Hand Tie Clips 2.1.0 is outMulti-hop MiniMax H3 from one node — this release adds a refine pass and an H3 Cache so the non-turbo base is actually usable. Refine pass for degradation prevention over a long chain - visible degradation will happen still, but very much further down the line in a same-frame-continuous clip. As always, dump the page into a LM if you need a summary and easier understanding of how things work.Take note that the refine pass will not work well on turbos with the stock values. What's new<ul><li>hop_refine — optional second sampler pass per hop (ships off). Short, under-converged; blends back into the pin the next hop inherits. Defaults frozen: denoise 0.50, 2 steps, audio freeze. (this is the degradation fix)</li><li>H3 Cache — new node on the MODEL wire. Speeds up the plain (non-turbo) base. Based on silveroxides' block cache, improved and ported by PlagueKind, redistributed with permission.</li><li>Shipped workflows use 10 steps (was 7). Node default stays 14.</li><li>Prompt pack got a trust boundary so scene text can't rewrite hop count or rules.</li><li>Every new widget was appended — a 2.0 graph loads and renders the same. --------------------------------------------------------------------------------------- Small fix : audio reference will now pin across multiple hops Major update Hand Tie Clips 2.0 — Native MiniMax H3 Chains in ComfyUI</li></ul>I highly suggest feeding the git/hf page to a LM to study all the respective settings for control within your planning itself. It has a LOT of fine tune controls available now. This is NOT a plug-and-play one click setup, requires some tinkering to achieve what you want. MiniMax H3 produces stunning six-second clips, but building longer sequences is notoriously messy—character faces drift, lighting shifts, and audio restarts on every generation. Hand Tie Clips runs your entire sequence in a single queue: it passes final frames forward and locks character consistency from start to finish. Key Features <ul><li>🎙️ Continuous Voiceover: Drop in a single audio file. Every shot automatically lip-syncs to its own slice without audio resets, seam cuts, or manual re-timing.</li><li>🧍 Persistent Character Consistency: Feed it one reference photo. The node carries the identity through every shot, preventing face drift across the chain.</li><li>🎬 Beat-Based Directing: Write simple one-line actions per shot ("She turns to the window," "He walks to the counter"). The node handles prompt generation, camera movements, and cut transitions.</li><li>🔄 SWAP Mode: Combine motion from a video with identity from a photo—the source clip drives movement and scene context, while your reference photo drives the person.</li><li>✍️ Built-in Scripting (WRITE Tab): Connect a local LLM to automatically draft multi-shot sequences directly from a short concept paragraph.</li></ul>Quick Start <ol><li>Drop the <code>ComfyUI-Hand-Tie-Clips</code> folder into your <code>custom_nodes/</code> directory and restart ComfyUI.</li><li>Load the included starter workflow.</li><li>Add a character photo to the REFERENCES tab, write 3 shots, and hit queue.</li></ol>Includes starter and showcase workflows out of the box, with detailed tooltips for every setting. Needs<ul><li>ComfyUI new enough to include Add Guide for MiniMax H3 (<code>MiniMaxH3AddGuide</code>)</li><li>A ref2va (or hybrid ref2va) checkpoint — fl2va has no reference rows</li><li>Video VAE + audio VAE + MiniMax text encoder</li><li>Recommended: <a target="_blank" rel="ugc" href="https://github.com/NikoDemon80/ComfyUI-H3-Motion-Context">ComfyUI-H3-Motion-Context</a> for the latent join. Without it the node falls back to <code>MiniMaxH3AddGuide</code>, which still works and still chains — it is a different join, so a seam you are comparing against someone else's render may not be the same code path.</li></ul>The two shipped workflows wire the turbo stack this node is actually run with, because an example without it is not the graph anyone uses:<pre><code>UNETLoader -&gt; LoRA Loader Stack -&gt; H3 AdaLN LoRA Fix -&gt; MiniMax H3 Low VRAM
 -&gt; H3 SLA Attention -&gt; Model Preview Override -&gt; Hand Tie Clips
</code></pre>packnodes inside workflow : <a target="_blank" rel="ugc" href="https://github.com/PlagueKind/ComfyUI-PlagueKind-Nodes">ComfyUI-PlagueKind-Nodes</a>LoRA Loader Stack, H3 AdaLN LoRA Fix, H3 SLA Attention <a target="_blank" rel="ugc" href="https://github.com/kijai/ComfyUI-KJNodes">ComfyUI-KJNodes</a>MiniMax H3 Low VRAM Attention (experimental), Model Preview Override Realistic Limitations <ul><li>⚠️ Quality Drift: Image quality degrades on long chains. 3 to 5 shots is the realistic ceiling before colors and textures begin to wander.</li><li>📊 Degradation Study: The README details a 10-run, 9-shot study measuring exact failure points and identifying settings that yielded no impact.</li></ul>Requirements &amp; Links <ul><li>ComfyUI Build: <code>v0.34.0</code> or newer (required for H3 support)</li><li>Dependencies: None (no extra Python packages needed)</li><li>🔗 GitHub Repository &amp; Full Docs: <a target="_blank" rel="ugc" href="https://huggingface.co/sandpies/ComfyUI-Hand-Tie-Clips">https://huggingface.co/sandpies/ComfyUI-Hand-Tie-Clips</a></li></ul>

Hand Tie Clips — multi-hop MiniMax H3 video + audio, one queue

Hand Tie Clips 2.1.0 is outMulti-hop MiniMax H3 from one node — this release adds a refine pass and an H3 Cache so the non-turbo base is actually usable. Refine pass for degradation prevention over a long chain - visible degradation will happen still, but very much further down the line in a same-frame-continuous clip. As always, dump the page into a LM if you need a summary and easier understanding of how things work.What's new<ul><li>hop_refine — optional second sampler pass per hop (ships off). Short, under-converged; blends back into the pin the next hop inherits. Defaults frozen: denoise 0.50, 2 steps, audio freeze. (this is the degradation fix)</li><li>H3 Cache — new node on the MODEL wire. Speeds up the plain (non-turbo) base. Based on silveroxides' block cache, improved and ported by PlagueKind, redistributed with permission.</li><li>Shipped workflows use 10 steps (was 7). Node default stays 14.</li><li>Prompt pack got a trust boundary so scene text can't rewrite hop count or rules.</li><li>Every new widget was appended — a 2.0 graph loads and renders the same.</li></ul>Credits (H3 Cache)<ul><li>silveroxides — original MiniMax-H3 block cache (ComfyUI-UtilsCollection)</li><li>PlagueKind — improvements, port, reference protection, production defaults (ComfyUI-H3-MiniMax-Cache)</li></ul>Install / update<pre><code>cd ComfyUI/custom_nodes/ComfyUI-Hand-Tie-Clips
git pull
</code></pre>Restart ComfyUI, hard-refresh the browser. Console should say <code>editor ui v2.1.0 loaded</code>.GitHub: <a target="_blank" rel="ugc" href="https://github.com/dntpi/ComfyUI-Hand-Tie-Clips">https://github.com/dntpi/ComfyUI-Hand-Tie-Clips</a> Full notes: <a target="_blank" rel="ugc" href="http://CHANGELOG.md">CHANGELOG.md</a> in the repo.

Files with SHA256: f044e7a798f3ecdebb76f42ebe1371690a10d8afce30ef8520b226e64c3d7a88