Wan AI 2.7 Reference-to-Video: Image to Video Online
Bottom line
Wan 2.7 Reference-to-Video uses your photo as an identity reference, not a first frame: 5 or 10 second clips at 720p or 1080p from 32 credits. Upload a starting frame, choose Wan AI 2.7 Reference-to-Video in the Image to Video tool, and animate it on VdoBloom. The credit price appears before every generation.

Wan 2.7 Reference-to-Video (R2V) generates a new clip around a subject you supply as a reference, instead of animating your photo as the opening frame. The photo tells the model who or what must appear (a face, an outfit, a product); the prompt tells it what happens and where. That split is what you want for series work: the same character in a kitchen, then on a street, then at a desk, across separate generations.
Clips are 5 or 10 seconds at 720p or 1080p. A 5-second clip costs 32 credits at 720p and 48 at 1080p; 10 seconds costs 64 or 95, the same rates as Wan 2.7 Image-to-Video. That is $1.60 to $4.75 at the Lite plan rate, the exact cost is shown before you generate, and a failed job is refunded automatically. No subscription is needed: a one-time 300-credit pack is $19.50 and never expires.
On the Image to Video tab, the photo you upload is sent as the reference. The same model also powers the Wan 2.7 Same Look option in Extend, where it keeps faces and outfits locked to your first clip. The sections below cover the spec table, worked costs, how R2V differs from its image-to-video siblings, where it runs, how to choose a reference photo and its limits.
Wan 2.7 Reference-to-Video specs at a glance
These figures come from the Wan 2.7 R2V endpoint, the service that sends it to the provider and VdoBloom's credit table. Two things differ from the other Wan 2.7 entries: the ceiling is 10 seconds, not 15, and the image is a reference rather than a frame.
| Spec | Wan 2.7 Reference-to-Video on VdoBloom |
|---|---|
| Input | A reference photo plus an optional text prompt |
| What the photo does | Defines the subject's identity and look; it is not used as the first frame |
| Clip length | 5 or 10 seconds in the composer; the API accepts 2-10 s and bills 5 or 10 |
| Resolutions | 720p, 1080p |
| Aspect ratio | No aspect-ratio control for this model |
| References accepted by the API | Up to 5 in total (images and videos combined); the Image to Video tab sends your one uploaded photo |
| Prompt length | Optional, up to 5,000 characters |
| Price | 32 / 64 credits at 720p; 48 / 95 at 1080p (5 / 10 s) |
| Content filter | Flexible (HOT label, shared by the Wan family); platform rules still apply |
| Provider | Hosted by VdoBloom through KIE; Alibaba has not released Wan 2.7 weights |
| Where it runs | Image to Video tab; Extend (as Wan 2.7 Same Look) |
What Wan 2.7 Reference-to-Video costs: worked examples
R2V shares Wan 2.7 Image-to-Video's per-clip prices for the lengths it offers. Dollar figures use the Lite plan rate of $15 for 300 credits ($0.05 a credit); bought once, the same 300 credits cost $19.50 and never expire, while plan credits reset each cycle. A 300-credit pack covers nine 5-second 720p clips, or six 5-second 1080p clips with credits left over. The 10 free signup credits do not reach the 32-credit minimum.
| Length | 720p | 1080p | Per second (720p / 1080p) |
|---|---|---|---|
| 5 s | 32 cr ($1.60) | 48 cr ($2.40) | 6.4 / 9.6 |
| 10 s (maximum) | 64 cr ($3.20) | 95 cr ($4.75) | 6.4 / 9.5 |
Read more: Credit packs and plans
Reference-to-video vs image-to-video: what the photo is for
In image-to-video, your photo is frame one: the clip starts on exactly that picture and the prompt describes the motion from there. In reference-to-video, the photo is a description of the subject: the model draws that person or product into a new shot, framed and posed however the prompt asks. Pick image-to-video when the composition of the photo is the shot. Pick R2V when the subject matters and the composition should change.
Within Wan, the siblings split like this: Wan 2.7 Image-to-Video costs the same per clip, runs to 15 seconds and adds a last-frame option so you can fix where a shot ends. Wan 3.0 accepts a photo too, runs any length from 2 to 30 seconds and generates audio natively. Wan 2.6 Image-to-Video animates a single starting image to 15 seconds at slightly higher prices.
| Model | Photo used as | Lengths | 5 s 720p | 10 s 1080p | Extra control |
|---|---|---|---|---|---|
| Wan 2.7 Reference-to-Video | Identity reference | 5 / 10 s | 32 cr | 95 cr | Subject held across new scenes |
| Wan 2.7 Image-to-Video | First frame (plus optional last frame) | 5 / 10 / 15 s | 32 cr | 95 cr | Fix the end frame |
| Wan 3.0 Video | First frame | 2-30 s | 32 cr | 128 cr | Native audio, 480p tier |
| Wan 2.6 Image-to-Video | First frame | 5 / 10 / 15 s | 35 cr | 125 cr | None |
Read more: Wan 2.7 Image-to-Video · Wan 3.0 Video · Wan 2.6 Image-to-Video · Best model for image to video, by constraint
Where to run Wan 2.7 Reference-to-Video on VdoBloom
R2V shows up in two places, both priced from the same table.
- Image to Video tab: pick Wan AI, then 2.7 Reference-to-Video, upload one photo of the subject and describe the scene. Choose 5 or 10 seconds and 720p or 1080p. The photo is sent as the reference image.
- Extend (the extend button on a finished video in My Creations): choose Wan 2.7 Same Look. It uses the last frame of your clip as the first frame and the clip's opening frame as the reference, so faces and wardrobe stay consistent in the added 5 or 10 seconds.
- Not on photo-effect tabs: those tabs animate the photo you upload, so reference-style models are left out of their pickers on purpose.
Read more: How to use Wan 2.7 on VdoBloom
Choosing a reference photo and writing the prompt
The reference does the identity work, so give it a clear subject: one person or one product, well lit, facing the camera or close to it, with the details you need kept (face, hair, outfit, label) visible and unobstructed. Busy backgrounds and group photos give the model more to misread. Crop to the subject before uploading.
Then let the prompt carry everything else: the setting, the action and the camera. Describe the scene as if the subject were already in it ("the woman walks through a rainy night market, camera tracking beside her") rather than re-describing her appearance, which can fight the reference. VdoBloom sends Wan's prompt-extend option switched on, so short prompts get expanded; write a full sentence or two if you want literal control. Draft at 720p and 5 seconds for 32 credits, and keep the same reference photo for every clip in a series.
Planning a series: what a set of clips costs
R2V is usually bought in sets, because the point is the same subject across several clips. Since each clip is a separate generation, a series costs the per-clip price times the number of clips; there is no batch discount and no setup charge for reusing a reference. Extend is the other way to build length around the same character: start with one clip, then add 5 or 10 seconds at a time with Wan 2.7 Same Look, each extension priced from the same table.
- Pick R2V when the subject must stay recognisable and each clip needs a new scene or pose.
- Pick Wan 2.7 Image-to-Video when the photo itself is the opening shot, or you need to fix the last frame too.
- Pick Wan 3.0 when one continuous take longer than 10 seconds matters more than reference locking.
- Draft every new reference at 720p and 5 seconds (32 credits) before committing the series to 1080p.
| Plan | 720p | 1080p | Fits in one 300-credit pack? |
|---|---|---|---|
| 6 clips x 5 s | 192 cr | 288 cr | Yes at both resolutions |
| 6 clips x 10 s | 384 cr | 570 cr | No; two packs, or the Lite plan plus a top-up |
| 1 clip of 5 s + two 5 s Anchored extensions (15 s total) | 96 cr | 144 cr | Yes |
| 1 clip of 10 s + one 10 s Anchored extension (20 s total) | 128 cr | 190 cr | Yes |
Limits and honest caveats
The ceiling is 10 seconds per generation; for a longer single shot use Wan 2.7 Image-to-Video (15 s) or Wan 3.0 (30 s), or chain clips with Extend. There is no aspect-ratio control. The provider API accepts up to five references, reference video and a reference voice, but the Image to Video tab sends one photo, so multi-subject scenes from several references are not available in the composer today. Identity is held by the model, not guaranteed: unusual angles, heavy occlusion or a low-resolution reference make drift more likely. The free 10 signup credits do not cover the 32-credit minimum. Failed jobs are refunded automatically, and a prompt stopped by the content check is stopped before any credits are taken.
VdoBloom does not rank models on subjective quality; the only reliable test of how well R2V holds your subject is a 32-credit draft with your own reference.
What creators use Wan AI 2.7 Reference-to-Video for
- A recurring mascot or presenter placed in a new scene per clip, from one clean reference photo.
- Product shots where the packaging must stay recognisable while the setting changes from clip to clip.
- Extending a finished clip with Wan 2.7 Same Look so the added seconds keep the same face and wardrobe.
- Story beats for a short series, each a separate 5 or 10 second generation built around the same character.
How to use Wan AI 2.7 Reference-to-Video on VdoBloom
- 1Open the Image to Video tool and select Wan AI 2.7 Reference-to-Video from the model picker.
- 2Upload your starting image and describe the motion you want in the prompt.
- 3Choose your duration (5–10s), quality (720p/1080p), review the credit cost, and generate.
Frequently asked questions
What is Wan 2.7 reference-to-video?
A Wan 2.7 mode that builds a new clip around a subject you provide as a reference photo. Unlike image-to-video, the photo is not the first frame; it tells the model who or what to show, and your prompt sets the scene and action. It is meant for keeping one character or product recognisable across separate clips.
How much does Wan 2.7 Reference-to-Video cost?
32 credits for 5 seconds at 720p and 64 for 10 seconds; at 1080p it is 48 and 95. That is $1.60 to $4.75 at the Lite rate of $0.05 a credit. The cost is shown before you generate and refunded if the job fails.
How long can a Wan 2.7 R2V clip be?
5 or 10 seconds. The API accepts 2 to 10 seconds but bills them as 5 or 10, so the composer offers only those two. For 15 seconds use Wan 2.7 Image-to-Video; for up to 30, Wan 3.0.
How is R2V different from Wan 2.7 Image-to-Video?
Image-to-video starts the clip on your photo (and can end on a second photo); R2V uses the photo only to define the subject and composes a new shot around it. They cost the same per clip at 5 and 10 seconds; image-to-video also offers 15 seconds.
Can I upload several reference images or a reference video?
The provider API accepts up to five references in total, images and videos combined, but the Image to Video tab sends one uploaded photo as the reference. Multi-reference and reference-video inputs are not in the dashboard composer today.
What is Wan 2.7 Same Look in the Extend tool?
It is this model used for extending a finished clip. Extend sends your clip's last frame as the starting frame and its opening frame as the reference, which keeps faces and outfits consistent in the added 5 or 10 seconds. It is priced from the same table: 32 or 64 credits at 720p, 48 or 95 at 1080p.
Can I choose the aspect ratio?
No, Wan 2.7 Reference-to-Video has no aspect-ratio control on VdoBloom. If the frame shape matters, Wan 2.7 Text-to-Video and Wan 3.0 both offer five aspect ratios.
What makes a good reference photo?
One clear subject, well lit, with the features you need kept (face, outfit, logo) visible and unobstructed, ideally cropped close. Then describe the scene and action in the prompt rather than re-describing the subject's appearance.
Can I try it with the free signup credits?
Not on their own: new accounts get 10 free credits and the minimum R2V render is 32. A one-time 300-credit pack is $19.50, never expires, and covers nine 5-second 720p clips.
Is Wan 2.7 open source?
No. Alibaba has not released Wan 2.7 weights. VdoBloom runs it through its hosted provider, so there is nothing to install and no API key; you pay in VdoBloom credits.
Wan AI 2.7 Reference-to-Video guides, prices and comparisons
This page is the Wan AI 2.7 Reference-to-Video hub. Each guide below goes deeper on one question and links back here.
How to use Wan 2.7
Text, image and reference modes, and when to pick each.
Best model for image to video
Pick by length, resolution, sound and end-frame control.
WAN vs Seedance
When reference images favour WAN, and when Seedance wins.
Image to video best practices
Getting clean motion and stable faces from a still photo.
Wan AI review (2026)
The Wan family on VdoBloom, version by version.