{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreicr6rcccoxuwfzzscx2okvz7cfqgychxt7jei4fi6ouo6a3btok7m",
    "uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mlnp7eh7yli2"
  },
  "path": "/t/wan2-2-i2v-clarifications-needed-regarding-settings-on-low-vram-system/175884#post_11",
  "publishedAt": "2026-05-12T09:23:30.000Z",
  "site": "https://discuss.huggingface.co",
  "tags": [
    "How to get the most out of prompts for WAN models",
    "Kijai ComfyUI-PromptRelay",
    "RunComfy Wan2.2 Prompt Relay workflow",
    "FFmpeg xfade filter documentation",
    "FFmpeg xfade crossfade guide",
    "ComfyUI official Wan2.2 guide",
    "Comfy Wan2.2 14B FLF2V workflow",
    "Comfy blog: Wan2.2 FLF2V native support",
    "Prompt Relay research/demo repo",
    "FizzNodes GitHub",
    "PromptSchedule documentation",
    "RunComfy PromptSchedule node page",
    "ComfyUI CLIPTextEncode documentation",
    "ComfyUI Community Manual: Text Prompts",
    "ComfyUI Wiki: prompt weighting",
    "ComfyUI Wiki: basic prompt syntax",
    "Phr00t WAN2.2 Rapid All-in-One"
  ],
  "textContent": "> is there any way to formulate the prompts that would imply . do action A, after action A is done do action B?\n\nThere seem to be several methods available, but some of them are difficult to use in an 8GB VRAM environment:\n\n* * *\n\n# Wan2.2 RapidBase I2V: sequential actions, prompt weights, and continuity-safe A → B workflows\n\nShort answer:\n\nYes, you can write prompts like **“do action A, then after A is done, do action B.”** The model can understand that language. But in a normal single-prompt I2V workflow, that instruction is usually a **soft temporal suggestion** , not a reliable frame-accurate command.\n\nFor the current RapidBase workflow, the safest ranking is:\n\n\n    1. One-clip two-beat prompt\n    2. Two clips with a handoff frame\n    3. Neutral overlap + short crossfade\n    4. FLF2V bridge clip\n    5. Prompt Relay\n    6. Prompt Schedule / FizzNodes\n\n\nFor prompt weights:\n\n\n    Avoid:\n      (((action:1.9)))\n\n    Prefer:\n      (same face and identity:1.10)\n      (preserve exact face:1.10)\n      (static camera:1.10)\n      (tiny natural smile:1.05)\n\n\nThe key rule is:\n\n> Weight preservation more than action.\n\nThe current workflow is working because it preserves the source image. Anything that pushes too hard toward complex action can also push the model into repainting, hallucination, or face drift.\n\n* * *\n\n## 1. Does the model understand “A, then B”?\n\nIt can understand the wording, but it does not necessarily execute it as an exact timeline.\n\nA prompt like this is understandable:\n\n\n    The same person first blinks once, then after a brief pause makes a tiny natural smile.\n\n\nBut in a normal I2V generation, the text prompt conditions the whole clip. It is not automatically split into exact frame ranges like:\n\n\n    frames 0-16:\n      action A\n\n    frames 17-33:\n      action B\n\n\nSo the model may interpret “first A, then B” loosely.\n\nPossible outcomes:\n\nPrompt | Possible model behavior\n---|---\n`blink once, then smile` | blink and smile happen in the right order\n`blink once, then smile` | smile starts before the blink finishes\n`blink once, then smile` | only the smile happens\n`look down, then look back` | gaze drifts vaguely instead of following exact order\n`A then B then C` | one action is skipped or the face starts drifting\n\nThis is normal for a single-prompt video model. The model sees the whole instruction, but it is not a strict animation timeline unless you use timeline-control tools.\n\nUseful background:\n\n  * How to get the most out of prompts for WAN models\n  * Kijai ComfyUI-PromptRelay\n  * RunComfy Wan2.2 Prompt Relay workflow\n\n\n\n* * *\n\n## 2. Best first method: one-clip two-beat prompting\n\nFor the current RapidBase workflow, this is the best first method.\n\nIt does not add nodes, LoRAs, bridge models, scheduling tools, or extra VRAM load. It also protects the main thing the current setup is good at:\n\n\n    same source image\n    same face\n    same lighting\n    same texture\n    same background\n    low hallucination\n\n\nThe limitation is that A → B order is only approximate.\n\n### Good use cases\n\nUse one-clip two-beat prompts for small actions:\n\n\n    blink once -> tiny smile\n    gentle breathing -> blink once\n    look slightly downward -> return eyes to camera\n    tiny smile -> neutral expression\n    eyes shift slightly left -> eyes return to camera\n    neutral expression -> tiny smile\n\n\n### Bad use cases\n\nAvoid large or multi-stage sequences:\n\n\n    turn head -> talk -> raise hand\n    walk forward -> gesture -> camera zooms in\n    look away -> laugh -> turn back\n    large smile -> speaking -> hair blowing\n    pose change -> lighting change -> background reaction\n\n\nEach extra action increases the chance of:\n\n\n    face drift\n    changed mouth shape\n    changed eye shape\n    new lighting\n    new camera angle\n    background mutation\n    AI-looking repainting\n\n\n* * *\n\n## 3. Good wording for “after A is done, do B”\n\nUse completion language, not just a loose list.\n\nWeak:\n\n\n    blink and smile\n\n\nBetter:\n\n\n    first blinks once, then after the blink is complete, slowly forms a tiny natural smile\n\n\nGood sequencing phrases:\n\n\n    first <action A>, then after a brief pause <action B>\n    after <action A> is complete, <action B>\n    begins still, then <action A>, then settles into <action B>\n    first holds a neutral expression, then gradually <action B>\n    after returning to neutral, <action B>\n\n\nAvoid vague or overloaded phrasing:\n\n\n    blink and smile naturally\n    perform a sequence of expressions\n    react emotionally\n    do a cute expression\n    move seductively\n    act naturally\n\n\nVague words invite the model to improvise. Improvisation is where identity drift usually starts.\n\n* * *\n\n## 4. Practical one-clip formula\n\nUse this structure:\n\n\n    [identity lock] + [starting state] + [action A] + [pause/settle] + [action B] + [camera lock] + [scene lock]\n\n\nExample:\n\n\n    The same person from the source image keeps the exact same face and identity. The video begins with a calm neutral expression. First, the person gently blinks once. After the blink is complete, the person slowly forms a tiny natural smile. Preserve the same hairstyle, clothing, lighting, colors, camera angle, and background. Static camera. No zoom. No pan. No scene change.\n\n\nThis is better than:\n\n\n    she blinks then smiles\n\n\nbecause it tells the model:\n\n\n    who must remain the same\n    what state to start from\n    what action comes first\n    what happens after\n    what must not change\n\n\n* * *\n\n## 5. One-clip two-beat prompt templates\n\n### Safest A → B template\n\n\n    The same person from the source image first blinks once, then after a brief pause makes a tiny natural smile. Preserve the exact same face, identity, hairstyle, clothing, lighting, colors, camera angle, and background. Static camera. No zoom. No pan. No scene change. Subtle natural motion only.\n\n\n### More explicit timing template\n\n\n    The video begins with the same person holding still. First, the person gently blinks once. After the blink is complete, the person slowly forms a tiny natural smile. Preserve the exact same face, identity, hairstyle, clothing, lighting, colors, camera angle, and background. Static camera. No scene change.\n\n\n### Face-first template\n\n\n    Preserve the exact same face and identity throughout the video. The same person first blinks once, then after a brief pause makes a tiny natural smile. Same hairstyle, clothing, lighting, colors, camera angle, and background. Static camera. No zoom. No scene change.\n\n\n### Short template\n\n\n    Same person, same face and identity. First one subtle blink, then a tiny natural smile. Static camera. Same lighting and background.\n\n\n### Very safe template\n\n\n    The same person keeps the exact same face and identity throughout the video. First, one small blink. Then, a tiny natural smile. Same hairstyle, clothing, lighting, colors, camera angle, and background. Static camera.\n\n\n* * *\n\n## 6. When a single prompt is not enough\n\nIf exact order matters, use two clips.\n\nDo not do this:\n\n\n    Clip 1:\n      original source image -> action A\n\n    Clip 2:\n      original source image -> action B\n\n\nThat creates two independent clips from the same original starting point. Clip 2 does not know where Clip 1 ended.\n\nBetter:\n\n\n    Clip 1:\n      original source image -> action A only\n\n    Handoff frame:\n      clean stable frame near the end of Clip 1\n\n    Clip 2:\n      handoff frame -> action B only\n\n\nThis is the most practical way to get reliable A → B ordering without adding complex nodes.\n\n* * *\n\n## 7. Handoff-frame workflow\n\n### Process\n\n\n    1. Generate Clip 1 with action A only.\n    2. Inspect the last 3-10 frames.\n    3. Do not blindly use the final frame.\n    4. Pick the cleanest stable frame:\n       - best face\n       - least blur\n       - stable lighting\n       - stable background\n       - expression suitable for the next action\n    5. Save that frame as PNG.\n    6. Use it as the source image for Clip 2.\n    7. Prompt Clip 2 for action B only.\n    8. Keep settings consistent:\n       - same resolution\n       - same FPS\n       - same VAE\n       - same text encoder\n       - same sampler\n       - same scheduler\n       - same CFG\n       - same steps\n       - same prompt style\n\n\n### Clip 1 example\n\n\n    The same person gently blinks once, then returns to a calm neutral expression. Preserve the same face, identity, lighting, clothing, colors, camera angle, and background. Static camera.\n\n\n### Clip 2 example\n\n\n    The same person begins from a calm neutral expression, then slowly forms a tiny natural smile. Preserve the same face, identity, lighting, clothing, colors, camera angle, and background. Static camera.\n\n\nThis is more reliable than trying to force a complex sequence into one prompt.\n\n* * *\n\n## 8. Neutral overlap and crossfade\n\nIf using two clips, make the join happen during a neutral moment.\n\nBad join:\n\n\n    Clip 1 ends during a blink.\n    Clip 2 starts with a smile.\n\n\nBetter join:\n\n\n    Clip 1 ends after returning to neutral.\n    Clip 2 starts from the neutral handoff frame.\n\n\nIf the join is slightly visible, use a short crossfade.\n\nTypical overlap:\n\n\n    4-8 frames\n    same FPS\n    same resolution\n    same color settings\n    same encoding settings\n\n\nFFmpeg example:\n\n\n    ffmpeg \\\n      -i clip1.mp4 \\\n      -i clip2.mp4 \\\n      -filter_complex \"xfade=transition=fade:duration=0.25:offset=2.75\" \\\n      -c:v libx264 -crf 18 -preset slow \\\n      output.mp4\n\n\n`offset` must be adjusted to match the length of Clip 1 and the desired transition point.\n\nReference:\n\n  * FFmpeg xfade filter documentation\n  * FFmpeg xfade crossfade guide\n\n\n\nImportant:\n\n> A crossfade can hide a small seam. It cannot fix a true face, lighting, or background mismatch.\n\nIf the face is different between the clips, a crossfade may create ghosting or a double-face dissolve.\n\n* * *\n\n## 9. FLF2V bridge clip\n\nFLF2V means **First-Last Frame to Video**.\n\nInstead of simply crossfading Clip A into Clip B, you provide:\n\n\n    first frame = stable end frame of Clip A\n    last frame  = stable start frame of Clip B\n\n\nThen the model generates the transition between them.\n\nConcept:\n\n\n    Clip A:\n      source -> action A\n\n    Clip B:\n      handoff/source -> action B\n\n    Bridge:\n      first frame = stable end frame of Clip A\n      last frame  = stable start frame of Clip B\n      prompt      = smooth subtle transition, same face, same lighting, static camera\n\n\nWhy it can help:\n\n\n    more natural transition than crossfade\n    can reduce a sudden jump between two clips\n    uses actual visual endpoints\n\n\nWhy it may not be ideal for the current 8GB RapidBase workflow:\n\n\n    separate workflow family\n    may be heavier\n    may not preserve the same RapidBase look\n    may introduce bloom or airbrushed style\n    may require more setup and testing\n\n\nUse FLF2V only if:\n\n\n    Clip A is good.\n    Clip B is good.\n    The join is visibly bad.\n    A simple crossfade is not good enough.\n    The bridge can be short.\n\n\nReferences:\n\n  * ComfyUI official Wan2.2 guide\n  * Comfy Wan2.2 14B FLF2V workflow\n  * Comfy blog: Wan2.2 FLF2V native support\n\n\n\n* * *\n\n## 10. Prompt Relay\n\nPrompt Relay is closer to the real solution for “A happens in one segment, B happens in another segment.”\n\nInstead of relying on a single prompt, Prompt Relay routes different prompts through different temporal segments.\n\nConcept:\n\n\n    Global prompt:\n      same person, same face, same identity, same lighting, same background, static camera\n\n    Segment 1:\n      blink once\n\n    Segment 2:\n      tiny natural smile\n\n\nWhy it is attractive:\n\n\n    A and B happen inside one timeline\n    less independent-clip continuity drift\n    global identity/camera constraints can stay active\n    different segments can receive different action prompts\n\n\nWhy it should be treated carefully:\n\n\n    changes the workflow structure\n    may not plug cleanly into the current RapidBase GGUF workflow\n    may increase complexity\n    8GB behavior is uncertain\n    could break the source-fidelity look\n\n\nDo not add it to the working workflow directly. Test only in a duplicate workflow.\n\nReferences:\n\n  * Kijai ComfyUI-PromptRelay\n  * RunComfy Wan2.2 Prompt Relay workflow\n  * Prompt Relay research/demo repo\n\n\n\n* * *\n\n## 11. Prompt Schedule / FizzNodes\n\nPrompt scheduling is the general concept of changing prompt conditioning over time.\n\nConcept:\n\n\n    Frames 0-16:\n      same person gently blinks once\n\n    Frames 17-33:\n      same person slowly forms a tiny smile\n\n\nWhy it may help:\n\n\n    more explicit temporal control\n    frame/segment-based prompt changes\n    better than hoping a single prompt follows order\n\n\nWhy it is not the first recommendation here:\n\n\n    not guaranteed to fit the current RapidBase GGUF workflow\n    can change conditioning behavior\n    may break the current source-fidelity look\n    adds complexity\n\n\nReferences:\n\n  * FizzNodes GitHub\n  * PromptSchedule documentation\n  * RunComfy PromptSchedule node page\n\n\n\n* * *\n\n## 12. Recommended method ranking\n\nRank | Method | A → B reliability | Source fidelity | 8GB friendliness | Recommendation\n---|---|---|---|---|---\n1 | One-clip two-beat prompt | Medium | High | High | Try first\n2 | Two clips + handoff frame | High | Medium-high | High | Best practical method\n3 | Two clips + neutral crossfade | Medium-high | Medium-high | High | Good polish\n4 | FLF2V bridge | High for transition | Medium | Medium-low | Separate experiment\n5 | Prompt Relay | High conceptually | Unknown | Unknown | Advanced experiment\n6 | Prompt Schedule / FizzNodes | Medium-high conceptually | Unknown | Medium | Experimental\n\nBest practical rule:\n\n\n    simple A -> B:\n      use one-clip two-beat prompt\n\n    strict A -> B:\n      use two clips with a handoff frame\n\n    smooth transition:\n      use handoff frame + optional crossfade\n\n    true timeline control:\n      test Prompt Relay or Prompt Schedule only in a duplicate workflow\n\n\n* * *\n\n## 13. Prompt weights: do they work?\n\nYes, ComfyUI prompt weights can work.\n\nCommon syntax:\n\n\n    (phrase:1.2)\n\n\nPlain parentheses also increase weight. ComfyUI’s CLIPTextEncode documentation says plain parentheses apply a default weight of `1.1`, and the ComfyUI Community Manual says nested weights multiply.\n\nExamples:\n\n\n    (phrase)\n      roughly increases emphasis\n\n    (phrase:1.2)\n      explicit weight\n\n    ((phrase:1.2):0.5)\n      nested weights multiply\n\n\nReferences:\n\n  * ComfyUI CLIPTextEncode documentation\n  * ComfyUI Community Manual: Text Prompts\n  * ComfyUI Wiki: prompt weighting\n  * ComfyUI Wiki: basic prompt syntax\n\n\n\n* * *\n\n## 14. Is `(((weighted prompts:1.9)))` useful here?\n\nProbably not. For this workflow, it is more likely to hurt than help.\n\nAvoid:\n\n\n    (((turns head and smiles:1.9)))\n\n\nAvoid:\n\n\n    (((first blinks then smiles:1.9)))\n\n\nAvoid:\n\n\n    (((action A then action B:1.9)))\n\n\nWhy? Because a huge action weight tells the model:\n\n\n    This action matters more than preserving the source image.\n\n\nThat can cause:\n\n\n    face drift\n    changed facial geometry\n    changed skin texture\n    changed lighting\n    hallucinated details\n    mouth/teeth weirdness\n    background changes\n    overcooked motion\n    loss of source fidelity\n\n\nThe current RapidBase workflow works because it is conservative. Heavy action weights fight that.\n\n* * *\n\n## 15. Better prompt-weight strategy\n\nDo not heavily weight the action. Lightly weight preservation.\n\nBetter:\n\n\n    The same person first blinks once, then after a brief pause makes a tiny natural smile. (Preserve the exact same face and identity:1.15). Same hairstyle, clothing, lighting, colors, camera angle, and background. (Static camera:1.10). No zoom. No scene change.\n\n\nRisky:\n\n\n    The same person (((first blinks then smiles:1.9))). Same face and background.\n\n\nThe first prompt says:\n\n\n    identity and camera are important\n    action is small\n    do not repaint\n\n\nThe second prompt says:\n\n\n    force this action even if the model has to invent\n\n\nFor face permanence, that is the wrong priority.\n\n* * *\n\n## 16. Suggested weight ranges\n\nWeight | Use\n---|---\n`1.00` | normal baseline\n`1.05` | tiny emphasis\n`1.10` | safe emphasis\n`1.15` | useful emphasis for identity/static camera\n`1.20` | upper normal test\n`1.25` | mild stress test\n`1.35` | risky; use sparingly\n`1.50+` | likely too strong\n`1.90` | avoid for source-faithful I2V\n\nFor this workflow, use mostly:\n\n\n    1.05-1.20\n\n\nMaybe test:\n\n\n    1.25\n\n\nAvoid:\n\n\n    1.50+\n    1.90\n    triple-parentheses action forcing\n\n\n* * *\n\n## 17. What to weight\n\n### Good things to weight\n\n\n    (same face and identity:1.10)\n    (preserve exact face:1.10)\n    (preserve source image:1.10)\n    (static camera:1.10)\n    (no scene change:1.05)\n    (same lighting and background:1.10)\n    (tiny natural smile:1.05)\n    (one subtle blink:1.05)\n\n\n### Risky things to weight\n\n\n    (turns head:1.4)\n    (speaks:1.4)\n    (laughs widely:1.4)\n    (raises hand:1.4)\n    (hair blowing:1.4)\n    (camera zooms in:1.4)\n\n\n### Very risky\n\n\n    (((wide smile:1.9)))\n    (((speaking:1.9)))\n    (((turning head:1.9)))\n    (((complex action sequence:1.9)))\n\n\nLarge facial motion and mouth motion are exactly where face permanence usually breaks.\n\n* * *\n\n## 18. Weighted A → B examples\n\n### Safe weighted A → B prompt\n\n\n    The same person from the source image first blinks once, then after a brief pause makes a tiny natural smile. (Preserve the exact same face and identity:1.15). Same hairstyle, clothing, lighting, colors, camera angle, and background. (Static camera:1.10). No zoom. No pan. No scene change.\n\n\n### Slightly stronger action prompt\n\n\n    The same person begins still, then (gently blinks once:1.05), then slowly forms a (tiny natural smile:1.10). Preserve the exact same face, identity, hairstyle, clothing, lighting, colors, camera angle, and background. Static camera. No scene change.\n\n\n### Face-first prompt\n\n\n    (Preserve the exact same face and identity:1.15). The same person first blinks once, then slowly forms a tiny natural smile. Same hairstyle, clothing, lighting, colors, camera angle, and background. (Static camera:1.10). No zoom. No scene change.\n\n\n### Minimal weighted prompt\n\n\n    (Same face and identity:1.15). First one subtle blink, then a tiny smile. Same lighting and background. (Static camera:1.10).\n\n\n* * *\n\n## 19. What not to do\n\nAvoid this:\n\n\n    (((The person first blinks, then smiles, then turns their head, then speaks:1.9)))\n\n\nThat stacks three problems:\n\n\n    too many actions\n    too much weight\n    weighting the part that causes identity drift\n\n\nAlso avoid:\n\n\n    First she blinks, then smiles, then speaks, then turns her head, while the camera zooms in and the lighting becomes cinematic.\n\n\nThat asks the model to solve:\n\n\n    facial motion\n    mouth motion\n    head rotation\n    camera motion\n    lighting change\n    identity preservation\n    background stability\n\n\nThat is too much for a source-faithful RapidBase clip.\n\n* * *\n\n## 20. Practical A → B workflow\n\n### Step 1 — choose the smallest version of the action\n\nInstead of:\n\n\n    turns head and smiles\n\n\nuse:\n\n\n    tiny eye movement and tiny smile\n\n\nInstead of:\n\n\n    speaks\n\n\nuse:\n\n\n    subtle mouth movement\n\n\nInstead of:\n\n\n    laughs\n\n\nuse:\n\n\n    tiny natural smile\n\n\n### Step 2 — write a two-beat prompt\n\n\n    The same person first <action A>, then after a brief pause <action B>. Preserve the exact same face, identity, hairstyle, clothing, lighting, colors, camera angle, and background. Static camera. No scene change.\n\n\n### Step 3 — add light preservation weights\n\n\n    (Preserve the exact same face and identity:1.15)\n    (Static camera:1.10)\n\n\n### Step 4 — batch seeds\n\n\n    8-16 seeds\n    same prompt\n    same settings\n    short preview\n\n\nPick by:\n\n\n    1. face permanence\n    2. correct order\n    3. natural motion\n    4. prompt obedience\n\n\n### Step 5 — split into two clips if order fails\n\nIf the model keeps blending A and B, use:\n\n\n    Clip 1:\n      action A only\n\n    Handoff:\n      clean stable frame near the end of Clip 1\n\n    Clip 2:\n      action B only\n\n\n* * *\n\n## 21. Best practical recommendation\n\nFor the current RapidBase workflow:\n\n\n    Use one-clip two-beat prompts first.\n    Use \"first A, then after a brief pause B.\"\n    Keep A and B very small.\n    Batch seeds.\n    Weight preservation, not action.\n    Avoid 1.9 weights.\n    Use handoff-frame two-clip generation when strict order matters.\n    Only test Prompt Relay / scheduling in a duplicate workflow.\n\n\n* * *\n\n## 22. Example final prompts\n\n### Blink → smile\n\n\n    The same person from the source image first blinks once, then after a brief pause makes a tiny natural smile. (Preserve the exact same face and identity:1.15). Same hairstyle, clothing, lighting, colors, camera angle, and background. (Static camera:1.10). No zoom. No pan. No scene change.\n\n\n### Look down → return gaze\n\n\n    The same person from the source image first looks slightly downward with only a tiny eye movement, then returns the eyes to the camera. (Preserve the exact same face and identity:1.15). Same hairstyle, clothing, lighting, colors, camera angle, and background. (Static camera:1.10). No zoom. No scene change.\n\n\n### Neutral → tiny smile\n\n\n    The video begins with the same person holding a calm neutral expression. Then the person slowly forms a tiny natural smile. (Preserve the exact same face and identity:1.15). Same hairstyle, clothing, lighting, colors, camera angle, and background. Static camera. No scene change.\n\n\n### Breathing → blink\n\n\n    The same person from the source image keeps the exact same face and identity. The person gently breathes with subtle natural motion, then blinks once after a brief pause. Same hairstyle, clothing, lighting, colors, camera angle, and background. Static camera. No zoom. No scene change.\n\n\n* * *\n\n## 23. Reference links\n\n### Wan prompting / CFG behavior\n\n  * How to get the most out of prompts for WAN models\n\n\n\n### Rapid / AIO background\n\n  * Phr00t WAN2.2 Rapid All-in-One\n\n\n\n### FLF2V / Wan2.2 workflows\n\n  * ComfyUI official Wan2.2 guide\n  * Comfy Wan2.2 14B FLF2V workflow\n  * Comfy blog: Wan2.2 FLF2V native support\n\n\n\n### Prompt Relay / timeline control\n\n  * Kijai ComfyUI-PromptRelay\n  * RunComfy Wan2.2 Prompt Relay workflow\n  * Prompt Relay research/demo repo\n\n\n\n### Prompt scheduling\n\n  * FizzNodes GitHub\n  * PromptSchedule documentation\n  * RunComfy PromptSchedule node page\n\n\n\n### Prompt weighting\n\n  * ComfyUI CLIPTextEncode documentation\n  * ComfyUI Community Manual: Text Prompts\n  * ComfyUI Wiki: prompt weighting\n  * ComfyUI Wiki: basic prompt syntax\n\n\n\n### Crossfade\n\n  * FFmpeg xfade filter documentation\n  * FFmpeg xfade crossfade guide\n\n",
  "title": "Wan2.2 i2v (clarifications needed regarding settings on low vram system)"
}