Save 4 hours! We auto-setup your workflow! Free!

Drop your workflow.json — we handle every dependency, custom node, and model. Just open the link and run.

Auto-Setup Workflow Json (Free) Now!
ComfyUI > Nodes > ComfyUI-FeiHou-Easy-H3 > FeiHou Easy H3 1st Sample Enhancer(一采)

ComfyUI Node: FeiHou Easy H3 1st Sample Enhancer(一采)

Class Name

FeiHouEasyH3SampleEnhancer1st

Category
FeiHou Easy H3
Author
FX-FeiHou (Account age: 2190days)
Extension
ComfyUI-FeiHou-Easy-H3
Latest Updated
2026-11-12
Github Stars
0.08K

How to Install ComfyUI-FeiHou-Easy-H3

Install this extension via the ComfyUI Manager by searching for ComfyUI-FeiHou-Easy-H3
  • 1. Click the Manager button in the main menu
  • 2. Select Custom Nodes Manager button
  • 3. Enter ComfyUI-FeiHou-Easy-H3 in the search bar
After installation, click the Restart button to restart ComfyUI. Then, manually refresh your browser to clear the cache and access the updated list of nodes.

Visit ComfyUI Online for ready-to-use ComfyUI environment

  • Free trial available
  • 16GB VRAM to 80GB VRAM GPU machines
  • 400+ preloaded models/nodes
  • Freedom to upload custom models/nodes
  • 200+ ready-to-run workflows
  • 100% private workspace with up to 200GB storage
  • Dedicated Support

Run ComfyUI Online

FeiHou Easy H3 1st Sample Enhancer(一采) Description

Prepares Easy H3's high-noise first sampling pass and preserves media context for continuation.

FeiHou Easy H3 1st Sample Enhancer(一采):

This node prepares the first, high-noise sampling pass for the FeiHou Easy H3 workflow. It uses the low-resolution canvas defined by the Easy H3 context and creates the sampling components needed by ComfyUI’s official advanced custom sampler. The first pass focuses on establishing the main visual, video, and audio structure while the signal is still highly noisy, after which the resulting denoised data can be previewed or continued into a second Sample Enhancer pass. The node also preserves the related video and audio model information, including frame rate, so downstream preview, decoding, and second-pass processing can remain synchronized. Optional semantic-bridge processing can help guide the generated result toward the conditioning information, while configurable scheduling, classifier-free guidance, negative prompting, and trigger-word handling provide control over quality, prompt adherence, and sampling speed.

FeiHou Easy H3 1st Sample Enhancer(一采) Input Parameters:

model

Selects the H3-compatible model used for the first sampling pass. This model works on the smaller canvas and the higher-noise portion of the generation process, where the broad structure and temporal or audio relationships are established. The selected model must be compatible with the connected Easy H3 context and the H3 sampling workflow. This parameter has no numeric range or default value specified in the node interface; it accepts a MODEL connection.

h3_context

Provides the Easy H3 context containing the latent representation, conditioning information, and media-related settings required by the first pass. The context determines the first-pass canvas size, so the resolution configured in the connected Easy H3 node is the resolution used here. A valid MINIMAX_H3_CONTEXT connection from the Easy H3 main node is required. This parameter has no numeric range or default value; it accepts a MINIMAX_H3_CONTEXT connection. If another type of context is connected, preparation fails because the node cannot safely determine the required H3 sampling information.

sampler_name

Selects the sampling method used during the first pass. The available choices come from dual_clock.SAMPLER_OPTIONS; the interface recommends dual_clock_euler, which allows the video and audio portions to follow their respective clocks. Other choices are the ComfyUI-supported sampling methods exposed by the installed H3 implementation. The selected sampler affects motion, detail formation, audio or video synchronization behavior, and the visual character of the result. The default is dual_clock_euler, and the available options are the entries supplied by dual_clock.SAMPLER_OPTIONS. The same sampler choice is intended to be continued by the second pass.

scheduler

Controls how the noise levels change across the sampling steps. The available choices come from dual_clock.SCHEDULER_OPTIONS. The default is simple. The native_flow option uses a uniform H3-native flow schedule with shift 12 and is described as being very similar to simple; beta57 uses a beta distribution with parameters beta(0.5, 0.7). The scheduler affects how quickly the first pass moves from highly noisy data toward a structured result, so changing it can alter detail, stability, and the way the second pass continues the generation. The exact available options are those exposed by dual_clock.SCHEDULER_OPTIONS.

total_steps

Sets the total number of noise-removal steps represented by the complete sampling schedule, including the portion that will be continued by a later pass. More steps can provide a finer progression through the denoising process but usually require more computation when the full schedule is used. The allowed range is from 2 to 200, and the default is 9. This value must be greater than first_steps.

first_steps

Sets how many steps the first pass actually performs from the beginning of the full schedule. The first pass is intended for the high-noise portion, so it normally uses only the initial part of the schedule and leaves the remaining steps for the second Sample Enhancer pass. The allowed range is from 1 to 199, and the default is 4. The value must be lower than total_steps; otherwise the node raises the error first_steps 必须小于 total_steps. Increasing this value gives the first pass more influence over the image, video, or audio structure but leaves fewer steps for the continuation pass.

cfg

Controls classifier-free guidance, which determines how strongly the positive conditioning and prompt influence the result. A value of 1.0 is intended for workflows using an acceleration LoRA and does not calculate negative conditioning. Values greater than 1.0 enable the negative prompt and generally increase prompt adherence, although excessive guidance can produce harsh, unnatural, or over-constrained results. The allowed range is from 1.0 to 20.0, with increments of 0.1, rounding to 0.01, and a default of 1.0. The selected value is carried forward for the second pass.

negative_prompt

Contains text describing unwanted content or qualities, such as unwanted artifacts, incorrect anatomy, or undesirable visual styles. This text is used only when cfg is greater than 1.0; at cfg = 1.0, the node does not calculate negative conditioning. When enabled, the negative conditioning automatically uses the same relevant reference-image or keyframe information as the positive conditioning, and it is reused by the second pass. The default is an empty string, and the field accepts multiline text.

semantic_bridge_on

Enables or disables semantic-bridge processing. When enabled, the selected semantic-bridge model can add an additional conditioning influence during the first pass and is also configured for use by the second pass. When disabled, the semantic_bridge selection and bridge_strength value have no effect, and the node reports a stable disabled state for change detection. The default is True, displayed as 开 when enabled and 关 when disabled.

semantic_bridge

Selects a semantic-bridge model from the models discovered in the models/semantic_bridge directory or registered semantic-bridge model locations. The bridge can provide additional semantic guidance to the Easy H3 conditioning, but it is applied only when semantic_bridge_on is enabled and bridge_strength is greater than zero. The default is NONE_OPTION, meaning that no bridge model is selected. Available options are NONE_OPTION plus the discovered semantic-bridge model filenames. A selected model must be present and readable; otherwise the node reports 找不到语义桥模型,请放到 models/semantic_bridge 后刷新.

bridge_strength

Controls how strongly the selected semantic-bridge model influences the conditioning. A value of 0.0 effectively disables the bridge effect even if the bridge switch is enabled, while larger values apply stronger semantic guidance. Excessive strength may overpower the original Easy H3 conditioning, so moderate values are generally more appropriate when you want to preserve the source prompt or reference structure. The allowed range is from 0.0 to 1.0, with increments of 0.01, and the default is 0.10.

default_trigger_words

Controls whether the node adds the built-in trigger words to the beginning of the prompt conditioning. The default is True, which enables the default trigger words without displaying their expanded text in the node. Set this to False when your prompt or model setup already includes the necessary trigger words, or when you need complete control over the exact conditioning text. This is a Boolean option displayed as Default Trigger Words for both its enabled and disabled states.

FeiHou Easy H3 1st Sample Enhancer(一采) Output Parameters:

guider

Provides the configured guidance object for the official ComfyUI advanced custom sampler. It contains the positive conditioning and, when cfg is greater than 1.0, the corresponding negative conditioning and guidance strength. Connect this output to the guider input of the advanced custom sampler. It determines how the sampler interprets the prompt while denoising the first-pass latent.

sampler

Provides the sampling method prepared for the selected H3 model and sampler option. Connect it to the sampler input of the official advanced custom sampler. This output contains the sampling behavior needed to process the first-pass portion of the schedule, including the selected dual-clock behavior when dual_clock_euler is used.

sigmas

Provides the noise schedule values for the first pass. It contains the beginning of the complete schedule through the step selected by first_steps, so the advanced custom sampler performs only the intended high-noise section. Connect it to the sampler’s sigmas input. The final value represents the noise level at which the first pass stops and the continuation workflow can resume.

latent

Provides the low-resolution latent representation from the connected Easy H3 context. Connect it to the latent_image or equivalent latent input of the official advanced custom sampler. This is the data that the first pass denoises into the initial structured result. The latent resolution is determined by the Easy H3 context rather than by this node itself.

h3_context

Returns the processed H3 context for downstream use. It contains the original Easy H3 information together with the first-pass sampling plan, including the schedule continuation data, selected model and sampler settings, guidance configuration, negative conditioning when enabled, and semantic-bridge configuration when active. Connect this output to the corresponding context input of the second Sample Enhancer node so the second pass can continue from the correct point.

video_vae

Provides the video VAE associated with the H3 context. Use it when decoding the sampler’s denoised_output for video preview or later processing. This output ensures that downstream decoding uses the VAE associated with the current workflow rather than requiring you to select it again manually. It has type VAE.

audio_vae

Provides the audio VAE associated with the H3 context. Use it when decoding or processing the audio portion of the generated result. Keeping this output with the first-pass sampling data helps downstream nodes use the correct audio decoder for the current H3 workflow. It has type VAE.

fps

Provides the frame rate stored in the H3 context as a floating-point value. Use it for video decoding, preview, or export so playback speed remains consistent with the generation settings. If the context does not contain an explicit frame rate, the implementation uses 24.0 as the fallback value. This output has type FLOAT.

FeiHou Easy H3 1st Sample Enhancer(一采) Usage Tips:

  • Begin with the default configuration of total_steps = 9 and first_steps = 4. This gives the first pass the initial high-noise section while leaving a continuation section for the second Sample Enhancer pass.
  • Keep first_steps lower than total_steps. A practical rule is to reserve several steps for the second pass instead of using nearly the entire schedule in the first pass.
  • Use dual_clock_euler when working with combined video and audio generation, because the node documentation specifically recommends it for allowing the two media streams to follow their respective clocks.
  • Start with the simple scheduler. Test native_flow when you want an H3-native uniform flow schedule, or test beta57 when you need to compare a differently distributed denoising progression.
  • Keep cfg at 1.0 when using an acceleration LoRA and when you do not need negative prompting. Increase it above 1.0 only when negative conditioning or stronger prompt adherence is needed.
  • Use a moderate bridge_strength, such as the default 0.10, when enabling a semantic bridge. Increase it gradually if the semantic guidance is too weak, and reduce it if the original prompt or reference structure is being overridden.
  • Place semantic-bridge model files in models/semantic_bridge, then refresh the model list before selecting them. Keep the same semantic-bridge settings for the first and second passes through the returned H3 context.
  • Keep default_trigger_words enabled unless you have deliberately included the model’s trigger words yourself. Disable it when manually controlling the complete prompt conditioning to avoid duplicated trigger words.
  • Use the returned video_vae, audio_vae, and fps when decoding a preview from the sampler’s denoised_output. This keeps decoding aligned with the media settings stored in the Easy H3 context.
  • Connect the returned h3_context to the second Sample Enhancer node and use the same sampling method when continuing the workflow, because the context stores the information needed to resume the remaining schedule.

FeiHou Easy H3 1st Sample Enhancer(一采) Common Errors and Solutions:

h3_context 必须接 ComfyUI-FeiHou-Easy-H3 主节点的 h3_context

  • Explanation: The connected context is not the H3 context produced by the Easy H3 main node. The first Sample Enhancer requires the latent, conditioning, and media information stored in that specific context type.
  • Solution: Connect the h3_context output from the FeiHou Easy H3 main node directly to this node. Do not connect a regular latent, text-conditioning object, or context from an unrelated workflow.

first_steps 必须小于 total_steps

  • Explanation: The first pass cannot use all of the schedule because the workflow is designed to leave a continuation section for the second pass.
  • Solution: Increase total_steps or reduce first_steps. For example, use total_steps = 9 and first_steps = 4.

找不到语义桥模型,请放到 models/semantic_bridge 后刷新

  • Explanation: A semantic-bridge model was selected, but the file could not be found in the configured model directories.
  • Solution: Place the required .safetensors file inside models/semantic_bridge, refresh ComfyUI’s model list or restart the application, and select the model again. Disable semantic_bridge_on or choose NONE_OPTION if you do not need semantic-bridge processing.

FileNotFoundError while loading a semantic bridge

  • Explanation: The selected bridge entry points to a file that is no longer available, was renamed, or cannot be accessed.
  • Solution: Verify that the model file still exists, that its filename is unchanged, and that ComfyUI can read the directory. Refresh the model list after correcting the file location.

Invalid or incompatible MODEL connection

  • Explanation: The selected model may not support the H3 workflow, or the connection may not be a valid ComfyUI MODEL object. This can prevent the node from preparing the dual-clock sampler or calculating the schedule.
  • Solution: Connect a model intended for the FeiHou Easy H3 workflow and confirm that the required H3-related custom nodes and dependencies are installed and loaded.

Invalid sampler_name or scheduler selection

  • Explanation: The workflow may contain a sampler or scheduler name that is not available in the installed version of the H3 implementation.
  • Solution: Reopen the dropdown and select one of the currently available entries. Use dual_clock_euler and simple as the initial settings, then save the workflow again after selecting valid options.

Negative prompt has no visible effect

  • Explanation: Negative conditioning is intentionally skipped when cfg is 1.0 or lower. The node’s minimum CFG value is 1.0, so an empty or unused negative result is expected at that setting.
  • Solution: Set cfg to a value greater than 1.0, such as 2.0 or 3.0, and provide text in negative_prompt. Reduce the value if the result becomes overly constrained or harsh.

Semantic bridge has no visible effect

  • Explanation: The bridge is ignored when semantic_bridge_on is disabled, when semantic_bridge is set to NONE_OPTION, or when bridge_strength is 0.0.
  • Solution: Enable semantic_bridge_on, select an available bridge model, and set bridge_strength above 0.0. Start with the default strength of 0.10.

Missing or incompatible media context

  • Explanation: The node returns media information from the connected H3 context. If the context was created without the expected video or audio settings, downstream decoding may not have the correct VAE or frame-rate information.
  • Solution: Recreate the context using the Easy H3 main node and ensure that the intended video and audio configuration is connected before running the Sample Enhancer workflow.

Out-of-memory or insufficient VRAM error during sampling

  • Explanation: The first-pass model, canvas size, guidance settings, and semantic-bridge model can require substantial memory. A larger Easy H3 canvas or a higher number of sampling steps increases the workload.
  • Solution: Reduce the Easy H3 canvas size, lower total_steps while keeping first_steps valid, disable the semantic bridge, or use a lighter compatible model. Keep the first pass focused on the intended high-noise section rather than assigning nearly all steps to it.

FeiHou Easy H3 1st Sample Enhancer(一采) Related Nodes

Go back to the extension to check out more related nodes.
ComfyUI-FeiHou-Easy-H3
RunComfy
Copyright 2025 RunComfy. All Rights Reserved.

RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Models, enabling artists to harness the latest AI tools to create incredible art.

FeiHou Easy H3 1st Sample Enhancer(一采)