Wan 3.0 Pro Prime · Image To Video:
This ComfyUI node uses Alibaba’s Wan 3.0 Pro Prime model through SpicyAPI to turn a still image into a short video. It is suited to animating artwork, developing a moving shot from a composition, or creating a transition toward a chosen ending image. You can guide the subject’s movement, scene activity, and camera behavior while choosing the clip’s format and whether to generate audio. The model renders natively up to 1080p; its 2k and 4k tiers enlarge that render rather than generating those sizes natively. Generation runs as a paid remote task, so the node requests a cost quote before submission and then waits for the finished media.
Wan 3.0 Pro Prime · Image To Video Input Parameters:
prompt
Describe what should happen in the video, especially the subject’s movement, changes in the scene, and camera motion. For example, you might describe a portrait blinking as the camera slowly moves closer. This is a multiline STRING input with an empty default. Enter a meaningful prompt: the model schema allows up to 20,000 characters and specifies at least one character for a supplied prompt. If you leave the field blank, the node omits it from the request, giving the model no written motion guidance.
image_url
Connect the IMAGE that should serve as the opening frame. This is a required input with no default; generation cannot proceed without it. The node uploads the connected image before submitting the task. If you connect an image batch, it uses only the first image, so select the intended frame before connecting it.
resolution
Choose the output resolution tier: 1080p, 2k, or 4k. The default is 1080p. Wan 3.0 Pro Prime’s native render tops out at 1080p; 2k and 4k enlarge that render and should not be interpreted as additional native detail. Resolution affects the cost quote, so choose a tier that fits how you plan to display the clip.
aspect_ratio
Choose the output’s shape from 16:9, 4:3, 1:1, 3:4, 9:16, or adaptive. The default is adaptive, which follows the input media and the intended result. Choose a fixed ratio when the clip must fit a particular canvas, such as 9:16 for a vertical composition or 16:9 for a wide one.
duration_seconds
Set the clip length using Smart or a whole number from 2 through 30 seconds. The default is 5 seconds. A fixed number requests that duration; Smart sends -1 to let the model choose a length from the prompt and input media. Duration affects the cost quote, so a short fixed length is useful when you need to control the scope of a run.
generate_audio
Choose whether to include a generated audio track in the video. This BOOLEAN input defaults to true. Turn it off when you plan to add your own sound or need a silent clip. It does not create a separate audio output socket.
enable_prompt_expansion
Choose whether the service expands your written prompt before generation. This BOOLEAN input defaults to true. Leave it on when you want the model to elaborate on brief direction, or turn it off when you want to submit your wording without that expansion step. The provided source does not specify exactly how an expanded prompt will be rewritten.
seed
Set the random seed as a whole number from 0 to 2,147,483,647, in steps of 1. The default is 0, and the node initially keeps it fixed. Keeping the seed and other settings the same is intended to make repeat runs consistent; changing it requests a variation. You can set ComfyUI’s “control after generate” behavior to randomize for a new seed on each run, but each newly executed paid run can be billed.
last_image_url
Optionally connect an IMAGE to guide the clip’s final frame. There is no default; leave it unconnected when you do not need a specified ending. The catalog describes the clip as ending exactly on this image. This input works alongside the required image_url opening frame, making it useful for a planned transition between two compositions. As with the opening image, the node uses only the first image if you connect a batch.
Wan 3.0 Pro Prime · Image To Video Output Parameters:
video
The generated VIDEO is the finished clip for previewing, saving, or passing to another compatible ComfyUI node. When a task succeeds, the node downloads the first returned video asset and converts it into a ComfyUI video value. If generate_audio is enabled, any included audio is part of this video rather than a separate output. A task that finishes without a video asset raises an error instead of returning an empty clip.
task id
This STRING identifies the SpicyAPI generation task. Keep it when checking console logs or contacting support about a run, particularly if generation fails or you stop waiting for a task that may still be running remotely. It is an identifier, not a video link or a second media result.
Wan 3.0 Pro Prime · Image To Video Usage Tips:
- Start with one clearly chosen opening image and a prompt that names the subject, its motion, and the camera movement. For example: “The dancer turns slowly while the camera holds a steady medium shot.”
- Use
last_image_urlwhen the ending composition matters. Choose opening and ending images that support the transition you want, then describe the movement between them inprompt. - Begin tests at
1080pand a short fixedduration_seconds, such as5. Move to a longer clip or an enlarged output tier after the motion and framing work for your project; both duration and resolution affect the quote. - Keep
aspect_ratioonadaptivewhen you want the model to follow the image and prompt. Select a fixed ratio before generating when you have a specific delivery format. - Keep
seedfixed while comparing prompt or setting changes. Change or randomize it when you deliberately want another variation; an executed new run can incur a charge. - Check the cost quote and your per-run limit before submission. Stopping ComfyUI from waiting after task creation does not necessarily stop the remote task or its charge.
Wan 3.0 Pro Prime · Image To Video Common Errors and Solutions:
No SpicyAPI key is set.
- Explanation: The node cannot request a quote or submit a generation without an API key. The source message directs you to Settings > SpicyAPI or the
SPICY_API_KEYenvironment variable. - Solution: Add your key in SpicyAPI settings or set the environment variable, then run the workflow again.
A required opening image is missing
- Explanation:
image_urlis required by the model. With no connected image, the node has no opening frame to upload. - Solution: Connect an IMAGE to
image_urland rerun the node.
The prompt is empty or is rejected as too long
- Explanation: The model schema specifies a supplied prompt of 1 to 20,000 characters. The node omits a blank prompt, which also leaves the model without written motion direction; the service may reject a prompt that exceeds its limit.
- Solution: Enter a concise description of the subject, motion, and camera, and keep it within 20,000 characters.
An input image is too large to upload even as a JPEG
- Explanation: The node first encodes the image as PNG, then tries several JPEG quality settings if it exceeds its 10 MiB image-upload ceiling. If none fits, it raises an error before generation.
- Solution: Scale down or simplify the connected image and try again.
This run could cost up to … above your limit … Nothing was charged.
- Explanation: The pre-submission quote exceeds the maximum cost per run set in SpicyAPI settings. The node stops before creating the task.
- Solution: Try a shorter fixed
duration_secondsor a lowerresolution, then review the new quote. Change your spending limit only if the quoted maximum is acceptable.
No deployment serves this parameter combination.
- Explanation: This is the source’s recovery guidance for API error
40004. The service cannot run the requested combination, even if the individual controls appear valid. - Solution: Check which parameter the full error message names, change that setting, and retry. If necessary, choose another model.
This model has no usable deployment right now.
- Explanation: API error
50301indicates a temporary service-side deployment problem; it does not establish that your image or prompt is faulty. - Solution: Try again later or use another model.
The generation failed upstream and was refunded.
- Explanation: API error
50302reports that the upstream generation failed and was refunded. - Solution: Run the node again. If failures repeat, retain the task id and any request id shown in the error for support.
The uploaded file did not match its upload ticket.
- Explanation: API error
40003means the uploaded image did not match the authorization issued for that upload. - Solution: Run the node again so it re-uploads the image. If it persists, try a smaller image and check your network connection.
task … finished without a video
- Explanation: The remote task reported success, but the node found no downloadable video asset in its result.
- Solution: Check the task in SpicyAPI console logs using its task id, then retry or provide that id to support.
task … did not finish within … seconds
- Explanation: The node stopped waiting before the remote task reached a usable result. The task may still be running on SpicyAPI.
- Solution: Look up the task id under console Logs before starting another run, so you can check its status and avoid an unnecessary duplicate task.
Stopped waiting. Task … keeps running on SpicyAPI.
- Explanation: ComfyUI processing was interrupted after submission. Stopping the local wait does not cancel the remote task; it can still be charged if it succeeds.
- Solution: Check the task id in console Logs for its outcome before generating again.
network request to SpicyAPI failed
- Explanation: A request to the service could not complete after the client’s retries. This can happen during quoting, submission, or status checks.
- Solution: Check your connection and proxy settings, then retry. If a task id was already shown, check that task in console Logs before resubmitting.
SpicyAPI answered HTTP … with a body that is not JSON
- Explanation: The source notes that a proxy or firewall may be intercepting the API response.
- Solution: Check whether your network, proxy, or firewall permits access to the SpicyAPI API, then try again.
