OpenRouter Simple (Text Output):
OpenRouterSimple sends your written instructions and optional image, video, or audio context to a compatible OpenRouter model and returns a text response. It checks that the selected model accepts the kinds of context you provide, prepares connected media within a time limit, and reports details that help you understand the request and its result. You can use it to describe visual material, work with supplied media as context for a prompt, or incorporate model-generated writing into a ComfyUI workflow.
OpenRouter Simple (Text Output) Input Parameters:
model
Selects the OpenRouter model that handles the request. The choices come from a cached model catalog; until choices are available, the menu may show “Loading OpenRouter models…”. There is no fixed default model ID in the source: the first available menu entry is the default, and you must choose a compatible model before queueing. The node checks the current catalog again when it runs and rejects a model that is no longer listed or does not accept every connected media type.
reasoning_effort
Requests a reasoning-effort setting when the request is built for the selected model. Options are auto (default), none, minimal, low, medium, high, xhigh, and max. Use auto when you do not need to specify an effort level; other choices express the level you want. The provided source does not establish that every model supports every setting, so check the request details in info if a setting matters to your workflow.
timeout_seconds
Sets the overall time allowed for this node’s work, including catalog lookup, media preparation, the model request, and any subsequent credit lookup. The default is 120 seconds; the minimum is 1 and the maximum is 3600. A short limit can stop a request before it finishes, particularly when media needs preparation, while a longer limit gives complex requests more time.
temperature
Sets the requested temperature for the model response. The default is 1.0; the minimum is 0.0 and the maximum is 2.0, in steps of 0.01. Lower values are generally useful when you want more consistent wording, while higher values request more variation. The effect depends on the selected model; the node passes this setting into its request-building step.
max_tokens
Sets the requested maximum number of tokens for the response. The default is 4096; the minimum is 1 and the maximum is 1,000,000. A token is a unit of text rather than necessarily a whole word. Choose a smaller limit for brief captions or answers and a larger one when you need a long response, keeping in mind that the selected model may impose its own limits.
response_format
Chooses the requested response format: text (default) or json_object. Use text for ordinary prose, such as a description or caption. Use json_object when you need a JSON object for a later workflow step, and tell the model what fields you want in your prompt. Whether a model can honor the requested format depends on its capabilities.
zdr
Sets the requested ZDR setting passed to the request-building step. The default is False; the options are True and False. ZDR relates to zero-data-retention handling, but the provided node source does not establish how the selected model or provider applies it. If retention requirements are important, verify the applicable OpenRouter and provider policies rather than treating this switch alone as a guarantee.
regenerate
Controls whether ComfyUI considers the node changed for execution purposes. The default is True; the options are True and False. With True, the node requests a fresh execution when the workflow runs. With False, ComfyUI may reuse a cached result when nothing relevant has changed. This setting does not instruct the model to rewrite an answer within a single request.
system_prompt
Provides the overarching written instructions for the model. The default is “You are a helpful assistant.” This is a multiline string with no stated length limit. Use it to establish the model’s role or a consistent style, such as asking for concise, neutral descriptions across many images.
user_prompt
Provides the main task or question for the model. The default is “Describe the supplied context.” This is a multiline string with no stated length limit. Be specific about the result you want—for example, ask for a one-sentence scene description or a list of visible details—so the response is useful for your next workflow step.
image
Optionally connects an IMAGE for the model to consider alongside your prompts. There is no default image or stated minimum or maximum size in this interface. Connecting it makes image support a requirement for the selected model; the node prepares the image before sending the request.
image_2
Optionally connects a second IMAGE as additional context. There is no default image or stated size range. Like image, it requires the selected model to accept image input; explain in your prompt how you want the model to use multiple images.
image_3
Optionally connects a third IMAGE as additional context. There is no default image or stated size range. Use it when another visual reference is relevant, and choose a model that accepts image input.
video
Optionally connects a VIDEO for the model to consider. There is no default video or stated duration or size limit in this interface. Connecting it requires a model that accepts video input, and preparing it uses part of the node’s overall time limit.
video_2
Optionally connects a second VIDEO as additional context. There is no default video or stated duration or size range. Choose a model that accepts video input and describe the comparison or analysis you want in your prompt.
video_3
Optionally connects a third VIDEO as additional context. There is no default video or stated duration or size range. Additional media must be prepared within the same overall time limit and accepted by the selected model.
audio
Optionally connects AUDIO context. There is no default audio or stated duration or size limit in this interface. Connecting it requires a model that accepts audio input; ask for the specific text response you need, without assuming the node produces an audio output.
audio_2
Optionally connects a second AUDIO input. There is no default audio or stated duration or size range. It adds audio context to the request and requires the selected model to accept audio input.
audio_3
Optionally connects a third AUDIO input. There is no default audio or stated duration or size range. Use your prompt to make clear what you want the model to do with the supplied audio context.
api_key
Optionally supplies an OpenRouter API key as a single-line string. The default is an empty string; no length range is stated. If you leave it blank, the node can use OPENROUTER_API_KEY or the legacy LLM_KEY from the ComfyUI environment. A key entered here is saved with the workflow, so consider using an environment variable if you plan to share the workflow.
OpenRouter Simple (Text Output) Output Parameters:
text
The model’s text response as a STRING. This is the main result to connect to nodes that display, save, or further process writing. Its content depends on your prompts, connected media, selected model, and request settings; even when you request json_object, this socket remains a string.
info
A STRING containing a compact JSON record about the completed request. It includes the selected model, response ID, reported usage, required input modalities, information about prepared media, request-parameter information, model-catalog status and timestamp, and elapsed time in seconds. Use it to inspect what happened or pass the details to a later step that can read JSON; it is metadata, not the model’s answer.
credits
A STRING containing the result of a post-request credit lookup. It is separate from the model’s response and may instead say “Credits not checked (node deadline exhausted)” when insufficient time remains. Do not interpret that message as a failed model response; inspect text and info for the request result.
OpenRouter Simple (Text Output) Usage Tips:
- Start with your task in
user_promptand keep reusable style or role instructions insystem_prompt. For example, ask for “a concise description of the visible scene” when connecting an image for captioning. - Choose a model that accepts every connected type of context. Connecting an image and audio together requires a model that accepts text, image, and audio; disconnect unused media if a model does not support them.
- Leave
reasoning_effortatautofor a first attempt, and useresponse_formatset totextfor ordinary writing. If you choosejson_object, specify the desired object fields in your prompt and inspect the returned text before relying on it downstream. - Use a modest
max_tokensfor short descriptions and increase it for detailed writing. If media preparation or generation times out, raisetimeout_secondswithin its 3600-second maximum or reduce the connected context. - Set
regeneratetoTruewhen you want a new request on each run; set it toFalsewhen reusing an unchanged result is preferable. For shareable workflows, leaveapi_keyblank and configure an environment key instead. - Check
infofor usage, media details, request-parameter information, and catalog warnings. Checkcreditsseparately; its lookup happens after the model response and has only the remaining deadline time available.
OpenRouter Simple (Text Output) Common Errors and Solutions:
Choose a compatible OpenRouter model before queueing
- Explanation: The model choice is empty or is a placeholder rather than a selectable model.
- Solution: Wait for model choices to load, then select a compatible OpenRouter model before running the workflow.
Enter or connect api_key, or set OPENROUTER_API_KEY (or legacy LLM_KEY) in the ComfyUI environment
- Explanation: The node could not find an API key in the input or supported environment variables.
- Solution: Connect or enter a valid key, or configure
OPENROUTER_API_KEYin the ComfyUI environment. Remember that a key typed into the node is saved with the workflow.
model '…' is absent from the current OpenRouter catalog
- Explanation: The selected model was not found when the node checked the current catalog, even if it previously appeared in the menu.
- Solution: Refresh or reopen the model choices and select an available model, then run again.
model '…' does not accept every connected modality (…)
- Explanation: The selected model cannot accept at least one connected input type, such as image, video, or audio.
- Solution: Select a model that supports all connected types, or disconnect the unsupported media inputs. The message lists the types the request requires.
OpenRouter node timed out after … seconds
- Explanation: The node exceeded its configured deadline during catalog lookup, media preparation, the model request, or another time-limited step.
- Solution: Increase
timeout_secondsand try again. If the request includes several media connections, try fewer inputs to reduce preparation time.
unsupported media input slot: …
- Explanation: Media preparation encountered a slot type it does not recognize. The listed node inputs use the supported image, video, and audio slot names, so this is not expected in a normal workflow.
- Solution: Check that you are using this node’s listed media sockets. If the error persists with valid connections, update or report the node implementation with the full error message.
OpenRouter request or media preparation fails with another error
- Explanation: The provided source calls separate catalog, media, payload, and network implementations whose full error messages are not shown here; a failed service request, unsupported setting, or unsuitable media may surface as an error from one of those steps.
- Solution: Read the complete error message first. Confirm your API key and model choice, try the default
textandautosettings, disconnect media to isolate a problematic input, and retry if the issue appears temporary.
