CentralFloridaAttorney-ComfyUI-ZMongo/example_workflows/00_README_usage.md

ZMongo LTXV Modular Workflows

These files split the original oversized workflow into independently runnable ComfyUI API/prompt JSON files.

The uploaded source workflow was an API-style prompt JSON, not a visual UI workflow with node positions and group boxes. For reliability, the modular version is separated into smaller runnable workflow JSON files instead of one giant graph.

Shared defaults

• Image collection: succubi
• Image field path: image_data
• Mask field path: mask_data
• Prompt collection: prompts
• LTXV positive prompt path: prompt.ltxv_i2v_121f_16fps_720p.positive
• LTXV system prompt path: prompt.system.ltxv_i2v_121f_16fps_720p
• Target video length: 121 frames
• Target frame rate: 16 FPS
• Target image/video prep size: 1280x720

The API key field was intentionally cleared. Paste your API key into node 260, titled:

00 API Key Session - paste key here

The session switch node 261 controls hosted API vs local file store.

Suggested use order

1. Select or preview source image

Open:

00_select_image_browser.workflow.json

Use this to list image records, browse thumbnails, and preview the selected image.

Important nodes:

• 270 — image collection name
• 244 — image field path
• 245 — list docs
• 279 — selected image index
• 291 — display selected image
• 290 — preview selected image

Run this workflow by itself when you only want to change or confirm the source image.

2. Save or update the LTXV system prompt

Open:

01_save_ltxv_system_prompt.workflow.json

This saves the Gemini prompt-constructor system prompt into:

prompt.system.ltxv_i2v_121f_16fps_720p

Important nodes:

• 325 — editable system prompt text
• 322 — system prompt field path
• 286 — save value node
• 287 — save-result preview

Run this only when changing the system instruction.

3. Manually save a positive video prompt

Open:

02_manual_save_ltxv_positive_prompt.workflow.json

Use this when you already have a finished LTXV prompt and want to save it without calling Gemini.

Important nodes:

• 900 — manual positive prompt text
• 295 — save generated/manual positive prompt
• 294 — save-result preview

The prompt saves to:

prompt.ltxv_i2v_121f_16fps_720p.positive

4. Generate a positive LTXV prompt from the selected image

Open:

03_generate_ltxv_prompt_from_selected_image.workflow.json

Use this when you want Gemini to look at the selected image and the rough concept, then produce a clean LTXV image-to-video prompt and save it.

Important nodes:

• 289 — rough user video concept
• 279 — selected image index
• 291 — selected image
• 325 — system prompt
• 301 — Gemini image + text prompt generator
• 295 — save generated prompt
• 293 — Gemini response preview
• 294 — save-result preview

This workflow does not render video. It only generates and saves the video prompt.

5. Select/load an existing video prompt

Open:

04_select_load_ltxv_video_prompt.workflow.json

Use this to select a prompt document and preview the saved positive prompt before rendering.

Important nodes:

• 262 — prompt collection name
• 267 — list prompt docs
• 283 — selected prompt document index
• 296 — get saved LTXV positive prompt
• 297 — loaded prompt preview

6. Render simple LTXV video

Open:

05_render_ltxv_video_simple.workflow.json

This is the smaller LTXV render path using:

• ltx-video-2b-v0.9.5.safetensors
• t5xxl_fp16.safetensors
• LTXVImgToVideo
• 121 frames
• 16 FPS

Important nodes:

• 279 — selected source image index
• 296 / 297 — loaded positive prompt
• 77 — LTXV image-to-video node
• 72 — sampler
• 81 — video save node

Note: node 77 uses 1280x736 because some LTXV/video latent paths require dimensions divisible by model-friendly latent sizes. The source image selection/prep is still 1280x720.

7. Render advanced LTXV video/audio path

Open:

06_render_ltxv_video_advanced.workflow.json

This is the larger advanced path using:

• ltx-2-19b-distilled.safetensors
• gemma_3_12B_it_fp4_mixed.safetensors
• latent upscaler
• audio latent path
• 121 frames
• 16 FPS

Important nodes:

• 279 — selected source image index
• 296 / 297 — loaded positive prompt
• 310 — resize selected image to 1280x720
• 316:62 — frame length
• 321 — frame rate
• 316:3 — loaded prompt text encode
• 315 — save final video

Use this only after the prompt and selected image are correct.

8. Generate and save a new reference image

Open:

07_generate_and_save_reference_image.workflow.json

Use this to generate a new still image from the selected/loaded prompt and save it back to ZMongo.

Important nodes:

• 296 / 297 — loaded prompt
• 236 — SD3.5 checkpoint
• 280 — image sampler
• 240 — save generated image
• 281 — save-result preview

9. Save and browse mask from selected image

Open:

08_save_and_browse_mask_from_selected_image.workflow.json

Use this only when you need a mask generated from the selected image and saved to the mask field path.

Important nodes:

• 291 — selected image
• 274 — image to mask
• 248 — mask to image
• 247 — save mask image
• 250 — browse saved masks
• 276 / 277 — previews

Practical workflow pattern

1. Run 00_select_image_browser.workflow.json
2. Run 01_save_ltxv_system_prompt.workflow.json only if the system prompt changed
3. Run 03_generate_ltxv_prompt_from_selected_image.workflow.json
4. Run 04_select_load_ltxv_video_prompt.workflow.json
5. Run either:
   • 05_render_ltxv_video_simple.workflow.json, or
   • 06_render_ltxv_video_advanced.workflow.json

Why this is more functional

The old workflow mixed these actions into one graph:

• source-image selection
• system-prompt saving
• Gemini prompt construction
• prompt saving
• prompt loading
• image generation
• mask generation
• simple LTXV rendering
• advanced LTXV rendering

That made ComfyUI try to evaluate unrelated nodes. These modular files prevent unnecessary model loading and let each workflow do one job.