You generate three clips for a 30-second narrative. The character wears a red hat in clip 1. In clip 2, the hat is suddenly blue. In clip 3, no hat at all. Same character, same supposed scene. The film falls apart — every cut breaks the viewer’s suspension of disbelief.
This is the multi-clip version of character drift: clip-level details (clothing, props, accessories, hairstyle) get re-invented each generation because nothing locks them.
Common causes
Ordered by hit rate, highest first.
1. Each clip generated independently with no shared anchor
Three separate prompts → three separate model outputs. Each output starts from scratch on details unless you anchor them.
How to spot it: each clip has its own prompt; nothing is shared except the broad character description.
2. Per-clip prompts describe details inconsistently
Clip 1 says red baseball cap. Clip 2 says red cap. Clip 3 says nothing about a hat. Even when you intend continuity, slight wording differences produce different results.
How to spot it: compare your prompts side-by-side. Different wording for the same element = continuity breaks.
3. No fixed seed across clips
Same character + same prompt + different seed = different fine details. Hat angle, clothing fit, hair style all vary.
How to spot it: each clip’s seed is random or different.
4. No reference image shared across clips
Even with identical prompts, no shared visual anchor means the model has to re-derive the look each time.
How to spot it: you’re not using image-to-video, or you’re using a different reference per clip.
5. Wardrobe / props described in only some clips
Clip 1: "woman in red dress and gold necklace, holding a leather bag"
Clip 2: "woman walking through park"Clip 2 doesn’t mention dress, necklace, or bag — model is free to drop them.
How to spot it: clip 2+ omits details mentioned in clip 1.
6. Clips generated in different tool sessions / days
Some tools cache style / character settings per session. Different sessions = different defaults = drift.
How to spot it: clips were generated days apart in different sessions.
Shortest path to fix
Step 1: Write a “production bible” once, reuse it
For any multi-clip narrative, write down once:
CHARACTER: 28-year-old woman, chin-length asymmetric bob,
single mole below right eye, sharp jawline
COSTUME: red wool baseball cap, oversized denim jacket,
white t-shirt, blue jeans
ACCESSORIES: gold thin chain necklace, brown leather crossbody bag
HAIRSTYLE: chin-length bob, tucked behind left earPaste this verbatim into EVERY clip prompt.
Step 2: Feed the same character reference image to every clip
# Process
1. Generate ONE reference image showing the character with full costume/props
2. Save as `production_reference.png`
3. Use as start frame (Runway/Kling/Pika image-to-video) for every clipStep 3: Lock the same seed for the same character across clips
# Runway: --seed 4242
# Pika: same seed setting
# Kling: same seed (where supported)Same seed + same reference + same character description = highly consistent output.
Step 4: Use one tool / one session for the whole project
Don’t mix Runway + Kling + Pika within one project. Pick one and stick with it. Generate all clips in the same session if possible.
Step 5: QC every clip immediately
After generating each clip:
1. Pause at frame 0 and frame N (last frame)
2. Open production bible
3. Verify every item: hat color, jacket, necklace, bag
4. If any item drifted, regenerate before continuingStep 6: Final post-production touch-ups
For minor inconsistencies (slight color drift on the hat):
# DaVinci Resolve Power Window
- Isolate the hat across the clip
- Color-shift to match other clips
# Photoshop / After Effects
- For single-frame drift, mask and patchThis is last-resort; better to regenerate.
Prevention
- Before any multi-clip project, write a production bible (character + costume + props)
- Always start by generating ONE master reference image; reuse it
- Lock the same seed for the same character throughout the project
- QC at clip-level before continuing to the next