If you’re just getting started, don’t overcomplicate it, biggest challenge is maintaining the same face across different tools. If you want, I can share a simple workflow I use

@Joshua Hayes It’s easier than before, agreed. But keeping identity consistent across different models/tools is still a real challenge in production workflows.

Yeah this happens because both models interpret faces differently, so the identity drifts even with the same reference. Simple fix that works:Use 3–5 reference images (not just one), keep image strength low in Seedream, and generate multiple angles from Nano Banana first to lock the face. If needed, do a final face swap to fix consistency. If you plan to reuse the character, training a small LoRA is the best long-term solution. If you want, I can share a simple step-by-step workflow to keep character identity consistent across tools

Step-by-step (works reliably): 1. Nano Banana: Generate 10-15 images of the same character (front, side, different expressions) 2. Pick 3-5 best images → use these as your main references 3. Seedream: Upload all references + keep image strength low (around 0.3-0.5) 4. Use a detailed prompt (face shape, eyes, hair, skin tone) instead of “same person” 5. Generate → pick best output 6. Final step → face swap / face restore to fix any small changes Pro tip: If you’ll reuse this character a lot, train a small LoRA once → saves a ton of time later. If you want, I can help you set this up step-by-step