Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

DirectML

AI overview

Sign in and the AI will write an overview from our coverage.

Headlines · 3

  1. Shengjing r2 generates short sound effects locally on Windows

    The author released Shengjing r2, a Windows desktop tool that generates short sound effects and ambient audio from text descriptions using TinyAudio-MF and ONNX Runtime. It can generate one to four versions, display waveforms, play previews, export 16-bit PCM WAV files, and use DirectML GPU decoding with CPU fallback; the author reports that decoding on an RTX 4060 Laptop GPU fell from 22.63 seconds to 0.43 seconds in one comparison. The package runs on Windows 10/11 x64 without an online generation API once its model resources are prepared, although English prompts are recommended and Chinese-prompt quality remains insufficiently tested.

    掘金 · 人工智能 · 🔥 0
  2. Microsoft's Nebula experiment turns photo libraries into 3D spaces

    Microsoft design director Lucas Couto has published Nebula, an experimental open-source Windows 11 photo app that turns a library into an explorable 3D space. It uses Google SigLIP 2 with Windows ML, ONNX Runtime, and DirectML for local semantic search, while its developer claims the renderer can display up to 100,000 images in real time. The Apache 2.0 project remains an experiment rather than a replacement for the Windows 11 Photos app, and users must compile the source themselves.

    cnBeta · 🔥 0
  3. Xiaoya Children's Voice Companion releases r4 with GPU speech synthesis

    The author released r4 of the Windows desktop app Xiaoya Children's Voice Companion, adding DirectML GPU speech synthesis, wake-word interaction, and an interface for self-hosted OpenAI-compatible language model services. The app can use DeepSeek's official service or a compatible self-hosted backend, while speech recognition and synthesis run locally; the app does not download or host the self-hosted model. In the author's test on an RTX 4060 Laptop, warmed-up speech synthesis averaged about 0.32 seconds on the GPU versus 5.16 seconds on the CPU, covering only the synthesis stage.

    掘金 · 人工智能 · 🔥 0

Experience and discussion from the community

Share my DirectML experienceAsk about DirectML

Nobody has shared their experience with DirectML yet.