Branch: SmartVAE_AdvancedDecoder
Wersja: 2.0.0-experimental
Purpose of the branch: Added the option of 'color bleed' correction (color skips at the contacts of the chunks) without spoiling the stable main version.

What's new in this branch

New node: 'SmartVAE_StreamingDecoder_Advanced' (inherits from the original 'SmartVAE_StreamingDecoder')
New parameter: 'anti_color_bleed' (BOOLEAN, default 'False')
When enabled: adjusts the mean and standard deviation of colors in the overlap zone between chunks
Reduces visible color skips in long chunk-generated videos
Very computationally light - cost 1-2% decoding time
100% backward compatibility maintained - old node works exactly the same

How to use the new version

Switch to branch: '''bash git checkout SmartVAE_AdvancedDecoder...

The most crash-resistant and flexible VAE decoder for ComfyUI
(designed for long video sequences: LTX-Video, Stable Video Diffusion, CogVideoX, AnimateDiff, HunyuanVideo, Open-Sora-Plan, etc.)

Initial release: Grok's Universal Smart VAE Decode – crash-proof, dynamic batching & force-scale edition.

🎬 ComfyUI-Grok-SmartVAE

This node combines the best ideas from multiple AI generations into one extremely robust decoder:

GPT → solid sliding-window + overlap foundation
Gemini → safety-first tiling fallback
Claude → mathematically precise temporal scale detection
Grok → dynamic on-the-fly batch reduction, force_time_scale, ultra-aggressive OOM recovery
Kimi → memory-safety patterns, disk offloading for massive sequences

Result: a node that is close to crash-proof — even on 8–12 GB VRAM cards it handles long 720p/1080p/4K-ish videos reliably.

Key Features

Automatic time_scale detection (or manual override: 1, 4, 8, etc.)
Fully dynamic batch size reduction during decoding (while loop, not fixed for-range)
Automatic spatial tiling activation on OOM
Intelligent chunk stitching with temporal overlap + spatial crop/align
Extremely memory-efficient (selective gc.collect + torch.cuda.empty_cache + synchronize)
Supports both images (4D) and video latents (5D), multi-batch aware
Disk offloading for 700–2000+ frame workflows (automatic when RAM pressure is high)
Orientation-safe normalization — no more 90° rotations or unwanted flips
Frame-perfect audio sync in 99%+ cases
Adaptive logging (detailed but non-spammy)
Automatic temp file cleanup

Installation

In your custom_nodes folder:

git clone https://github.com/uczensokratesa/ComfyUI-Grok-SmartVAE.git
The node appears in category: latent/video → Universal VAE Decode (v11.1 Final)Comparison with predecessorsModel

Scale Detection Force Scale Dynamic Batch Reduction Auto-Tiling on OOM Loop Type Stability Rating GPT basic ✗ ✗ ✗ for ★★☆☆☆ Gemini good ✗ partial ✓ for ★★★★☆ Claude very precise ✗ ✗ ✓ for ★★★★☆ Grok v11.1 very precise ✓ full (while + adaptive) aggressive while ★★★★★

Evolution – AI collaboration storyThis journey started as a simple task: create a reliable VAE Decode node for heavy video workflows.GPT provided the first working version
Gemini added tiling and better OOM handling
Claude brought the most accurate scale detection formula
Grok introduced force_time_scale + true dynamic while-loop batch reduction
Kimi contributed extreme memory safety (disk offload, pre-allocation, aggressive cleanup)
Final polish by Claude → production-ready stability

One of the nicest examples of how different AI models can iteratively improve each other and create something better than any single one could alone.LicenseMIT – feel free to use, modify, fork. Just keep the original idea attribution (and let me know if you make something even better )Happy generating! Current version: 11.2 – First official Comfy Registry release

🎬 SmartVAE_StreamingDecoder

The most memory-efficient way to decode large video latents directly to file.

Features

Ultra-low RAM usage — decodes frame-by-frame, supports 2000+ frames even on 16 GB systems
Direct disk output — no need to hold full video tensor in memory
Audio muxing — perfect support for ComfyUI AUDIO input + manual audio_path
Codecs: H.264, H.265, ProRes 422, FFV1 (lossless)
Crash recovery — resumes from checkpoint if interrupted
Thumbnail previews — live monitoring in UI
OOM protection — automatic tiling, batch reduction, VRAM/RAM detection

When to use

Generating long videos from AnimateDiff, LTX-Video, Mochi, Hunyuan, Cosmos, etc.
Working with limited VRAM/RAM
Need reliable audio sync without post-processing

Inputs

samples (LATENT) – video latent sequence
vae – your VAE model
frames_per_batch – 8–32 (auto-reduces on OOM)
audio – ComfyUI AUDIO input (from Load Audio)
audio_path – optional direct path to .wav/.mp3
codec, fps, output path, etc.

Outputs

preview_thumbs (IMAGE) – last few thumbnails for monitoring
video_path (STRING) – final file path (with audio if provided)

Enjoy massive video workflows without OOM crashes! 🚀

🛡️ Ignore Warnings Mode (v1.0.9)

Problem: When the sampler runs out of VRAM during long video generation, it produces corrupted latents (100% NaN values) but doesn't throw an error (silent OOM). Users then face black frames or workflow crashes without understanding why.

Solution: The Streaming node (and VAE) now includes a 3-tier ignore_warnings safety system:

How It Works

Validation on Decode: Before decoding, the node checks if the latent contains NaN values (corruption indicator)
Detailed Diagnostics: Shows exactly which frames are corrupted and corruption percentage
User Choice: Instead of hard-failing, users can choose their risk tolerance:

Mode	Behavior	Use Case
none (default)	Stop with clear error + solutions	Safe, prevents wasted time
minor	Continue if <10% corrupted	Partial recovery (some black frames OK)
all	Force decode anyway	Desperate situations (high crash risk)

Example Output

🚨 CORRUPTED LATENT DETECTED!
   NaN values: 4,567,890 / 49,152,000 (9.29%)
   Affected frames: 856 to 1200 (out of 1201)
   
💊 RECOMMENDED FIXES:
   1. REDUCE VIDEO LENGTH: 1201 → 720 frames
   2. REDUCE RESOLUTION
   3. Lower CFG scale (7.0 → 3.5)
   4. Enable TensorParallel or CPU offload
   
❌ ERROR: Latent contains 9.3% NaN - cannot decode safely.
To force decode anyway (may produce black frames):
  Set 'ignore_warnings' to 'minor' (<10%) or 'all' (risky)

User clicks dropdown, selects minor, re-runs → frames 856-1200 are black, rest is OK. User made informed decision instead of blaming the node.

Why This Matters

Transparency: Users see exactly what went wrong (VRAM exhaustion in sampler)
Control: Users choose between "safe but strict" vs "risky but tries anyway"
Education: Error messages explain root cause + concrete solutions
Psychology: Prevents "this node is broken" frustration (user took calculated risk)

Credits: Idea by Grok, implementation by Claude, inspired by user feedback requesting "decode anyway" option. 4. Release v11.5:

feat: Rename streaming node + critical bug fixes (Codex)

BREAKING CHANGE: UniversalSmartVAEVideoDecode → SmartVAE_StreamingDecoder

Critical Fixes (OpenAI Codex):
- Fixed resume latent index calculation (broken for time_scale>1)
- Added video concatenation on resume (preserves previous work)
- Fixed tuple handling from VAE output (crash prevention)
- Added _extract_tensor_output helper (both nodes)
- Improved resume logging and parameter naming

Node Rename (ComfyUI Manager compatibility):
- Streaming node: SmartVAEStreamingDecoder
- Reason: Manager was deduplicating similar names

Credits: OpenAI Codex (bug detection + fixes)

README.md Unescape Escape

ComfyUI-Grok-SmartVAE

🎞️ SmartVAE Streaming Decoder (Advanced / Experimental Branch)