Building a Digital AI Influencer with Qwen: Reference Images Keep Faces Consistent
tryku2 · reddit · 2026-09-27
The author shares early experiments using Qwen to create a digital AI human/influencer, with decent results so far.
The notable detail: when a face is inserted as a reference image, Qwen maintains strong face consistency even with the face detailer node disabled in ComfyUI — suggesting the model's reference-image capability is doing the work without post-processing fixes. The author notes their prompts and workflow are still simple, so there's headroom to improve.
More from Multimodal
- TeleOCR Trends on Hugging Face: A Qwen2.5-VL-Based Chinese Document OCR Model — XingChen-AGI · 2026-09-28
- Higgsfield ships 11 production skills that leave Claude with editable project files — xiaohu · 2026-09-28
- Opus 5.5 makes a music video, drawing love from AI circle — repligate · 2026-09-28
- Under $1 per 66-second reel: open-model pipeline with Qwen3-TTS and MiniMax H3 — victor_explore · 2026-09-28
- Music generation is '110% solved', says researcher as AI song quality stuns — teortaxesTex · 2026-09-28
- 282 viral Claude Opus 5.5 videos with exact prompts, all in one open-source repo — dotey · 2026-09-28