<h2 id="support">Support</h2>Everything here is free and stays free — the format spec, the nodes, the workflows, the cartridges, the LoRAs. If it saved you a night of debugging (it contains several hundred of mine), tips keep the 5090 warm:<ul><li>☕ <a target="_blank" rel="ugc" href="https://ko-fi.com/joeygambino">Buy me a coffee on Ko-fi</a></li><li>💛 <a target="_blank" rel="ugc" href="https://github.com/sponsors/jlucasmcrell">Sponsor on GitHub</a></li><li>🔁 <a target="_blank" rel="ugc" href="https://liberapay.com/joeygambino">Liberapay</a> (recurring)</li><li>⚡ Or right here: the Civitai tip button on this page sends Buzz directly.</li></ul><h2 id="type-a-story.-get-one-continuous-video-with-sound.-multi-shot-scenes-render-as-a-single-take-no-last-frame-chaining-no-quality-loss-from-shot-to-shot.-that's-the-whole-pitch.">Type a story. Get one continuous video, with sound. Multi-shot scenes render as a single take - no last-frame chaining, no quality loss from shot to shot. That's the whole pitch.</h2>Both workflows now read left to right: numbered lanes, and you only ever touch lanes 2-4. Everything below the main row is optional.<h2 id="what-you-need">What you need</h2><ul><li>ComfyUI + this node pack (Manager: MiniMax-H3 Multishot, or the zip on this version).</li><li>A MiniMax-H3 checkpoint (links on this page). 24 GB card? Take a GGUF.</li></ul><h2 id="seamless-chain-multi-shot-scenes-as-one-take">SEAMLESS CHAIN - multi-shot scenes as one take</h2>Lane by lane:<ul><li>README - the quick start lives on the canvas itself.</li><li>1 - MODELS - pick your H3 checkpoint and text encoder. VAEs are preset; LoRA slots are empty until you fill one.</li><li>2 - ANCHORS (optional) - a photo to open shot 1 on (enable its gate), and a short voice clip to lock the speaker's voice.</li><li>3 - YOUR PROMPTS - type your idea in the box, or point the switch at a prompt file. The writer expands it into shot prompts. Writing your own? Set the writer to <code>passthrough (raw JSON, skip LLM)</code> and paste shots separated by <code>---</code> lines.</li><li>4 - CONTROLS - size, frames per shot, steps, and <code>take_seconds</code> (total length; 30 is a good first run). The switches stay off unless you installed the pack a switch names.</li><li>5 - ENGINE - nothing to change. The remote encoder lives here if you want the text encoder on a second PC: enter its address, flip the encoder switch, free ~15 GB.</li><li>6 - OUTPUT - your video and its audio save here.</li></ul>Optional panels below the main row: reference images (your character, ref2va checkpoints - folder per character + AUTO REFS on), V2V reference (a clip whose look guides the render), FFLF plates (flf_chain mode only), audio spine (a soundtrack the take follows).<h2 id="extend-take-one-person-talking-as-long-as-you-want">EXTEND TAKE - one person talking, as long as you want</h2>Same lanes, different job: one premise becomes ONE continuous speech cut across windows.<ul><li>1 - MODELS - same as above.</li><li>2 - ANCHORS - a photo of your speaker (shot 1 opens on them) and a voice clip. More useful here than anywhere: one person carries the whole take.</li><li>3 - YOUR PROMPT - ONE premise, one speaker. The writer writes the whole speech. num_shots 0 = it decides. Passthrough works here too.</li><li>4 - CONTROLS - <code>take_seconds</code> is the star: 30 ships, 60 clears TikTok's minute. <code>window</code> stays on auto - it sizes itself to your card.</li><li>5 - ENGINE / 6 - OUTPUT - same as above.</li></ul>Keep takes to about 4 windows for now - very long takes slowly sharpen.<h2 id="rules-of-thumb-(both-workflows)">Rules of thumb (both workflows)</h2><ul><li>Spoken lines: 8-12 words per shot. Short lines sync; long lines garble.</li><li>Say the sounds you want ("rain on the roof, a fridge hum") or it invents its own.</li><li>Keep your character's face in frame - faces carry identity between shots.</li></ul><h2 id="if-something-breaks">If something breaks</h2><ul><li>Red node? Update the pack in Manager, restart, reload the workflow from disk.</li><li>Render crawls at low wattage? Lower resolution or frames per shot, or use the remote encoder.</li><li>Only one of the two workflows shows in your sidebar? Fixed in 2.6.5 - re-download both.</li><li>Still stuck: comment with your console log. I answer.</li></ul>Deep dives: the two articles linked on this page. Every lane also has a short note on the canvas. Detailed guide for people that can read good: <a target="_blank" rel="ugc" href="https://civarchive.com/articles/34046/every-setting-explained-the-seamless-chain-deep-manual">Every setting explained: the Seamless Chain deep manual | Civitai</a>

MiniMax-H3 Multishot — Seamless Chain: multi-shot scenes that render as one continuous take (picture + audio)

v1.1Image-to-video, a long-form memory sampler, and a VRAM fix that cuts render time roughly 4x on 32 GB cards.Image-to-videoThe sampler now takes an optional <code>start_image</code>. The AIO workflow ships with <code>LoadImage → H3 Optional Image → start_image</code>, so one graph does both: flip the toggle on to begin from your frame, off for pure text-to-video. Shot 1 keeps its first frame, so the image you supply is the image you get.Why a custom toggle? A normal switch node cannot express “no image” — both of its branches are required — so turning I2V off that way ends up feeding a black placeholder frame. That is not text-to-video, it is video that starts from black. <code>H3 Optional Image</code> emits nothing when disabled.~4x faster on 32 GB cardsThe text encoder is now evicted before sampling. The Qwen3-VL encoder (~16.5 GB even at Q4) and the H3 DiT (~25 GB) do not co-fit on a 32 GB card, so the DiT was loading partially and streaming ~19 GB from system RAM on every step. If you ever saw this in your log, that was it:<pre><code>loaded partially; 6423 MB usable, 5847 MB loaded, 19363 MB offloaded</code></pre>Conditioning is already computed before sampling starts, so the encoder is safe to evict there. Measured on an RTX 5090: ~60 min → ~15 min for the same render. The node prints <code>[H3Multishot] TE evicted; NN.N GB free for the DiT</code> each shot so you can confirm it.New: long-form memory samplerFor 2–5 minute videos (12–30 shots), where plain chaining drifts. Stock chaining shows each shot one image — the previous shot’s last frame — so every hop can only see one hop back and identity error compounds.<code>H3 Multishot Sampler + Memory</code> splits the two jobs stock chaining conflates:<ul><li>Keyframe — what the video physically continues from. Still the most recent frame, so seams stay smooth.</li><li>Memory — what the encoder actually looks at: a persistent anchor from the start of the piece plus the last N shot-end frames. The anchor never changes, so drift cannot compound.</li></ul>Knobs: <code>anchor_frames</code> (1 = on, the long-chain fix) and <code>memory_frames</code> (recent frames the encoder also sees; 0 = stock behaviour). H3’s encoder takes multiple images natively, so this uses the model’s own mechanism, just deeper.Script parser self-repairsLong JSON scripts that lose their closing brace or bracket, end on a trailing comma, or leave a string unterminated are now auto-closed with a console warning instead of failing the render. A 4,500-character script missing one <code>}</code> is not a typo the author can see. Genuinely malformed scripts still fail loudly.Tip from testingRender at H3’s native resolution and upscale afterwards. Rendering natively at 1920x1088 scored worse than 960x544 in blind review — softer detail, and a reviewer called it “looks upscaled” — while costing about 4x the time.Models<ul><li>DiT quants — <a target="_blank" rel="ugc" href="https://huggingface.co/joeygambino/MiniMax-H3-GGUF">joeygambino/MiniMax-H3-GGUF</a></li><li>Text encoder + mmproj — <a target="_blank" rel="ugc" href="https://huggingface.co/joeygambino/MiniMax-H3-encoder-GGUF">joeygambino/MiniMax-H3-encoder-GGUF</a>. The mmproj sidecar is required for multi-shot, not just for reference images: shot chaining feeds the previous frame through the encoder’s vision path.</li><li>VAEs — <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/MiniMax-H3">Comfy-Org/MiniMax-H3</a></li></ul>SupportEverything here is free and stays free. If it saved you a night of debugging (it contains several hundred of mine), tips keep the 5090 warm: <a target="_blank" rel="ugc" href="https://ko-fi.com/joeygambino">Ko-fi</a> · <a target="_blank" rel="ugc" href="https://github.com/sponsors/jlucasmcrell">GitHub Sponsors</a> · <a target="_blank" rel="ugc" href="https://liberapay.com/joeygambino">Liberapay</a>.<a target="_blank" rel="ugc" href="https://github.com/sponsors/jlucasmcrell">nsors</a> · <a target="_blank" rel="ugc" href="https://liberapay.com/joeygambino">Liberapay</a>.

Files with SHA256: cc159fcd205edf5fb06ec0e2c8c203b9e2188ae6e256b04bc3cac1a11b9a0ab8