rookiestar28-ComfyUI-Sigmax/docs/COMPATIBILITY.md

Compatibility

This page summarizes the supported user-facing boundary for tagged ComfyUI-Sigmax 1.0.0 and the current unreleased development additions described below.

Environment

Component	Supported boundary
Python	3.10 or newer
ComfyUI package requirement	0.29.0 or newer
General validated host baseline	ComfyUI 0.29.0
MiniMax H3 validated host baseline	ComfyUI 0.30.0 with the upstream H3 nodes
Operating systems covered by project gates	Windows and Linux/WSL
Mandatory additional Python packages	None
Host runtime dependency policy	Record the selected host's compatible package versions; current ComfyUI-recommended comfy-aimdo versions (including 0.4.13) are accepted without an exact-version gate

A newer ComfyUI version may work, but is not automatically promoted to the validated host baseline; reproduce workflows on that baseline first after an update.

Supported model profiles

Profile	Supported schedule	User selection
Krea 2 Turbo	Fixed-shift external sigma schedule, 8-step official recipe	Select Turbo explicitly when model evidence is ambiguous
Krea 2 RAW	Resolution-aware external sigma schedule, official 52-step and framework-reference 28-step recipes	Select RAW explicitly and provide the actual width and height
Z-Image Base	Fixed-ratio external sigma schedule, 28-50 steps	Select Base explicitly
Z-Image Turbo	Fixed-ratio external sigma schedule, 8-step official recipe	Select Turbo explicitly
FLUX.1-schnell	Unshifted external sigma schedule, 1-4 steps	Use the dedicated FLUX.1-schnell node
Original Qwen Image	ComfyUI fixed 1.15 shift, or Diffusers dynamic shift with explicit image_seq_len; original T2I only	Use Sigmax.QwenImageSigmaScheduler; select the mode explicitly
Original Stable Diffusion 3	Original SD3 Medium T2I only; explicit publisher-reference 1.0 or ComfyUI/Diffusers fixed 3.0 shift	Use Sigmax.SD3SigmaScheduler; select the source mode explicitly
Original AuraFlow v0.2	Fixed unit-flow 1.73 ratio shift, original 50-step recipe	Use Sigmax.AuraFlowSigmaScheduler; select Official Fixed (1.73) explicitly
Lumina-Image 2.0	Fixed unit-flow 6.0 ratio shift, original 50-step recipe	Use Sigmax.Lumina2SigmaScheduler; select Official Fixed (6.0) explicitly
HunyuanImage 2.1 Base	Fixed unit-flow 5.0 ratio shift, official 50-step recipe; schedule-only	Use Sigmax.HunyuanImage21SigmaScheduler; select Base (5.0) explicitly
HunyuanImage 2.1 Distilled	Fixed unit-flow 4.0 ratio shift, official 8-step publisher recipe; native host path unqualified	Use Sigmax.HunyuanImage21SigmaScheduler; select Distilled (4.0) explicitly
MiniMax H3 Base FL2VA/Ref2VA	Diffusers endpoint-inclusive external video sigmas with shift 12.0; model-owned audio mapping with shift 3.0; post-v1.0.0 qualification slice	Use Sigmax.MiniMaxH3SigmaScheduler with ComfyUI's upstream MiniMaxH3SigmaShift; select FL2VA or Ref2VA explicitly
Anima Base / Aesthetic	Fixed unit-flow rational 3.0 shift, 30-50 step framework-reference recipe; schedule-only	Use Sigmax.AnimaSigmaScheduler; select Base or Aesthetic explicitly
Anima Turbo	Fixed unit-flow rational 3.0 shift, 8-12 step framework-reference recipe; schedule-only	Use Sigmax.AnimaSigmaScheduler; select Turbo explicitly
Wan 2.1 T2V	Source-qualified unit-flow direct-ratio shift (5.0 official, 8.0 ComfyUI-native, 3.0 Diffusers reference), 50-step recipes	Use Sigmax.WanSigmaScheduler; select generation, task, source, and None resolution explicitly
Wan 2.1 I2V	Resolution-qualified official/Diffusers-reference paths: 480P 3.0 and 720P 5.0, 40-step recipes	Use Sigmax.WanSigmaScheduler; select I2V and the actual 480P or 720P class
Wan 2.2 TI2V 5B	Native and Diffusers-reference unit-flow 5.0 shift, 50-step recipes	Use Sigmax.WanSigmaScheduler; select TI2V and the source explicitly
Wan 2.2 A14B T2V/I2V	Source-qualified unit-flow paths with caller-owned boundaries; native T2V/I2V 12.0/5.0, Diffusers references 3.0	Use Sigmax.WanSigmaScheduler; boundary output is metadata only and never routes experts
Wan 2.1 FLF2V 14B 720P	Official-native unit-flow 16.0 shift, 50-step recipe	Use Sigmax.WanSigmaScheduler; select FLF2V, Official native, and 720P explicitly
Wan 2.1 VACE 1.3B / 14B	Official-native unit-flow 16.0 shift, 50-step recipes	Use Sigmax.WanSigmaScheduler; select the VACE model size and Official native explicitly
Wan 2.2 S2V 14B	Official-native unit-flow 3.0 shift, 40-step recipe	Use Sigmax.WanSigmaScheduler; select S2V and Official native explicitly
Wan 2.2 Animate 14B	Official-native unit-flow 5.0 shift, 20-step recipe	Use Sigmax.WanSigmaScheduler; select Animate and Official native explicitly
Wan Animate 2 Base / Distilled 14B	Official-native unit-flow 5.0 shift, 40-step Base or 10-step Distilled recipes	Use Sigmax.WanSigmaScheduler; select Wan Animate 2 and the explicit Base/Distilled task
LTXV 0.9.8 / LTX-2 19B / LTX-2.3 22B	Dev token-count adaptive shift; LTX-2/LTX-2.3 distilled Stage 1/2 immutable vectors; schedule-only	Use Sigmax.LTXSigmaScheduler; select generation and stage explicitly

The experimental Sigmax.Krea2ConditioningRebalance node accepts explicit RAW or Turbo CONDITIONING tensors with shape (batch, sequence, 30720). It preserves standard ComfyUI metadata and performs no scheduler or model mutation. Its profiles are community-derived and experimental; this boundary does not claim prompt adherence or image-quality improvement.

The generic advanced FlowMatch node constructs explicit schedule math only. It is experimental and does not establish compatibility with an arbitrary model.

MiniMax H3 Base FL2VA/Ref2VA is an accepted post-v1.0.0 development profile on dev, not part of the tagged 1.0.0 boundary. Public steps counts transitions and produces steps + 1 endpoint-inclusive video sigmas. The Sigmax scheduler owns that externally shifted video lane; ComfyUI's upstream MiniMaxH3SigmaShift supplies the matching video/audio shifts to the model so audio remapping remains model-owned. These are complementary responsibilities, not two schedule transforms.

Usage boundary

• Sigmax produces sigma schedules; it does not replace the numerical sampler.
• External Sigmax sigmas must not be shifted or scheduled a second time.
• Krea 2 Auto selection fails closed when evidence identifies the family but not RAW versus Turbo.
• Checkpoint header inspection does not load weights and cannot confirm a model variant from weak filename or structural evidence.
• Slicing, concatenating, or resampling a schedule changes its evidence status to modified.
• Model weights must be obtained separately under their own licenses and access terms.
• Qwen Image dynamic mode requires an explicit packed image sequence length; it never falls back to the fixed mode when that input is absent.
• SD3 source modes are intentionally non-composable. The publisher 1.0 and ComfyUI/Diffusers 3.0 values are separate evidence lanes, not a silent default selection.
• AuraFlow v0.2 uses one fixed model-native ratio shift. Do not apply a second shift or pass already-shifted sigmas to the node.
• Lumina-Image 2.0 uses one fixed model-native ratio shift. Do not apply a second shift or pass already-shifted sigmas to the node.
• HunyuanImage 2.1 uses one direct-ratio shift per explicit variant. The Base lane is host-compatible at the pinned baseline; Distilled is schedule-only and must not be presented as native host parity.
• HunyuanImage 2.1 model weights remain under Tencent's community license and are not distributed by Sigmax; this package does not include model code, weights, encoders, or conditioning.
• MiniMax H3 workflows must keep the Sigmax external video schedule and upstream MiniMaxH3SigmaShift values aligned at 12.0/3.0. Do not add BasicScheduler, apply another shift, or change only the model-side controls. The current Sigmax profile does not expose arbitrary H3 shift overrides.
• Anima applies one fixed rational 3.0 shift and rejects already-shifted composition. The node does not load Anima weights, run conditioning, or establish image-quality or prompt-adherence claims; Anima weight files remain under CircleStone Labs and applicable derivative licenses.
• Wan profiles keep ComfyUI-native, official-native, and Diffusers-reference shift ownership separate. A Wan 2.1 I2V resolution is mandatory, and unsupported derivatives fail closed. The released FLF2V, VACE, S2V, Wan 2.2 Animate, and Wan Animate 2 rows are official-native schedule lanes only; they do not imply model loading, conditioning, expert routing, or quality parity.
• Wan 2.2 A14B boundaries are descriptive caller-owned split metadata; Sigmax does not select high or low experts, load video weights, patch the model, or implement a video sampler.
• Wan Diffusers-reference rows document scheduler sigma/timestep construction only; they do not claim UniPC solver parity when consumed by an Euler sampler.
• LTX generation and stage are explicit; Dev token shifting and distilled vectors are separate lanes. LTX profiles do not load video weights or claim video-quality parity.
• Host E2E receipts record ComfyUI, Python, PyTorch, and runtime package versions for diagnosis; historical parity pins are reproducibility metadata, not a blocker for a current recommended host.

Not currently claimed

• General real-model GPU compatibility or image-quality parity. One bounded local Krea 2 H4 execution/provenance lane completed, but blind scoring and threshold review were explicitly waived, so it does not support a quality or profile-promotion claim.
• Stochastic, resumable, or interrupted sampler state.
• General partial-denoise execution or advanced model-patch compatibility beyond the explicitly documented MiniMax H3 host workflow.
• Automatic compatibility with unlisted model families.
• Guaranteed compatibility with every future ComfyUI release.
• AuraFlow v0.1/v0.3, PonyFlow, unsupported Wan derivatives (including Fun-Control and community finetunes), or real-model image-quality parity.

If a workflow falls outside these boundaries, treat it as unvalidated rather than silently substituting another profile or schedule.