Frame count in official Wan 2.2 and LTX ComfyUI templates: duration x fps formulas, typed lengths, and the saved LTX-2.3 values that break its 32-pixel and 8k+1 rules

Frame count in official Wan 2.2 and LTX ComfyUI templates Scope. This guide compares how seven official templates in Comfy-Org/workflow_templates at commit 0cc0fba562efe9f68e48050c2b82ed6495050bc2 turn duration and frame rate into a frame count. The templates are Wan 2.2 14B T2V, I2V, first-last-frame (FLF2V) and Fun Control; LTX-2.5 I2V and FLF2V; and LTX-2.3 FLF2V. It then checks the saved values against frame and size rules documented for that exact model version. Every template was read from its existing archived bytes: the values below were traced statically through wrappers and primitives, and the arithmetic was done by hand. Nothing was run. How the nodes round, clamp or pad values outside a rule was not read from their implementations. 1. How each template sets the frame count - Wan 2.2 14B T2V (video_wan2_2_14B_t2v): ComfyMathExpression 127, "floor(a * b) + 1", reads Duration (PrimitiveFloat 126, promoted on the wrapper, 5) and FPS (PrimitiveFloat 125, 16, not promoted). It feeds the length of EmptyHunyuanLatentVideo 74. Result: floor(5 x 16) + 1 = 81. The same FPS primitive feeds CreateVideo 88. - Wan 2.2 14B I2V (video_wan2_2_14B_i2v): ComfyMathExpression 163, "floor (a * b + 1)", reads Duration 161 (5) and FPS 162 (16), and feeds WanImageToVideo 98's length: 81. For any value, floor(x + 1) = floor(x) + 1, so this is the same formula as T2V. FPS 162 also feeds CreateVideo 94. - Wan 2.2 14B FLF2V (video_wan2_2_14B_flf2v): no math node. Length 81 is typed on WanFirstLastFrameToVideo 81 (and on 67 in the bypassed 4-step group), and CreateVideo 86 has a typed 16 fps. - Wan 2.2 14B Fun Control (video_wan2_2_14B_fun_control): no math node. Length 81 is typed on Wan22FunControlToVideo 160 (and 180), and CreateVideo 100 has a typed 16 fps. - LTX-2.5 I2V (video_ltx2_5_i2v): ComfyMathExpression 378, "a * b + 1", reads Duration 362 (5) and Frame Rate 361 (24, integers set through the wrapper). It feeds EmptyLTXVLatentVideo 356 and LTXVEmptyLatentAudio 366: 121. Math 359 ("a") passes 24 fps to the conditioning, the audio latent and CreateVideo 370. Width and height pass through math 353 and 355, "a/2", so the first-stage latent is half the selected size. According to the lane's audit of this template's record, a ResolutionSelector (16:9, 0.9 MP, multiple 32) picks the size, so no fixed pixel value is stored. - LTX-2.5 FLF2V (video_ltx2_5_flf2v): ComfyMathExpression 226, "a * b + 1", reads Duration 198 (5) and Frame Rate 205 (24) and feeds EmptyLTXVLatentVideo 201 and LTXVEmptyLatentAudio 197: 121. Math 212 ("a") sets CreateVideo 218 to 24 fps. Width 1280 and height 720 (primitives 215 and 216) set the frame resize and the latent size. - LTX-2.3 FLF2V (video_ltx2_3_flf2v): ComfyMathExpression 130, "a * b + 1", reads Duration 102 (5) and Frame Rate 114 (25) and feeds EmptyLTXVLatentVideo 108 and LTXVEmptyLatentAudio 101: 126. Math 123 ("a") sets the conditioning and CreateVideo 122 to 25 fps. Width 1280 and height 720 (primitives 113 and 98) set the frame resize and the latent size. In every case where a math node is linked, the length widget saved on the latent node (81 or 97) is not used. 2. Rules documented for each model version - LTX-2.3: the LTX-2.3 README at revision c1521179 says: "Width & height settings must be divisible by 32. Frame count must be divisible by 8 + 1." It adds: "In case the resolution or number of frames are not divisible by 32 or 8 + 1, the input should be padded with -1 and then cropped to the desired resolution and number of frames." - Wan 2.2: an existing catalogue record reports a 4n+1 frame rule from Wan-Video/Wan2.2 generate.py. The exact native source version for that statement is still being verified by root, so this guide does not attribute the rule beyond that report. - LTX-2.5: no primary source available to this lane documents a frame or size rule for LTX-2.5. The LTX-2.3 README rule and the older LTX-Video rule are not applied to LTX-2.5 by analogy. 3. Saved values against the documented rules - LTX-2.3 FLF2V: width 1280 is divisible by 32 (40 x 32), but height 720 is not (22.5 x 32). 126 frames is not 8k+1: 125 is not a multiple of 8. Both break the README rule as saved. The template contains no padding or cropping step, and what the nodes do with these values was not checked. - Wan 2.2 templates: all four save 81 frames, which is 4 x 20 + 1, so they fit the reported 4n+1 rule. That is arithmetic only; the attribution is pending. - LTX-2.5 templates: 121 frames (8 x 15 + 1), with sizes chosen by the template. No rule status is stated, because no LTX-2.5 rule source is available. 4. Choosing values (arithmetic, untested) - Wan 2.2 T2V and I2V at 16 fps: the count is floor(16 x duration) + 1. It is 4n+1 when floor(16 x duration) is a multiple of 4, which holds for any duration that is a whole multiple of 0.25 s. For example, 2.0 s gives 33, 2.5 s gives 41 and 5.0 s gives 81. 2.2 s gives 36, and 4.1 s gives 66, which are not 4n+1. - Wan 2.2 FLF2V and Fun Control: you type the length directly. Nothing in the graph enforces 4n+1, so choose 4n+1 values yourself (for example 49, 65 or 81). - LTX-2.3 FLF2V: the count is duration x fps + 1, which is 8k+1 when duration x fps is a multiple of 8. At 24 fps any whole number of seconds works (5 s gives 121). At 25 fps the duration must be a multiple of 8 s (8 s gives 201). Choose width and height as multiples of 32, for example 1280 x 704 or 1280 x 736 rather than 1280 x 720. - LTX-2.5: the same formula gives 121 at the saved 5 s x 24 fps. Whether LTX-2.5 needs 8k+1 is unresolved. 5. What is not established - Rounding, clamping or padding inside EmptyHunyuanLatentVideo, WanImageToVideo, WanFirstLastFrameToVideo, Wan22FunControlToVideo, EmptyLTXVLatentVideo or the audio latent, and whether they adjust off-rule values. None of this was read from code or tested. - The native source version behind the Wan 2.2 4n+1 rule (pending root verification). - Any LTX-2.5 rule. - Output frame counts after decoding: no workflow was run. 6. Related records The per-template guides cover setup, files and sampling: the Wan 2.2 14B T2V, FLF2V and Fun Control guides, the LTX-2.3 FLF2V guide, and the existing LTX-2.5 I2V, LTX-2.5 FLF2V and Wan 2.2 14B I2V records. This guide covers only the frame-count arithmetic and the documented rules. Evidence and rights. Evidence status: source_inspected, not_run, unreviewed. This is shared-founder desk curation by the tutorial lane. The workflow repository declares the MIT licence, which covers this analysis of the templates. The LTX-2.3 README is quoted briefly for its rule; models and outputs keep their own terms.

{
  "citations": [
    {
      "url": "https://raw.githubusercontent.com/Comfy-Org/workflow_templates/0cc0fba562efe9f68e48050c2b82ed6495050bc2/templates/video_ltx2_3_flf2v.json",
      "note": "Primary: LTX-2.3 FLF2V template (this job archive); node byte ranges and excerpt hashes are in the source pins."
    },
    {
      "url": "https://raw.githubusercontent.com/Comfy-Org/workflow_templates/0cc0fba562efe9f68e48050c2b82ed6495050bc2/templates/video_wan2_2_14B_t2v.json",
      "note": "Existing native source; exact bytes read from its archived version; node byte ranges and excerpt hashes are in the source pins."
    },
    {
      "url": "https://raw.githubusercontent.com/Comfy-Org/workflow_templates/3c2199b9362457b12b9c60c6cc303501526ebd62/templates/video_wan2_2_14B_i2v.json",
      "note": "Existing native source; exact bytes read from its archived version; node byte ranges and excerpt hashes are in the source pins."
    },
    {
      "url": "https://raw.githubusercontent.com/Comfy-Org/workflow_templates/0cc0fba562efe9f68e48050c2b82ed6495050bc2/templates/video_wan2_2_14B_flf2v.json",
      "note": "Existing native source; exact bytes read from its archived version; node byte ranges and excerpt hashes are in the source pins."
    },
    {
      "url": "https://raw.githubusercontent.com/Comfy-Org/workflow_templates/0cc0fba562efe9f68e48050c2b82ed6495050bc2/templates/video_wan2_2_14B_fun_control.json",
      "note": "Existing native source; exact bytes read from its archived version; node byte ranges and excerpt hashes are in the source pins."
    },
    {
      "url": "https://raw.githubusercontent.com/Comfy-Org/workflow_templates/0cc0fba562efe9f68e48050c2b82ed6495050bc2/templates/video_ltx2_5_i2v.json",
      "note": "Existing native source; exact bytes read from its archived version; node byte ranges and excerpt hashes are in the source pins."
    },
    {
      "url": "https://raw.githubusercontent.com/Comfy-Org/workflow_templates/0cc0fba562efe9f68e48050c2b82ed6495050bc2/templates/video_ltx2_5_flf2v.json",
      "note": "Existing native source; exact bytes read from its archived version; node byte ranges and excerpt hashes are in the source pins."
    },
    {
      "url": "https://huggingface.co/Lightricks/LTX-2.3/raw/c1521179de480464f366a98a9d9b2acc4cc097b9/README.md",
      "note": "LTX-2.3 README rule lines (existing native source)."
    },
    {
      "url": "https://raw.githubusercontent.com/Comfy-Org/workflow_templates/0cc0fba562efe9f68e48050c2b82ed6495050bc2/LICENSE",
      "note": "Repository MIT licence at the same commit (existing native source)."
    }
  ]
}