PixRoute

Reference to Video

NEW

Reference to Video runs on the strongest reference engines available — Seedance 2.5, Seedance 2.0 and 2.0 Fast, and MiniMax H3, switchable in one click. Upload reference images and video clips, cite them in your prompt as @Image1, @Video1, and generate clips up to 30 seconds where your characters, motion, and visual style stay locked from first frame to last.

Model
Input
Upload reference images — up to 10JPEG · PNG · WebP · GIF — max 10MB · 0/10
Upload reference videos — up to 10 clips · 30s totalMP4 · MOV — max 100MB · 0/30s
0/5000
Ratio
Length
Resolution

Video workspace

Example · MiniMax H3 · text-to-video run · 768p · 5s

A single ripe lemon on a wet slate countertop, a drop of water falls and splashes in slow motion, soft window light from the left.

Key Features of Reference to Video

Why creators pick PixRoute for reference to video ai.

Image and video references

Feed the model character sheets, location stills, and product shots — plus video clips whose motion, camera moves, or effects you want the new shot to inherit.

@-reference control

Cite references directly in the prompt — "@Image1 walks into the cafe from @Image2" — so the model knows exactly which material plays which role.

Character and style lock

Seedance 2.5's reference engine holds identity, wardrobe, set, and palette consistent across the whole clip — the hardest part of AI video, solved with references.

Up to 30-second takes

Generate 5 to 30 seconds in a single run — enough for a full scene, not just a moment — with no stitching artifacts between beats.

Synced audio included

Sound effects, ambience, and music are generated in sync with the visuals at no extra credit cost.

Draft cheap, finish clean

Iterate at 480p to stretch your credits, then rerun the winning prompt at 720p for delivery.

Four reference engines

Seedance 2.5 for the longest takes, Seedance 2.0 for 1080p/4K output, 2.0 Fast for cheap iteration, and MiniMax H3 for 4K with stereo audio — switch models in one click.

Reference to Video Pricing

Priced in credits, shared across every model on PixRoute.

ByteDance

Seedance 2.5

Reference to Video

QualityCredits / second5s clip
480p Draft57283
720p113565
Alibaba

Wan 2.7

Reference to Video

QualityCredits / second5s clip
720p24120
1080p36180
Alibaba

HappyHorse 1.1

Reference to Video

QualityCredits / second5s clip
720p34170
1080p45221
Google

Veo 3.1

Reference to Video

QualityCredits / second8s clip
720p96768
1080p96768
4K1441152
Google

Veo 3.1 Fast

Reference to Video

QualityCredits / second8s clip
720p36288
1080p36288
4K85677
Google

Gemini Omni Flash

Reference to Video

Credits / second5s clip
32160
MiniMax

MiniMax H3

Reference to Video

QualityCredits / second5s clip
480p Draft1260
768p1680
2K32160
4K40200
ByteDance

Seedance 2.0

Reference to Video

QualityCredits / second5s clip
480p Draft33165
720p73365
1080p165822
4K3731862
ByteDance

Seedance 2.0 Fast

Reference to Video

QualityCredits / second5s clip
480p Draft27131
720p58290

See credit packs and subscriptions →

Reference to Video FAQ

What is AI video to video generation?+

Video to video generation creates new footage guided by reference material instead of text alone. On PixRoute you upload reference images — characters, scenes, styles — cite them in your prompt as @Image1, @Image2, and Seedance 2.5 renders a new clip that keeps those references consistent throughout.

Is Reference to Video the same as video to video?+

They're related but distinct. Reference to Video generates a new shot guided by your reference images and clips. Video to Video on PixRoute is motion transfer — your character image performs the exact actions of a reference video, powered by Kling 3.0 Motion Control. If you want a new scene inspired by references, use this page; if you want to re-enact a specific performance, use Video to Video.

How is Reference to Video different from Image to Video?+

Image to video animates one image as the starting frame. Reference to Video uses up to 10 references as guidance rather than a literal first frame — the model composes a new shot where your referenced characters, scenes, and styles appear consistently, in whatever action you describe.

Can I upload a video as the reference?+

Yes. Upload MP4 or MOV clips alongside images — up to 10 clips totalling 30 seconds on Seedance 2.5, or 3 clips totalling 15 seconds on Seedance 2.0, 2.0 Fast, and MiniMax H3 — and cite them in the prompt as @Video1. The model carries over their motion, camera work, and effects. On Seedance models the clip duration is billed with the run (input video seconds count toward the total, with a 40% discount applied to the whole request).

How long can the generated video be?+

Up to 30 seconds per run on Seedance 2.5 — the longest single take on PixRoute — or 4–15 seconds on Seedance 2.0, 2.0 Fast, and MiniMax H3.

Why does Reference to Video cost more credits than Text to Video?+

The reference engine processes every reference image alongside the generation, which is substantially more compute than plain text-to-video upstream. Draft at 480p — half the credits of 720p — and rerun only your best prompt at full quality.

Is there a free trial?+

PixRoute includes free starter credits with no card required — they cover image generation. Reference-to-video runs start around 131 credits (Seedance 2.0 Fast, 480p, 5 seconds); credit packs and subscriptions unlock video runs. Start with a short 480p draft to test the reference engine cheaply.

Models Behind It

Ready to create with Reference to Video?

One credit balance across every top image and video model.

Start Reference to Video Free

Free credits to start. No card required.