Seedance 2.5 image to video: your photo decides the aspect ratio
Here is the fact that saves the most wasted credits on this model. Select Seedance 2.5 in VdoBloom's Image to Video tab and the photo you upload becomes the clip's first frame. On Seedance 2.5 a first-frame job cannot carry a chosen aspect ratio at all — the upstream API returns a 400 reading "first-frame and first-last-frame tasks only support adaptive aspect ratio" — so the request is coerced to adaptive rather than failing after the credits are charged. The output then inherits its shape from the image you uploaded.
The practical consequence is simple and worth acting on before you generate anything: crop the photo to the ratio you want the video to be. Upload a 1080x1920 crop and you get a vertical clip suitable for Reels, Shorts and TikTok. Upload an untouched 4:3 phone photo and you get a 4:3 video, whatever the ratio control said. This coercion is specific to Seedance 2.5 — Seedance 2, 2 Fast and 2 Mini pass your chosen ratio through unchanged.
There is a way around it inside the same tab. With a Seedance model selected, the Images panel offers a Frames (I2V) mode and a Multimodal mode, and it opens on Frames. Frames is the first-frame path described above, so it is the one that goes adaptive. Multimodal sends your photos as reference images instead — up to nine — and reference images do not trigger the coercion, so the ratio you pick is the ratio you get. The Seedance Effects tab behaves the same way for the same reason: effect uploads travel as reference images, and the ratio you choose there is honoured.
What this tool does
- In VdoBloom's Image to Video tab a Seedance 2.5 photo upload becomes the job's first frame, and first-frame and first-and-last-frame jobs on 2.5 are forced to "adaptive" — the upstream API rejects every other value, so the request is coerced rather than failing after the charge.
- Under adaptive, the output video's shape is inherited from the uploaded image. Crop the photo to the ratio you want before you upload it: 1080x1920 for 9:16, 1920x1080 for 16:9, 1080x1080 for 1:1.
- The adaptive coercion is specific to Seedance 2.5. Seedance 2, Seedance 2 Fast and Seedance 2 Mini pass your chosen aspect ratio through unchanged.
- Seedance 2.5 is selectable in the Image to Video picker, in the same Seedance group as Seedance 2, Seedance 2 Fast, Seedance 2 Mini, Seedance 1.5 Pro and the ByteDance V1 image-to-video variants.
- The Images panel for Seedance models has two modes and opens on Frames (I2V). Frames takes a first frame plus an optional last frame — two images at most — and is what triggers the adaptive coercion on 2.5. Multimodal takes up to 9 reference images and does not, so your chosen ratio is used. The two modes cannot be combined in one request.
- The aspect ratio control in Text to Video and Image to Video offers six values — 16:9, 4:3, 1:1, 3:4, 9:16 and 21:9. Adaptive is not one of them; it is applied server-side on first-frame Seedance 2.5 jobs. The Seedance Effects tab offers all seven, adaptive included.
- For Seedance 2.5 on the Seedance Effects tab, when an effect's stored prompt contains no @Image tag, VdoBloom prepends deterministic reference mentions in the form "Reference @Image1 for the person's face and appearance; reference @Image2 for the scene and visual style."
- On that tab the @ImageN numbering is deterministic: your uploads are numbered first, then the effect's admin reference images. Your clearest face photo should therefore be the first file you upload.
- First-frame and last-frame inputs are mutually exclusive with the reference image, reference video and reference audio inputs — one job cannot mix the two modes. Reference images alone are capped at 9 per job.
- Aspect ratio does not change the price. Seedance 2.5 is billed on resolution and duration only: 9.2 credits per second at 480p and 20.0 at 720p, so a 10-second image-driven clip is 92 or 200 credits regardless of shape.
How it works
1.Crop the photo to your target ratio before anything else
Do this in whatever editor you already use. For a vertical Reel or Short, export 1080x1920. For YouTube or a landing page, 1920x1080. For a feed square, 1080x1080. Under adaptive this crop is what sets the output shape — you are not going to fix it afterwards with a dropdown.
2.Open Image to Video and pick Seedance 2.5
Go to /dashboard/video-creation/image-to-video/ and select Seedance 2.5 from the Seedance group in the model list. Doing so expands the duration grid to every second from 4 to 30 and narrows the resolution list to 480p and 720p. If you had 1080p or 4K selected on another Seedance variant, it resets to 480p.
3.Choose Frames or Multimodal deliberately
The Images panel opens on Frames (I2V): your photo becomes the first frame and you can add an optional last frame, two images at most. That is the mode that forces adaptive on 2.5, and the mode you want when the photo should be the literal opening shot. Switch to Multimodal if you would rather keep control of the aspect ratio — it sends up to nine photos as reference images instead. The two cannot be combined in one request, and switching modes clears the other one's uploads.
4.Write the prompt for the motion, not the look
This tab has a prompt box and Seedance 2.5 accepts up to 30,000 characters. The photo already sets the subject and the palette, so spend the prompt on what moves: the camera path, the performance beat, how the action resolves. Over a long take, say what happens at roughly which second — 30 seconds is a lot of time to fill with one instruction.
5.Set duration and resolution, then check the cost
An image-driven job runs the full 30 seconds if you want it: 276 credits at 480p, 600 at 720p. Draft at 480p to confirm the photo animates the way you expected, then re-render the approved take at 720p. Credits are deducted at submission and refunded automatically if the job fails.
6.Check the shape of what comes back
In Frames mode the returned clip matches your photo, not the ratio control. That is the coercion working as designed rather than a bug — which is why the crop in step one is the real setting. If the shape is wrong, re-crop and re-run; changing the ratio button will not do it.
Frequently asked questions
Why is my Seedance 2.5 video not the aspect ratio I chose?
Because your photo went in as the job's first frame, and on Seedance 2.5 first-frame and first-and-last-frame tasks are coerced to "adaptive" before the request is sent. The upstream API refuses any other value for those tasks on this model, so VdoBloom rewrites the field instead of letting a charged job fail on a 400. Adaptive means the output takes its shape from the source image, so the video matches your photo, not your selection. That is the default Frames (I2V) mode in the Image to Video tab; the Seedance Effects tab sends photos as reference images instead and honours the ratio you pick.
How do I get a vertical 9:16 Seedance 2.5 video?
Crop the source photo to 9:16 before you upload it — 1080x1920 is the safe export. In Image to Video's Frames mode the output follows the image, so a 9:16 upload produces a 9:16 clip. If you would rather set the ratio explicitly, switch the Images panel to Multimodal so your photos travel as reference images, or write the shot in Text to Video, where there is no image for the output to inherit from and the 9:16 button is simply obeyed.
Can I use Seedance 2.5 on the Image to Video tab?
Yes. Seedance 2.5 is in the Image to Video model list, in the Seedance group alongside Seedance 2, Seedance 2 Fast, Seedance 2 Mini, Seedance 1.5 Pro and the ByteDance V1 image-to-video variants. Upload a photo, write the prompt, set duration and resolution, and generate. The one thing to know before you start is the adaptive coercion: in the panel's default Frames mode your photo becomes the first frame and its shape becomes the video's shape.
How many photos can I upload?
It depends on the mode. In Image to Video's Frames mode, one photo as the first frame plus an optional last frame — two at most. In Multimodal mode, up to nine reference images. On the Seedance Effects tab the effect decides: each entry declares exactly how many photos it needs from you and adds its own admin reference images on top. The upstream cap for reference images is 9 either way, and frames cannot be mixed with references in a single job.
What do the @Image1 and @Image2 tags in Seedance 2.5 prompts mean?
They are positional references to the images attached to the job. On the Seedance Effects tab, when an effect's stored prompt has no @Image tag of its own and the model is Seedance 2.5, VdoBloom prepends a deterministic line such as "Reference @Image1 for the person's face and appearance; reference @Image2 for the scene and visual style." Your uploads are numbered first and the effect's reference images follow, which is why the order you upload in matters there.
Does uploading an image change what Seedance 2.5 costs?
No. The price is set by resolution and duration only — 9.2 credits per second at 480p and 20.0 at 720p — so an image-driven 10-second clip is 92 or 200 credits, the same as a prompt-only job of that length. There is a cheaper tier at 6.8 and 15.2 credits per second, but it applies only when a reference video is supplied and is billed on input plus output duration; a photo upload does not qualify.
Can I give Seedance 2.5 both a start and an end image?
Yes — Frames mode takes a first frame and an optional last frame, and that combination is exactly where the adaptive coercion applies, so crop both to the same ratio and the output will follow it. Note that this mode is mutually exclusive with the reference image, reference video and reference audio inputs: a single job uses one approach or the other, never both.
Related tools
Seedance 2.5: the 30-second model, and how to run it
ByteDance Seedance 2.5 renders 4-30 seconds at 480p or 720p only. Pick it in Text to Video, Image to Video or Effects. 30s at 480p is 276 credits.
Seedance 2.5 vs Seedance 2: duration against resolution
Seedance 2.5 runs 30s but caps at 720p; Seedance 2 stops at 15s and reaches 4K. 15s at 720p costs 300 vs 244 credits. The full maths.
Seedance 2.5 price: the full credit table, per second and per clip
Seedance 2.5 costs 9.2 credits/sec at 480p and 20 at 720p. Cheapest render is 37 credits (4s, 480p); a full 30s at 720p is 600 credits.
Best AI model for image-to-video: pick by constraint, not by hype
40+ image-to-video variants. Cheapest 720p: ByteDance V1 Pro Fast at 7 credits per 5s. Longest: Vidu Q3 at 16s. Highest: Gemini Omni at 4K.
AI Video Generator for Instagram Reels
Creator Reel renders 1080×1920 9:16 with word-timed captions burned in, in 30s/1/1.5/2-minute lengths and 6 languages including Hindi and Tamil.
Ready to try it?
Open Image to Video