Qwen 2.1 Image's Default Workflow Criticized; CFG=3 and 12 Steps Beat Krea 2 on Detail
Dear-Spend-2865 · reddit · 2026-09-29
The author argues Qwen 2.1 text-to-image is solid, but its official demo workflow (CFG=1, steps=20) put many users off.
He points to community-tested settings: CFG=3, steps=12, sampler=dpmppm3, scheduler=beta.
Hands-on impressions: Qwen 2.1 still often botches text and fuses limbs in some poses, but shows more detail than Krea 2 (especially at 2K) and understands some "adult" terms; Krea 2 needs LoRAs that oddly slow down his workflow. Qwen 2.1's editing capabilities are also a plus.
More from Multimodal
- Kling 4 Flash vs Seedance 1.5: head-to-head video generation comparison — Smart-Proof-1861 · 2026-09-30
- Flux2Klein Enhancer ComfyUI Node Boosts Krea 2 Prompt Adherence for Free — No-Bad-4838 · 2026-09-30
- GEAR Uses Geometry as Address to Route Attention, Enabling Minute-Long Camera-Controlled Video Generation — Zesong Yang · 2026-09-30
- Caltech's DynaTokens teaches dynamics to camera-controlled video models at test time, NeurIPS accepted — georgiagkioxari · 2026-09-30
- AI-generated 'day in the life in Ancient Greece' video goes viral — venturetwins · 2026-09-30
- Claude Opus 5.5 directed a full 3-minute anime music video for ~$198 in API costs — wanderingtofu · 2026-09-30