Creating a consistent character across multiple AI-generated images is genuinely difficult without fine-tuning or custom models. But prompt chaining — combined with seed locking and style references — gets you 80% of the way there.
What Is Prompt Chaining?
Prompt chaining means using the output of one generation as a reference or seed for the next. In Midjourney, this is done with --sref (style reference) and --cref (character reference, v6.1+). In Stable Diffusion, you use img2img with controlled denoising strength.
Step 1: Create Your Anchor Image
Your anchor image is the definitive reference for your character. Spend time on this one. Get the hair, face, style, and lighting exactly right before you move to other scenes.
portrait of a young woman with [specific hair], [specific eyes], [specific style], neutral expression, simple background, studio lighting, --ar 1:1 --v 6 --style rawNote down the job ID and seed from your anchor image immediately. You'll need these for all subsequent generations.
Step 2: Use --cref for Character Consistency
In Midjourney v6.1+, --cref [image URL] tells the model to use that image as a character reference. Combine with --cw (character weight, 0-100) to control how strictly it follows the reference.
same woman walking through a rainy Tokyo street at night, cinematic, neon lights --cref [your-anchor-url] --cw 80 --ar 16:9 --v 6Step 3: Lock the Aesthetic with --sref
Character reference handles the face and body. Style reference (--sref) locks the visual aesthetic — colour grade, lighting style, rendering approach — across all scenes.
When Prompt Chaining Isn't Enough
For commercial-grade consistency — same character in dozens of varied scenes — you'll eventually need to fine-tune a model (LoRA in SD, Niji for anime characters, or professional tools like HeyGen for video). Prompt chaining is a great starting point but has limits.
