Complete Guide: AI Video Prompts with Face Reference (2026)
AI video generation has evolved dramatically in 2026. Tools like RunwayML Gen-3, Luma Dream Machine, and OpenAI's Sora can now produce stunning cinematic video clips from a single prompt. But the most powerful feature — and the one most creators overlook — is face reference prompting.
What Is a Face Reference Prompt?
A face reference prompt is a set of instructions that tells an AI video or image model to replicate specific facial features, hairstyles, and expressions from a source image. Rather than describing a face from scratch, you upload a photo and the AI learns the exact structure of that face — bone structure, eye shape, skin tone, and hair texture — and applies it to a brand-new scene.
This is transformative for content creators, YouTubers, Instagram influencers, and filmmakers who need consistent character appearances across multiple shots. Brands use face reference prompts to create AI ambassadors with recognizable, unchanging faces.
Midjourney Face Reference: The --cref Command
In Midjourney v6 and above, the --cref [IMAGE_URL] parameter
enables character reference mode. Pair it with --cw 100 (character weight 100)
to lock the face and hair exactly as shown in your reference image.
Example Midjourney prompt: "Professional woman sitting at a minimalist desk, soft window light, modern office, --cref [your_image_url] --cw 100 --ar 16:9 --style raw"
Lower --cw values (like 50 or
75) give the AI more creative freedom while still maintaining some similarity to the reference. Use this
when you want the character adapted to a fantasy or stylized setting.
RunwayML: Image-to-Video with Face Consistency
RunwayML's Gen-3 Alpha is one of the best tools for maintaining face consistency in video. When using Image-to-Video mode, always upload the same base image as your starting frame. This gives the model a strong anchor for your character's appearance before applying the motion described in your prompt.
Combine this with our generator's "Slightly shaky handheld follow" camera motion for a documentary-style feel, or use "Cinematic slow drone" for epic, wide-format content.
Camera Motion Styles Explained
- Handheld follow — Adds organic, real-world movement. Best for vlog, travel, documentary styles.
- Cinematic slow drone — Smooth aerial movement with wide-angle perspective. Ideal for landscape or establishing shots.
- Static tripod — No camera movement. Perfect for interview-style or product shots.
- Smooth dolly push-in — Camera glides toward the subject. Great for dramatic reveals, emotional moments.
- Orbiting around subject — 360-degree circular motion. Excellent for product showcases and character introductions.
Tips for Better AI Video Prompts
- Be specific with lighting — "Golden hour sunlight through trees" beats "natural lighting" every time.
- Describe the emotion — AI video models respond well to emotional direction: "joyful contentment," "somber reflection."
- Set the scene with sound — Many tools use your text to influence ambient audio. Mention "BGM: jazz piano" or "sound of ocean waves."
- Control speed — Add "slow motion, 0.5x speed" or "time-lapse" to the prompt for temporal effects.
- Use professional cinematography terms — "shallow depth of field," "anamorphic lens flare," "rack focus" all have real effects.
Internal Links: Explore More AI Prompt Tools
- AI Travel Video Prompt Generator — Create cinematic travel footage prompts
- AI Holiday Photo Video Generator — Animated family photo prompts
- Futuristic Rose AI Prompt Generator — Artistic AI prompts with rose aesthetics
- Best Midjourney Prompts 2026 — Comprehensive guide with 100+ prompts