Updated every morning
The journal. What changed today.
New models, checkpoints, LoRAs and ComfyUI workflows for people who run an AI influencer: news, reviews, comparisons and top lists, with every source linked.
- Troubleshooting
2026-10-14 · 9 min
ComfyUI out of memory: the Dynamic VRAM fix that works
ComfyUI says out of memory with free VRAM? Match the error text, test --disable-dynamic-vram, pin comfy-kitchen, and drop flags that no longer work.
- Top list
2026-10-13 · 9 min
Best open video models for vertical reels in 2026
Wan 2.2, LTX-2.5, MiniMax H3 and Kandinsky 6.0 ranked for 9:16 character reels: licenses, sound, VRAM and the one place each of them loses.
- Comparison
2026-10-12 · 9 min
Krea 2 vs Qwen-Image 2.1: which keeps the same face
Krea 2 Turbo against Qwen-Image 2.1 for an AI influencer: portrait realism, identity from reference photos, VRAM, speed and the two very different licenses.
- Model review
2026-10-10 · 9 min
MiniMax H3 in ComfyUI: specs, VRAM and the license catch
MiniMax H3 makes 4 to 15 second clips with sound in ComfyUI. What it does for an AI character, what 12 and 16 GB cards can run, and who its license excludes.
- News
2026-10-09 · 7 min
ComfyUI news, October 2026: Qwen-Image 2.1, Comfy Agent and more
What changed for AI influencer creators in early October 2026: Qwen-Image 2.1 and its Turbo, Comfy Agent, Kandinsky 6.0 with audio, ComfyUI 0.39, Civitai.