Gradium speech models run 26x realtime in iPhone Safari, under 40 MB, Apache-2.0 open source
blixt · x · 2026-11-04
Developer blixt announces all Gradium AI speech models now support multiple languages (e.g. French). Key specs: 26x real-time speech in Safari on an iPhone, model under 40 MB brotli'd, same voices and quality as Gradium's own runtime. Every word is timestamped so captions and read-alongs just work. Open source under Apache-2.0.
More from Multimodal
- Combining Midjourney styles with HeyGen video yields striking results — umesh_ai · 2026-11-04
- Ailleurs tool uses World Labs Atlas API to restyle real 3DGS captures while preserving geometry — theworldlabs · 2026-11-04
- Visual essay: AI cut 3D draft generation by 4 orders of magnitude, value shifts to topology — cpaik · 2026-11-04
- BluePixel-Turbo released: native 4K instruction-based image editing in 4 steps — MonkeyBoyPoop · 2026-11-04
- Pika H3 Flex debuts #1 on DesignArena; community ships 2-step Turbo LoRA within 24 hours — _akhaliq · 2026-11-04
- Civitai ends paywalls for Flux.2 Klein checkpoints as BFL enforces non-commercial license — Mabuse046 · 2026-11-04