VELVET VICE LTX — v0.1.12 Stable NVENC Output + Preview

VelvetViceLTXAutoVideoCombineV0112 is the newest of the pack's version-bound final-output nodes, the one the portable v0.1.12 workflow pins, and - because the pack's current default implementation version is 0.1.12 - effectively the one you should reach for when building fresh. It's the grown-up of the family.

What v0.1.12 changed

The encode path is the same proven one from v0.1.10: a single NVENC-or-CPU H.264 encode with the real 256×256 probe, plus the fast FP16→RGB8 delivery. The change is in the preview wiring. In earlier variants, the preview came along with a side effect that could shift the values of serialized encoder widgets - i.e., the act of previewing could nudge your bitrate or CRF settings as stored in the workflow file. v0.1.12 pairs the same encode with a post-configuration direct preview that cannot shift those serialized widget values. Your settings stay put, no matter how many times the preview refreshes.

That's the "Stable" in the display name, and it's exactly the kind of bug fix that only makes sense once you've had a workflow silently change its own encoder settings on you. Annoying enough to warrant a dedicated node version, honestly.

Everything else is the family behavior: AUTO mode probes h264_nvenc with a test frame and falls back to CPU libx264 at your CRF; NVIDIA NVENC ONLY raises when the probe fails; CPU H.264 ONLY forces software encoding; bitrate scales from nvenc_bitrate_mbps_at_24fps with FPS; the filename prefix is Windows-safe with %date:...% token support.

Inputs and outputs

Same as the base node: images, frame_rate, loop_count, filename_prefix (video/LTX23-FINAL), encoder_mode, nvenc_bitrate_mbps_at_24fps (60), cpu_crf (15), pix_fmt (yuv420p), pingpong, save_metadata, trim_to_audio, save_output, optional audio. Output: Filenames (VHS_FILENAMES), an output node.

Installing it

Part of VELVET VICE - LTX (velvet-vice-ltx) via ComfyUI Manager, or:

cd ComfyUI/custom_nodes
git clone https://github.com/Velvet-Vice/velvet-vice-ltx

Restart, hard-refresh with Ctrl+F5. No pip deps in the pack; VideoHelperSuite and FFmpeg required at runtime.

Which one to pick

• Loading a v0.1.12 workflow: use this node as-is. Version mixing is the one thing the README explicitly warns against.
• Building your own graph: this is the sensible default - you get the fast delivery and the preview that doesn't stomp your settings. The unversioned VelvetViceLTXAutoVideoCombine behaves the same today; the older pinned variants (v0.1.9, v0.1.10) exist purely so old workflows keep their original behavior.

Troubleshooting

• Encoder settings "reset" after a preview - if you're on an older variant, that's the bug v0.1.12 fixed. Update the pack (and switch to this node) and re-verify your values.
• Missing node on load - stale or merged install; delete the custom-node folder and reinstall clean, then restart.
• NVENC ONLY raises - the probe found no working h264_nvenc; use AUTO for the CPU fallback.

It's a humble output node, but it's the one the pack's author actually settles on - which is as good a recommendation as a custom node gets in a pack this niche.