To check a ComfyUI workflow after FlipSigmas, compare the input and output sequence order, inspect the first and last values, and verify which sequence is connected to the next node. The official documentation says the node reverses the input SIGMAS and ensures the first output value is non-zero when the original sequence ends with zero. It does not specify a general sampling-quality effect or explain the adjustment algorithm, so treat this as a sequence-boundary check, not a quality prediction.
Trace direction and boundary values
Start when a downstream stage is meant to receive sigma values in reverse order, or when the visible result makes you unsure which direction the workflow passes. Find FlipSigmas in the graph and follow its sigmas input and output. Use the port’s SIGMAS type label to confirm the connection; node color alone is only a visual clue.
- Record the upstream sequence’s first and last values, and note whether its last value is zero. If the values are not visible in your workflow, use an available sequence inspection method; do not infer the values from the node’s color or the rendered image.
- Confirm that the upstream wire enters the
sigmasinput onFlipSigmasand that its output is connected to the intended downstreamSIGMASinput. - Compare the output order with the input in reverse. Check the endpoints explicitly rather than assuming that a correctly connected node guarantees the intended route through the whole graph.
- If the input ends in zero, check whether the output begins with a non-zero value, as described in the documentation. Record what you observe; do not guess the replacement value or how many entries are changed.
- Confirm that the downstream stage is supposed to consume this flipped sequence, rather than the original upstream sequence or a separate branch.
ComfyUI documents the behavior and the SIGMAS input/output type on the FlipSigmas node page. The official node index can help locate related built-in node documentation if you need to check another stage’s documented inputs. Neither source establishes that using the flipped sequence will improve an image.
Record the check in a worksheet
The following values are a hypothetical inspection example, not a promise about exact node output or a report of a software test. Here, A, B, and C stand for distinct non-zero values; 0 is a literal zero. The table shows what to record and what decision follows from the observation.
| Check | Hypothetical observation | Decision to record |
|---|---|---|
| Input endpoints | First A; last 0; zero present at the end | Expect a reverse-order check and note the zero boundary |
| Output endpoints | First is non-zero; last is A | Record the observed boundary; do not infer the adjusted value |
| Direction | Remaining observed values follow reverse order where inspectable | Mark direction as confirmed only against the available sequence evidence |
| Downstream receiver | Flipped output wire reaches the intended SIGMAS input | Record that this stage receives the transformed sequence |
For your own graph, replace the symbols with observed values or a clear note that they are not inspectable. Keep expected direction separate from observed direction. This makes a mismatch actionable: you can identify whether the uncertainty is the sequence values, the zero boundary, or the wire to the downstream receiver.
Limits and what to do when it fails
The documentation describes reversal and a non-zero first output when the original sequence ends in zero. It does not specify the replacement value, the sequence length, an indexing convention, a complete numerical example, or a guaranteed change to the diffusion result. Do not fill those gaps with assumptions about sampling behavior.
If the endpoints do not match the expected direction, first recheck the actual SIGMAS ports and trace the wire from the node output to the receiving input. If the sequence itself is not inspectable, mark direction and boundary as unverified rather than treating a plausible image as proof. If the input does not end in zero, do not expect the documented zero-boundary adjustment to apply. If a zero-ending input appears to produce a zero first output, capture the input and output observations and consult current ComfyUI documentation or the node’s maintained issue or contribution channel; the page notes that its content is AI-generated. Keep the original and transformed routes distinct while diagnosing, so you can tell which one the downstream node actually receives.
Common follow-up questions
Does a non-zero first output prove the whole sequence is correct?
No. It checks one documented boundary condition. Verify the reverse ordering and downstream connection separately.
Should I expect the same values in reverse, including a trailing zero?
Do not assume that. The documentation describes a first-value adjustment when the original sequence ends in zero, but does not state the exact resulting values. Record what is observable.
Does FlipSigmas improve image quality?
The cited documentation explains sequence manipulation, not a quality guarantee. Choose the direction required by the workflow and verify the connected input; assess the rendered result separately.
Sources and next step
This guide interprets public documentation and uses explicitly hypothetical acceptance examples, not application tests. Controls and availability can change; verify the cited manual against your installed version.
For work created in Imagild, carry the input requirements and acceptance checklist into the photo editor. The external software controls discussed above belong to their respective products.