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 > velvet-vice-minimax-h3 > VELVET VICE MiniMax H3 — Director

ComfyUI Node: VELVET VICE MiniMax H3 — Director

Class Name

VelvetViceMiniMaxH3Director

Category
VELVET VICE/MiniMax H3
Author
velvet-vice (Account age: 0days)
Extension
velvet-vice-minimax-h3
Latest Updated
2026-09-22
Github Stars
0.0K

How to Install velvet-vice-minimax-h3

Install this extension via the ComfyUI Manager by searching for velvet-vice-minimax-h3
  • 1. Click the Manager button in the main menu
  • 2. Select Custom Nodes Manager button
  • 3. Enter velvet-vice-minimax-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

VELVET VICE MiniMax H3 — Director Description

Plans MiniMax H3 video dimensions, frame count, seed, and render settings for downstream nodes.

VELVET VICE MiniMax H3 — Director:

VelvetViceMiniMaxH3Director prepares the creative and technical settings for a MiniMax H3 render in one place. It plans the video’s dimensions and frame count, manages the seed used for repeatability, and packages the selected render, Turbo, audio, and ending settings for downstream nodes. It also provides a readable summary and readiness status so you can check the planned render before proceeding. This node prepares and passes along settings; it does not itself load models or generate the finished video.

VELVET VICE MiniMax H3 — Director Input Parameters:

first_frame

Required IMAGE. This is the starting image for the planned render. The node uses its dimensions when calculating an automatic format and passes the image through unchanged. No default is specified; connect an image to use the node.

render_preset

Required selection with options TEST, FAST, BALANCED, QUALITY, and CUSTOM; default BALANCED. Your choice is recorded in the render configuration for downstream use. This node does not itself change its dimension or frame-count calculations based on the preset, so the precise rendering tradeoffs depend on the connected workflow.

duration_seconds

Required FLOAT specifying the requested video length; default 5.0, minimum 1.0, step 0.1, and a source-defined maximum whose numeric value is not included in the provided code. The node limits the request to its supported duration range, then calculates a frame count on H3’s 17k+5 frame grid. Consequently, the actual planned playback time, calculated as frame count divided by 24 FPS, can differ from the number you enter. The source tooltip describes approximately 5–15 seconds as the commonly trained range and longer clips as untested by upstream ComfyUI.

format_mode

Required selection with options AUTO, 16:9, 9:16, 1:1, and CUSTOM; default AUTO. This controls the aspect ratio used to plan the output dimensions. AUTO uses the first image’s proportions; the numbered choices request a specific ratio; CUSTOM derives a ratio from custom_width and custom_height. Planned dimensions are rounded to multiples of 32, so the final ratio may be approximate.

resolution_mode

Required selection with options PRESET, CUSTOM MP, and CUSTOM SIZE; default PRESET. This determines how the planned width and height are calculated. PRESET uses the selected resolution preset, CUSTOM MP uses a target megapixel area, and CUSTOM SIZE uses the entered width and height directly after rounding and minimum-size checks. The resulting dimensions are shown in the outputs and render configuration.

resolution_preset

Required selection; default 900p. Available options are 144p, 240p, 360p, 480p, 540p, 576p, 720p, 900p, 1024p, 1080p, 1152p, 1440p, 2160p, 2K, 4K, ``0.26 MP

  • Preview, 0.36 MP - Small, 0.52 MP - SD, 0.65 MP - Balanced, 0.83 MP - HD, 1.00 MP - 1024p, 1.05 MP - HD+, 1.20 MP - HD++, 1.35 MP - 2K lite, 1.55 MP - 2K, 1.65 MP - 2K+, 1.75 MP - QHD, 2.10 MP - FHD, 3.30 MP - QHD+, 4.75 MP - 2K Pro, 6.50 MP - Production, 8.30 MP - UHD, and CUSTOM. The selection is used for dimension planning when resolution_modeisPRESET``. Preset names are targets, not a guarantee that either calculated dimension will exactly match the number in the name.

rotate_format

Required BOOLEAN; default False. In CUSTOM SIZE mode, enabling this swaps the planned width and height. In the other resolution modes, it inverts the ratio used to calculate dimensions, changing a landscape-oriented plan toward portrait orientation or vice versa. It affects the planned dimensions, not the pixels of the incoming image.

custom_width

Required INT; default 1344, minimum 256, maximum 4096, step 32. Together with custom_height, this supplies the requested dimensions in CUSTOM SIZE mode or defines the aspect ratio when format_mode is CUSTOM. In CUSTOM SIZE mode, the node rounds the value to a multiple of 32 and enforces a minimum planned dimension of 256 pixels. The entered value may also appear in the human-readable summary even when rotation changes the final planned width.

custom_height

Required INT; default 768, minimum 256, maximum 4096, step 32. Together with custom_width, this supplies the requested dimensions in CUSTOM SIZE mode or defines the aspect ratio when format_mode is CUSTOM. In CUSTOM SIZE mode, the node rounds the value to a multiple of 32 and enforces a minimum planned dimension of 256 pixels. Check the height output for the final value, particularly when rotation is enabled.

custom_megapixels

Required FLOAT; default 0.4, minimum 0.1, maximum 4.0, step 0.1. This sets the target pixel area when resolution_mode is CUSTOM MP. The node combines that area with the selected aspect ratio to calculate width and height, then rounds both dimensions to multiples of 32. In PRESET mode, it is also the calculation’s fallback value if a preset has no matching megapixel entry in the source-defined preset mapping.

turbo_enabled

Required BOOLEAN; default False. This records whether Turbo is requested in the render configuration and passes the same choice to the dedicated output. The summary labels the choice TURBO or BASE. The Director does not itself load or apply a Turbo model, so confirm that downstream nodes support the selected setting.

seed_mode

Required selection with options RANDOM, LOCKED, and REUSE LAST; default RANDOM. RANDOM generates a new seed when the node runs. LOCKED uses the number entered in seed. REUSE LAST uses the seed from this node instance’s previous execution; if none exists yet, it generates a new one. The resolved number is available in the seed output for recording or reuse.

seed

Required INT; default 1, minimum 0, maximum 9223372036854775807, step 1. This number is used directly when seed_mode is LOCKED. In RANDOM and REUSE LAST modes, check the output rather than assuming the entered number was used. A seed helps you repeat the same configuration, although it does not by itself guarantee identical results from every downstream rendering setup.

native_audio_output

Required BOOLEAN; default True. This passes your audio-output preference into the render configuration and dedicated output. The summary describes the choice as AUDIO WITH SOUND when enabled or AUDIO MUTED / VIDEO ONLY when disabled. This node does not itself produce, remove, or verify an audio track.

ending_mode

Required selection with options AUTO, NO CLIMAX, CLIMAX, and LOOP / CONTINUOUS ACTION; default AUTO. This records the requested style of ending in the render configuration and passes it through as text. The Director does not itself edit the ending of a video; its effect depends on how downstream nodes use the setting.

fps

Required INT; default, minimum, and maximum are all 24. H3 planning in this node runs at 24 frames per second: the execution method explicitly sets FPS to 24 before calculating frames and outputs. This is not an adjustable frame-rate control in the provided implementation.

last_frame

Optional IMAGE with no specified default; when unconnected, its value is None. Connect an image if your downstream workflow uses a target final frame. The node passes it through and records whether it was connected, but the provided code does not inspect or transform that image.

system_ready

Optional BOOLEAN input with no specified default; it is configured as a connection-only input. When unconnected, the node treats the system as ready. A connected false value changes the status prefix to SYSTEM NOT READY; a true value produces READY TO RENDER. This is a reported readiness signal, not a check performed by the Director itself.

system_status

Optional STRING input with no specified default; it is configured as a connection-only input. If a nonempty value is provided, the node appends it to its status text so you can see an upstream system message alongside the planned render details. It does not alter the calculated dimensions, seed, or frame count.

VELVET VICE MiniMax H3 — Director Output Parameters:

first_frame

IMAGE output containing the incoming starting image, passed through for connection to later nodes. The Director reads its dimensions for planning but does not modify the image.

last_frame

IMAGE output containing the optional final image as supplied. When no last_frame is connected, this output carries the unconnected value rather than a generated image.

width

INT output giving the final planned video width in pixels. It reflects the chosen format, resolution mode, rounding, and any rotation setting; use this value rather than an entered custom value when configuring downstream dimensions.

height

INT output giving the final planned video height in pixels. Like width, it reports the calculation’s result, which can differ slightly from a target because dimensions are rounded to multiples of 32.

duration_seconds

FLOAT output giving the requested duration after the node limits it to its supported range. It is not necessarily the exact playback length: divide frame_count by fps to find the planned duration on H3’s frame grid.

fps

FLOAT output containing 24.0. This is the frame rate used for the node’s frame-count calculation and for interpreting the planned playback duration.

frame_count

INT output containing the planned number of H3 frames. The node rounds the duration-based count onto its 17k+5 grid, starts from a minimum of five frames, and caps the result at its source-defined maximum valid count.

seed

INT output containing the seed actually selected for this execution. Record this value when you want to repeat a run that used RANDOM or REUSE LAST, because it may differ from the number shown in the input control.

turbo_enabled

BOOLEAN output passing through the selected Turbo preference. It tells downstream nodes whether Turbo was requested; it does not confirm that Turbo resources were loaded or applied.

native_audio_output

BOOLEAN output passing through the selected native-audio preference. It represents the requested output setting, not confirmation that the finished render contains or lacks sound.

ending_mode

STRING output containing the selected ending option. Downstream nodes can use it to interpret the requested ending style; the Director does not alter video content itself.

render_config

VELVET_VICE_H3_RENDER_CONFIG output containing a structured configuration for downstream nodes. It includes the requested and calculated timing, dimensions, format and resolution choices, resolved seed, Turbo and audio preferences, ending mode, and whether a last frame was connected. Its schema value is VELVET_VICE_H3_RENDER_CONFIG_V1.

summary

STRING output containing a readable account of the planned render, including requested and calculated duration, frame count, dimensions, 24 FPS, format and resolution selections, Turbo state, audio preference, and resolved seed. Use it to check the plan quickly; the calculated dimensions and frame count are the operative values.

status

STRING output containing a readiness prefix followed by the summary and, when supplied, the upstream system_status message. READY TO RENDER means no false readiness signal was received; it does not independently verify that models or other downstream requirements are available.

VELVET VICE MiniMax H3 — Director Usage Tips:

  • Start with BALANCED, AUTO format, and a short duration, then inspect width, height, and frame_count before connecting the configuration to a render workflow. The source describes roughly 5–15 seconds as the commonly trained duration range.
  • Use CUSTOM SIZE when exact planned dimensions matter. Enter dimensions in 32-pixel steps and check the outputs, especially after enabling rotate_format.
  • Use CUSTOM MP when you want to choose an approximate pixel budget while letting the node calculate dimensions for your preferred aspect ratio. For a specific portrait or landscape composition, select 9:16 or 16:9 rather than relying on the source image’s ratio.
  • Choose LOCKED and keep the seed value when comparing settings. For a random run you want to revisit, note the resolved seed output. REUSE LAST applies to this node instance’s prior execution and generates a seed if it has no previous value.
  • Connect system_ready and system_status from your system-checking workflow when available. Read the resulting status before rendering, but treat it as a reported signal rather than a guarantee that rendering will succeed.
  • Connect last_frame only when the later workflow supports it. The Director records its presence and passes it along; it does not create an end-frame transition on its own.

VELVET VICE MiniMax H3 — Director Common Errors and Solutions:

SYSTEM NOT READY

  • Explanation: This status appears when the connected system_ready value is false. It reports the upstream readiness signal; the Director does not diagnose the underlying issue.
  • Solution: Read any appended system_status text, resolve the issue in the connected system workflow, and run the check again.

READY TO RENDER appears when the system has not been checked

  • Explanation: With system_ready unconnected, the Director assumes readiness. This message alone does not confirm that required models or downstream nodes are available.
  • Solution: Connect a reliable readiness signal if your workflow provides one, and verify the downstream setup before starting a render.

Duration differs from the requested seconds

  • Explanation: This is an expected planning result, not necessarily an error. The node snaps the frame count to H3’s 17k+5 grid at 24 FPS, so frame_count / 24 may not equal the requested duration exactly.
  • Solution: Use frame_count and fps to check the planned playback time, then adjust duration_seconds if you need a nearby supported length.

Planned dimensions differ from the entered size or preset name

  • Explanation: The node rounds planned dimensions to multiples of 32. Preset and megapixel modes calculate dimensions from a target area and aspect ratio rather than setting an exact named width or height; rotate_format can also change the result.
  • Solution: Check the width and height outputs. For a specific size, select CUSTOM SIZE, enter 32-pixel-step dimensions, and verify the rotation setting.

REUSE LAST gives an unexpected seed

  • Explanation: The mode reuses the seed from this node instance’s previous execution. If there is no previous seed, it generates a new one.
  • Solution: Use LOCKED with a known seed for an explicit repeat, or take the resolved value from the seed output.

A render has no sound despite native_audio_output being true

  • Explanation: The Director records and passes along an audio preference; it does not generate an audio track or confirm that downstream rendering honors the preference.
  • Solution: Check the audio-capable parts of the downstream workflow and confirm that they use the render_config or native_audio_output setting.

Turbo or the selected ending mode has no visible effect

  • Explanation: This node passes these choices to downstream nodes but does not itself apply a Turbo model or edit a video’s ending.
  • Solution: Confirm that the connected rendering workflow supports and consumes the relevant setting.

VELVET VICE MiniMax H3 — Director Related Nodes

Go back to the extension to check out more related nodes.
velvet-vice-minimax-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.

VELVET VICE MiniMax H3 — Director