The clip is the prompt
Reference footage carries the direction — no prompt engineering needed to describe a camera move.
Reference-to-Video AI
Show the edit, do not describe it. A reference clip sets the motion; your photos set the cast.
Modelo de IA 4 créditos
GPT Image 2Geração de imagens versátil em 4K Instruções0/2000
Evite personagens, logotipos e músicas protegidos por direitos autorais, bem como conteúdo sensível ou explícito — pode ser bloqueado.
Resolução
Proporção
Geração em lote1
Custo estimado: 4 créditos
Sua edição aparece aqui
Escolha uma referência, adicione os sujeitos e toque em Gerar.
Why Reference-to-Video AI
Why creators choose Reference-to-Video AI
Reference-to-Video AI runs inside the Katana AI studio, so you can generate, edit, and export images without juggling separate tools or subscriptions.
Your subjects stay yours
Subject photos anchor identity — faces, products and outfits survive the rebuild.
Motion that already works
Presets and your own clips supply choreography that is proven to land on feeds.
Share-ready MP4
Vertical, landscape or square — download and post with your own soundtrack.
Workflow
Reference-to-video, from upload to MP4
Reference-to-video replaces the prompt with footage: a clip you supply directs the output, and subject photos decide the cast. This page walks the full pipeline — what to upload, what transfers, and how to review the result.
What the reference contributes
Everything you cannot easily write down: the camera path, the subject's movement, the framing, the cut points, the rhythm. A ten-second clip carries a shot list, a choreography and an editing decision in one file.
That is why reference quality dominates the result. Stable framing, legible motion and a clear subject produce edits that hold together.
What your photos contribute
Identity. The engine looks at your subject photos — up to eight — and rebuilds the reference so that the person, product or outfit in frame is yours. More angles mean a steadier likeness through turns and movement.
The inputs divide the work cleanly: reference for motion, photos for cast, an optional audio file for sound, an optional instruction for emphasis.
Settings worth touching
Two choices shape the deliverable: aspect ratio and resolution. Match the aspect to where the clip will post — 9:16 for Reels and Shorts, 16:9 for YouTube, 1:1 for feeds. Resolution is a budget lever: draft at 480p, finish at 720p or 1080p.
Duration is not a choice at all — it follows the reference. That constraint is the point: the rhythm you picked is the rhythm you get.
Review like an editor
Check identity first — faces, logos, fabric details — then timing against the source. AI output varies run to run, so keep the take where both hold; a 480p draft makes that check cheap.
Label the output as AI-generated where the context calls for it, and only upload footage and photos you have rights to use.
How to use Reference-to-Video AI
How to use Reference-to-Video AI in three steps
Upload the reference clip
4–15 seconds of footage — yours or a preset — sets the motion and rhythm.
Add your subject photos
1–8 images of the person, product or outfit that should appear.
Generate and review
Watch the rebuild, check the subjects, download the MP4.
Who uses Reference-to-Video AI
Who uses Reference-to-Video AI for image work
Creators & editors
Rebuild trending formats with your own cast — no reshoot.
Brands & sellers
Product photos step into proven ad-style motion.
Social teams
Format-matched clips sized and paced for feeds.
Fans & communities
Put yourself inside the edits you already love — with rights confirmed.
More models
Reference-to-Video AI and the rest of the studio in one place
Katana Edit handles the video pipeline end to end; companion image models cover thumbnails, covers and stills.
Reference-to-videoMotion transferSubject photosOptional audio4–15s editsImage generationMP4 exportPreset library
FAQ
Reference-to-Video AI FAQ
01What is reference-to-video?
A workflow where a video — not a text prompt — directs the output. The reference supplies camera moves, cuts and timing; your subject photos supply who or what appears.
02How is it different from text-to-video?
Text describes; reference demonstrates. Pacing, transitions and framing are notoriously hard to write out — a 10-second clip carries all of it exactly.
03What makes a good reference?
Clear motion, stable framing and a readable subject. Clips with one strong move — a walk, an orbit, a reveal — rebuild most reliably.
04What makes a good subject photo?
Sharp, well-lit images with the whole subject in frame. Multiple angles of the same subject help the engine hold identity across the edit.