Phosphene update makes Hailuo H3 2.3x faster on Mac, lighter on 16GB machines
cocktailpeanut · x · 2026-10-23
Phosphene shipped v4.19.10 and v4.19.11 with two notable improvements for local generation on Apple Silicon:
- 4.19.11 (LightX2V on Mac): H3 at High quality renders 2.3x faster — a 5-second image-to-video clip on an M4 Max 64GB drops from 35 to 15 minutes, powered by ModelTC's lightx2v v1.3 (1.4GB download). H3 Fast/Turbo get sharper mouths, previews arrive earlier, clips play in Safari, and 1-click install is available via Pinokio.
- 4.19.10: On Macs with ≤16GB RAM, Auto picks a lighter image engine when needed; video uses less memory on ≤24GB machines with identical results, and long clips are converted to frames in smaller chunks.
- v4.19.8 added license screens for Hailuo H3, Ideogram 4, YuE2, FLUX.1 Krea [dev] and Bria FIBO.
Caveat: Hailuo H3's license excludes the EU, UK, South Korea, and the US.
Related event: Open-source Phosphene Speeds Up Hailuo H3 on Mac by 2.3x(2 posts)→
More from Infra
- GLM-5.3-Flash RL training pushed to 1M-token context in SkyRL, 2.1x faster at 256K — AndrewLBeam · 2026-10-23
- Perplexity's Portable Computer lands on AMD Ryzen AI Max, running local 27B model — AravSrinivas · 2026-10-23
- Liquid AI Launches RIPL: Ultra-Low-Latency Inference Engine With Sub-50ms End-to-End Responses — JosephJacks_ · 2026-10-23
- OpenJet claims Qwen3.8-27B coding runs 1.9x faster than llama.cpp on one RTX 3090 — JeremyCMorgan · 2026-10-23
- Jensen Huang lays out NVIDIA's full Vera Rubin agentic system: NVL72, MGX, CMX and RTX — rohanpaul_ai · 2026-10-23
- ONNX to Transformers.js: running models directly in your browser — JosephJacks_ · 2026-10-23