Browse categories
Explore
Fiverr Pro
English
$
USD
I will create a consistent ai character with comfyui and flux
Zane S
About this gig
I build you a reusable pipeline that generates the SAME character across any scene, outfit, or pose using reference-based identity (ComfyUI + FLUX). No slow LoRA training, no per-image guesswork.
What you get:
A custom ComfyUI/FLUX workflow tuned to your character or style
Consistent identity from a reference image the same face across scenes (Standard and up)
Sample images, documentation, and the workflow files all yours
Optional: your pipeline deployed as a RunPod serverless API (Premium)
Why reference-based: no training wait, no huge datasets. Give me one clean reference and I lock the identity natively faster and cheaper than LoRA, and easy for you to re-run yourself.
Important consent & likeness: for any real person's likeness you must have that person's explicit consent, used only within the scope they agreed to. For public figures or third parties without consent, I'll use fictional, non-existent personas instead. No deepfakes, no impersonation.
How I work: text-first and async clear specs, sample previews, and reviewable increments.
Message me before ordering with your character concept or reference, and I'll confirm the approach and the right package.
Delivery style preference
Please inform the freelancer of any preferences or concerns regarding the use of AI tools in the completion and/or delivery of your order.
Get to know Zane S
Zane S
Senior AI Engineer
- FromSouth Korea
- Member sinceJul 2026
- Avg. response time1 hour
Languages
English, Korean
I am a Senior AI Engineer and former CTO with 7 years of experience building production-ready LLM agents, RAG systems, and generative AI pipelines. I specialize in turning complex AI experiments into measurable business impact, such as cutting workflows from 60 minutes to 5 and reducing infrastructure costs by 80%.
FAQ
Do you train a LoRA?
No — I use reference-based identity (ReferenceLatent) in ComfyUI/FLUX, which locks the character from a single reference with no training wait. Faster, cheaper, and easy to re-run. LoRA training is available on request.
Can I run the pipeline myself afterward?
Yes. You get the ComfyUI workflow files and documentation, so you can generate new images yourself. Premium also deploys it as a RunPod serverless API you can call.
Can you copy a specific real person's face?
Only with that person's explicit consent, used within the scope they agreed to. Without consent (e.g. celebrities), I'll create a fictional persona instead. No deepfakes or impersonation.
What reference image do you need?
For a consistent character, one clean, well-lit reference works best. For a brand-new persona, just describe the look and I'll generate the reference first.
