Full-time aero engineer ✈️ | Part-time maker 🚀 💬 chattier.dev 📊 plotpink.xyz 🎾 padelamix.com 🖱️ bit.ly/lima-app

Madrid, Comunidad de Madrid
Joined July 2012
While building Lima & Chattier, I learned a ton about face landmark estimation from video and 3D face blendshapes for speech animation. Here are the tools and papers that helped me the most 🧵👇 #AI #ComputerVision #3DAnimation #GameDev #IndieDev #BuildInPublic
📊 Building PlotPink - engineering data viz that doesn't take forever to set up. Shipping fast this time (shoutout to @marc_louvion). Chasing the dream tool I’d actually want for my full-time engineering work. 🚀
1
1
3
Nvidia's Audio2Face-3D blows me away, but needs dedicated inference servers + Nvidia AI Enterprise license (ouch for solo builders) Meta's Oculus Lipsync is lightweight enough to run locally on anything, just basic visemes without expressions though Is there a middle ground?🤔
1
My endless cycle of ambitious projects... Get sucked in for months trying to force these big ideas to work. Then realize I probably should've just shipped something simple instead. Sound familiar?
Why does Android’s on-device speech to text cut off after a few seconds of silence or even mid-speech, even though it’s running locally? Docs say it’s “likely to stream audio to remote servers… not intended for continuous recognition” but that is not the case here.
1
I run Chattier, a SaaS for multilingual AI chatbots. Now I’m turning that tech into an English speaking practice app. 🎙️ 10 hours in, aiming to launch in ~3 weeks. Excited to share the journey and updates along the way.
2
4
My recent favorites 📚 1️⃣ The Personal MBA - practical business insights. 2️⃣ Fall in Love with the Problem, Not the Solution - solve the right problems. 3️⃣ The Lean Startup - test ideas fast and smart. Loved these reads! @joshkaufman @UriLevine1 @ericries
1
1
Every Build in Public community on X: - Share your work! ✅ - No self-promotion! ❌ Can anyone explain how this works in practice? 😅
🚀 Really struggling to pinpoint the right niche users for Chattier (my SaaS for 3D AI avatars + customer support). Where do you think I could find the people who’d benefit most?
5
If you’re working on avatars, facial tracking or speech animation these tools are gold. Got more resources? Drop them below 👇 If you found this thread useful, share it so others can discover it too 🙌 #AI #GameDev #BuildInPublic
1
Audio-driven facial animation in MetaHuman Animator Converts audio to realistic facial animation, real-time or offline. Fully integrated in @UnrealEngine. 🔗metahuman.com/ 🔗dev.epicgames.com/documentat…
Azure Speech Viseme Generation Another cloud-based TTS service with visemes output. Supports multiple voices, styles, and real-time streaming, making it easy to power interactive avatars or games. @Azure 🔗learn.microsoft.com/en-us/az…
Amazon Polly - TTS with viseme generation One of the few TTS services to output viseme history, letting you animate lips movement in sync with speech. Used in Chattier’s speech animations! @awscloud 🔗docs.aws.amazon.com/polly/la…
Microsoft Facial Dense Landmarks Another great release from Microsoft Research, concerning 3D face reconstruction with dense landmarks. 🔗microsoft.github.io/DenseLan…
Meta Sapiens Human Vision Models Awesome models for: – 2D pose estimation – Body-part segmentation – Depth estimation – Surface normal prediction Licensed for non-commercial use. @RealityLabs 🔗about.meta.com/realitylabs/c…
ICT-FaceKit Morphable face model + toolkit with identity & expression blendshapes. Free version from USC ICT’s Vision & Graphics Lab. @USC_ICT 🔗arxiv.org/abs/2004.03458 🔗github.com/USC-ICT/ICT-FaceK…
Ready Player Me Create free customizable avatars (even for commercial projects). Web-ready (three.js compatible) & equipped with expression blendshapes. They power Chattier’s avatars! @readyplayerme 🔗readyplayer.me/avatar
Fake It Till You Make It (Microsoft) Showed how face landmarks detection models can be efficiently trained with synthetic data. Photorealistic images rendered with Blender Cycles. Dataset available for non-commercial research. @MSFTResearch 🔗microsoft.github.io/FaceSynt…
Google MediaPipe Face Landmarks Detection Runs locally on CPU at high FPS, highly customizable, and rock-solid. The core of Lima’s facial tracking ❤️ @GoogleAI 🔗ai.google.dev/edge/mediapipe…