it all worth it. Meet the winning projects: 🥇 Fact-check Live – a real-time video content analysis system combining media streaming with generative AI: https://lnkd.in/dRd_FDaW 🥈 Jutsu – Naruto Hand-Sign Battle (yes, it's exactly as fun as it sounds): https://lnkd.in/dh7TMhMA 🥉 PrivyMeet – turning raw meetings into actionable insights with on-device privacy filtering and human-in-the-loop AI: https://lnkd.in/d_PTTtDJ Honorable mentions: 🏅 PINGwin – a ping pong game you play with your hand in front of the camera instead of a paddle: https://lnkd.in/dh6K-A33 🏅 Live Embeddings Studio – contextually relevant ads on live video streams using multimodal embeddings: https://lnkd.in/dB_j2zFM Thank you to every single participant. Congratulations to the winners and all teams who submitted. A special thank you to Google DeepMind for co-hosting the event with us and to our judges: Kamil Stanuch, Krzysztof Magiera, Piotr Skalski, Amit Vadi, Prince Canuma, Maria Eckes We're already thinking about what to organize next. Stay tuned 🚀
-
-
Another great addition to React Native Enriched Markdown! Check it out 🔥
Stop fighting with limited native menus! 🛠️ I’m excited to announce the new contextMenuItems prop for react-native-enriched-markdown. You can now trigger custom actions - like the "Summarize with AI" demo below - directly from any selected text in both EnrichedMarkdownText/Input components. 💡 Why this matters: Modern apps need to be more than just static text. By combining this new prop with our other library, react-native-streamdown, you can create dynamic, interactive experiences with smooth, streaming outputs. Key Highlights: 🔹 Custom Actions: Add any menu item (Summarize, Search) to the native selection menu. 🔹 Full Text Access: Easily grab selected strings to pass into your custom logic. 🔹 Consistency: Works seamlessly across both viewing and editing modes. Improving the native feel of your React Native apps just got a lot easier. Links to the repos are in the comments. 👇
-
Keeping up with all those releases and news in tech is a full-time job. We're building MailBoyAI to change that! ✉️ 1️⃣ Tell us what you're working on 2️⃣ The AI agent scans hundreds of sources on LinkedIn, GitHub, Substack, HuggingFace, and more 3️⃣ Get a tailored digest every week, straight to your inbox Sounds interesting? Join the waitlist: https://lnkd.in/d66nFuqt
-
Software Mansion x Gemini Hackathon is on – here are some sneak peeks! 📸 https://lnkd.in/dzYpsh_y
-
-
React Native Audio API is a powerful, flexible, and truly cross-platform way to add audio to your React Native app. 🎵 Want proof? Check out the video where we share a bit about our work with Perch – a newsletter and content aggregator app. The client reached out because they wanted to take the UX to the next level with reliable audio that stays perfectly in sync with the UI. The goal was simple: enable natural text-to-speech so any article could be listened to like a podcast. During our collaboration, we cut audio preloading from ~10 seconds to ~1.5 seconds and ensured smooth multi-sound and background playback, along with perfectly synchronized audio highlights and progress tracking. ✅
-
We expected a lot of interest. We didn't expect to close registrations early. 260+ applications. Almost 200 accepted participants. 5 teams walking away as winners. The Software Mansion x Gemini Hackathon starts tomorrow. Let's see what happens when you put this many builders in one room with Google DeepMind and Software Mansion tech 🔥 Kamil Stanuch, Krzysztof Magiera, Maria Eckes, Piotr Skalski, Prince Canuma, Amit Vadi, Thorsten Schaeff, Google AI for Developers
-
-
We just shipped React Native ExecuTorch v0.8.0 – our biggest release yet! 🚀 Includes: Vision Camera integration, multiple computer vision models (RF-DETR by Roboflow, Liquid AI’s Vision Language), Bare React Native Support, and more. Full release notes: https://lnkd.in/dWTKHHEk
Earlier this year we completed a computer vision project for a client where on-device inference speed was critical. I can't share the details, but I can say this: that project convinced us that real-time CV on mobile is not a novelty, it's a real product need. And it pushed us to go much deeper on the computer vision side of 𝗥𝗲𝗮𝗰𝘁 𝗡𝗮𝘁𝗶𝘃𝗲 𝗘𝘅𝗲𝗰𝘂𝗧𝗼𝗿𝗰𝗵. Today we're releasing 𝘃𝟬.𝟴.𝟬 and it's our biggest and at the same time the most CV-heavy release to date. We added 𝗥𝗙-𝗗𝗘𝗧𝗥 and 𝗬𝗢𝗟𝗢 support for object detection and instance segmentation. Both work with real-time camera frames through VisionCamera integration. YOLO26 Nano runs object detection in 29ms on iPhone 17 Pro, RF-DETR Nano in 101ms. Instance segmentation with YOLO26 Nano comes in at 90ms. In the demo video you can see it all running live at 9 FPS with full bounding boxes and per-pixel masks on multiple objects. On top of that, this release brings Vision Language Models to mobile. You can now pass images alongside text to on-device LLMs, meaning the model can see what the user sees. Think visual Q&A, document understanding, or scene description. All running locally, all private. And here's what excites me beyond this release. Phones are just one edge device. The expertise we're building around real-time CV inference on constrained hardware applies to a much broader world. Drones, embedded systems, industrial devices. We've already started exploring this. Huge kudos to the Software Mansion AI team. From the initial v0.1.0 where we were just proving on-device inference was feasible, to shipping real-time instance segmentation and VLMs on a phone. The progress over the past year has been remarkable. Take a look at the demo below to see it in action 👇 Full changelog and detailed benchmarks in the comments. If you're exploring an on-device use case or just want to bounce ideas, my DMs are open.