Qwen - Image to Image:
This node uses KIE’s Qwen image-to-image service to create a new image from an existing one. It is suited to edits such as changing a scene’s season, adjusting its style, or replacing an element while describing what should remain recognizable. The node submits the generation task, waits for a result, and makes the finished image and task details available in your ComfyUI workflow. KIE’s example shows a winter scene changed to spring while retaining its composition; the degree to which a particular detail survives still depends on the image and instructions. (docs.kie.ai (https://docs.kie.ai/cn/market/qwen/image-to-image))
Qwen - Image to Image Input Parameters:
prompt
Required STRING. Describe the image you want to produce and, for a targeted edit, state what should stay unchanged. For example: “Change the snowy landscape to spring; keep the cabin, lake, mountains, and camera angle.” A specific prompt gives the model clearer direction than a general request such as “make it better.” The supplied node interface does not state a default, minimum length, or maximum length.
image_url
Required IMAGE socket. Connect the source image that the model will transform. Despite the name image_url, the authoritative interface identifies this as an image connection, not a field where you type a URL. Choose an image whose subjects and details are clear enough to guide the edit. No default, image-size limit, or supported file-format list is specified in the supplied interface.
strength
Optional FLOAT. Controls the strength of the image-to-image transformation, so it is a useful setting to adjust when the result changes too little or departs too far from the source. Change it gradually and compare results from the same source image. The supplied interface does not specify a minimum, maximum, default, or a guaranteed relationship between a particular number and the amount of change; the value 0.8 in KIE’s request example is an example, not an established node default. (docs.kie.ai (https://docs.kie.ai/cn/market/qwen/image-to-image))
output_format
Optional COMBO[STRING]. Selects the generated image’s file format, which matters when you plan to save or use the result outside ComfyUI. Choose from the options displayed by the node in your installation. The supplied interface does not list those options or a default; KIE’s example uses png, but that does not establish the full option list or the node’s default. (docs.kie.ai (https://docs.kie.ai/cn/market/qwen/image-to-image))
acceleration
Optional COMBO[STRING]. Selects a provider-exposed acceleration setting for generation. Use the choices displayed by the node, and compare their results if processing time matters to your workflow. The supplied interface does not list the available choices, their performance effects, or a default; none appears in KIE’s example request only. (docs.kie.ai (https://docs.kie.ai/cn/market/qwen/image-to-image))
negative_prompt
Optional STRING. Describes qualities you would prefer the generated image to avoid, such as “blurry” or “distorted.” Keep it focused on a few unwanted qualities and use the main prompt to describe what you do want. The supplied interface states no default or length limit, and it does not guarantee that every unwanted quality will be eliminated. KIE’s example includes a negative prompt. (docs.kie.ai (https://docs.kie.ai/cn/market/qwen/image-to-image))
num_inference_steps
Optional FLOAT. Sets a generation-step parameter sent to the model. It may be useful when comparing the model’s available generation settings, but the supplied source does not establish how changing it affects quality or processing time. No minimum, maximum, or default is provided; the 30 shown in KIE’s example is not a confirmed node default. (docs.kie.ai (https://docs.kie.ai/cn/market/qwen/image-to-image))
guidance_scale
Optional FLOAT. Adjusts a model guidance parameter associated with how the generation follows your instructions. If you experiment with it, change one setting at a time so you can judge its effect on your image. The supplied interface provides no minimum, maximum, default, or guaranteed interpretation of specific values; KIE’s example uses 2.5 without identifying it as this node’s default. (docs.kie.ai (https://docs.kie.ai/cn/market/qwen/image-to-image))
enable_safety_checker
Optional BOOLEAN. Controls a safety-checker setting passed with the generation request. A checker may affect whether a request or result is accepted, but the supplied source does not define this setting’s exact checks or its relationship to nsfw_checker. No default is specified in the interface; KIE’s example sets it to true for that request. (docs.kie.ai (https://docs.kie.ai/cn/market/qwen/image-to-image))
seed
Optional INT. Supplies a fixed random seed so you can make more controlled comparisons while adjusting other settings. Keep the source image, prompt, and settings the same when testing whether a seed helps reproduce a result; identical results are not guaranteed by the supplied source. The interface does not specify a minimum, maximum, or default.
nsfw_checker
Optional BOOLEAN. Controls a separate checker setting exposed by this node. The supplied interface does not define precisely how it differs from enable_safety_checker, whether the provider applies it for this model, or what its default is. Leave it at the node’s displayed setting unless you have a specific reason to change it, and do not assume either switch overrides the provider’s content policies.
timeout_seconds
Optional INT. Sets how long the node waits for the submitted task to finish before timing out. The supplied source specifies a default of 1200 seconds, a minimum of 30 seconds, a maximum of 7200 seconds, and a step of 30 seconds. Increase it when valid tasks regularly take longer than your current limit; a longer timeout allows more waiting but does not make generation faster.
callback_url
Optional STRING. Provides a URL for KIE to notify when a task finishes. Its tooltip says to leave it empty for normal ComfyUI use, because this node already waits for the result. Enter a URL only if you operate a service that can receive callbacks. The supplied source specifies an empty-string default and no length limit. KIE documents callbacks for its task API. (docs.kie.ai (https://docs.kie.ai/cn/market/qwen/image-to-image))
Qwen - Image to Image Output Parameters:
image
IMAGE. The generated image, ready to connect to another ComfyUI image-processing or preview node. This is the most direct output for visually assessing whether the requested change worked. It represents the returned result, not the original source image.
url
STRING. A URL associated with the generated result. It is useful when another step needs a link rather than a ComfyUI image connection. Treat it as a result location; the supplied source does not specify how long that URL remains available.
all_urls_json
STRING. A JSON-formatted record of result URLs. Use it when you need to inspect or pass along the complete set of returned links instead of relying only on url. The supplied source does not guarantee how many links a task will produce.
task_id
STRING. The identifier assigned to the submitted KIE task. Keep it when investigating a failed or delayed generation, or when you need to look up that particular task’s status through KIE’s task system. It identifies the job, not the image content. (docs.kie.ai (https://docs.kie.ai/cn/market/qwen/image-to-image))
raw_json
STRING. The provider response rendered as JSON text. It is chiefly a troubleshooting output: inspect it for task details or an error explanation when the visible image alone does not tell you what happened. Its exact fields can depend on the response.
credits_consumed
FLOAT. The reported credits used for this execution. Check it when comparing the cost of runs or recording usage. Interpret it as the node’s reported usage value, rather than assuming a particular price or billing formula.
credits_left
FLOAT. The reported remaining credit balance after the execution’s accounting. It helps you judge whether you have enough balance for further runs. The supplied source does not specify when account balances refresh or how concurrent tasks affect this figure.
Qwen - Image to Image Usage Tips:
- Start with a clear source image and a prompt that names both the change and the elements to preserve. For a seasonal edit, specify the new season and explicitly retain the composition or important objects. (docs.kie.ai (https://docs.kie.ai/cn/market/qwen/image-to-image))
- For a restrained edit, try adjusting
strengthgradually while keeping the image and prompt fixed. Compare the results rather than assuming that one value works for every image. - Use
negative_promptfor a short list of unwanted qualities, while keeping the desired subject, scene, and edit inprompt. This makes each instruction easier to interpret. - Change only one of
strength,num_inference_steps,guidance_scale, oraccelerationbetween comparison runs. Because the supplied source does not define their numeric ranges or performance tradeoffs, rely on the controls shown by your installed node and the results you observe. - Leave
callback_urlempty for an ordinary ComfyUI workflow. If a task is timing out but later finishes, increasetimeout_secondswithin its documented 30–7200-second range. - Save
task_idand inspectraw_jsonwhen troubleshooting. They provide a more useful starting point than repeatedly submitting the same request without checking what happened.
Qwen - Image to Image Common Errors and Solutions:
Required input is missing: prompt or image_url
- Explanation: The node interface requires both an instruction and a connected source image; an unconnected required socket can stop execution before a task is submitted.
- Solution: Enter a nonempty
promptand connect anIMAGEoutput toimage_url. Do not type a web address into the image socket.
KIE model ID has not been resolved yet. Refresh the live catalog once.
- Explanation: This exact error in the supplied source occurs when the node reaches the task-service path but cannot determine the model identifier needed to submit the job.
- Solution: Refresh KIE Next’s live catalog once, then retry. If it persists, check that the installed extension and its catalog data are available and up to date.
Invalid or unsupported parameter value
- Explanation: The node validates inputs before submission, and the provider can also reject settings it does not accept. The supplied interface does not state the numeric limits or full dropdown options for several model settings, so a value from another Qwen model may not work here.
- Solution: Use this node’s displayed dropdown choices, restore unfamiliar settings to their displayed values, and retry with only
promptandimage_urlbefore adding optional controls one at a time.
Image upload or source-image retrieval failed
- Explanation: The image must be made available to the generation service. An unreadable, unsupported, or inaccessible image can prevent processing; the supplied source does not give this node’s exact image-upload error wording or size limit.
- Solution: Confirm that the connected image loads in ComfyUI, try a conventional PNG or JPEG source, and check the detailed error or
raw_jsonfor a provider-specific requirement before resizing or converting it.
Authentication failed or API key is invalid
- Explanation: KIE’s protected task endpoint requires authorization. The exact error text depends on the service response. (docs.kie.ai (https://docs.kie.ai/cn/market/qwen/image-to-image))
- Solution: Check that KIE Next has a valid KIE API key configured and that the associated account has access to the service. Keep the key private.
Insufficient credits
- Explanation: The account may not have enough available credits to run the task. The node exposes
credits_leftwhen a balance is reported, but a rejected request may not produce normal outputs. - Solution: Check your KIE account balance and usage, then retry after sufficient credits are available. Avoid repeated submissions while a task’s status is uncertain.
Rate limit or temporary service error
- Explanation: The provider may decline a burst of requests or encounter a temporary processing failure. The supplied source does not define an exact error string for this node.
- Solution: Read the returned error details, reduce simultaneous runs, and retry after a pause. If the same task repeatedly fails, simplify the request and retain the
task_idfor support.
Task timed out or KIE task ended without a result.
- Explanation: The node waits for completion, and the supplied source includes the exact “ended without a result” error if execution reaches that state. A timeout can also occur when the task takes longer than
timeout_seconds; neither condition by itself proves that the provider never accepted the task. - Solution: Check any available
task_idor task history before resubmitting. If processing is simply slow, raisetimeout_secondswithin its documented range and retry after checking the existing task’s status.
Callback URL is invalid or unavailable
- Explanation: A nonempty
callback_urlis intended for a reachable callback receiver. An unsuitable URL can complicate a workflow that does not need callbacks; the supplied source does not specify the provider’s exact rejection message. - Solution: Leave
callback_urlempty for normal ComfyUI use. If you deliberately use callbacks, confirm that the destination is a valid, reachable endpoint that can receive the provider’s notification.
