Kokon 0.10 reuses Qwen-Image 2.1's text encoder to write prompts, runs on 8GB cards
Ok-Director4174 · reddit · 2026-10-22
Kokon, an open-source settings-free desktop photo editor for Qwen-Image 2.1, adds Prompt+ — detailed prompt writing powered by the already-loaded Qwen3-VL 8B text encoder, no second model needed (2s/image on RTX 5070 Ti, 13s on M2 Pro, works on 8GB cards) — plus beta support for up to three Qwen-Image 2.1 LoRAs. The author honestly notes popular 2.1 LoRAs don't beat Turbo with good prompts, so no curated catalog yet. Built on a stable-diffusion.cpp fork adding text generation (127 tok/s on 5070 Ti), bit-identical image output, AMD support via Vulkan, AGPL-licensed, non-commercial weights.
Related event: Kokon 0.10 reuses Qwen-Image text encoder to auto-write prompts(2 posts)→
More from Infra
- TSMC surveys suppliers on rare-earth-free motors for factory automation — pstAsiatech · 2026-10-22
- Zoho cries poor over AI costs, then applies to build $200M semiconductor fab — kingslyj · 2026-10-22
- NVIDIA recaps GTC Berlin dev community: open models to physical AI — NVIDIAAI · 2026-10-22
- GPU payback math: B300 rigs recoup costs in 10 months, 1GW build could earn ~$38B a year — tctjr · 2026-10-22
- What do customers actually get for billions in AI capacity prepayments? 20 companies reviewed — tengyanAI · 2026-10-22
- Flint 40B hits 128 tok/s on M4 Pro via Kapitsa, 1.9x faster than llama.cpp — DotaMate · 2026-10-22