media-workflow

vg generate workflow production

Runtime: All endpoint calls use the vg generate CLI (npm install -g vibedgames, or pnpm dogfood in this repo). The API key lives on the vibedgames server, so there is no per-machine setup. See the generate skill for the command reference.

Use this skill when a single model call is not enough. There are two ways in:

• Use-case recipe — your task matches a known kind of content production (commercial, character, lip-sync, restoration…). Start from the recipe table below; each recipe lists inputs, the vg generate call sequence, and a quality bar.
• Custom pipeline — no recipe matches. Design the workflow from scratch using the orchestration patterns in this skill.

A workflow either way is a planned sequence of vg generate calls with clear inputs, outputs, dependencies, and quality checks.

Use-case recipes

Match the user's intent to a recipe, then load that reference. If two apply (e.g. "commercial featuring a consistent character"), load both and run the more specific one first. If the task is a single endpoint call, skip recipes and go straight to the right model-catalog reference.

Reference	Use for
cinematography.md	Cinematic stills and video, shot language, lighting, lens, color grade
character-design.md	Original characters with consistent identity across shots
commercial.md	Product photography, ads, e-commerce batches, hero shots
storytelling.md	Multi-shot narratives, short films, ads, brand films, social stories
character-lipsync.md	Talking head / lip-sync video (TTS → animated portrait)
image-restoration.md	Smart-dispatch restoration, deblur, denoise, dehaze, fix faces, document restore
virtual-tryon.md	Apply a garment onto a person photo (with optional cleanup chain)
video-with-audio.md	Add narration / SFX / music to a silent video
product-shot.md	Hero product photography from a packshot reference
realism.md	Photoreal stills (candid, editorial, documentary, archival, food, nature, architectural) with an anti-AI-look checklist

Each recipe links to model-catalog for endpoint defaults rather than listing models inline, so the catalog stays the single source of truth.

3D assets have no recipe here — they are routed by model-catalog (text-to-3d / image-to-3d): rigged characters → regenerate-3d; rigged or openable props as procedural code → image-to-threejs.

Custom pipelines

Load these references as needed:

• references/pipeline-patterns.md
• references/node-rules.md
• references/utility-endpoints.md
• references/recipes.md — generic workflow recipes (multi-scene video, dataset, social batch…)
• model-catalog for creative model defaults

Use model-catalog for default creative model choices. Still inspect schemas, check pricing when cost matters, and use exact endpoint fields.

Inputs to collect

Ask only for missing information that changes the pipeline:

• Final deliverable: image set, video, clips, audio, subtitles, dataset, social batch, product campaign, storyboard, style exploration.
• Source assets: product images, character references, first frames, video, audio, logo, transcript, brand guide.
• Runtime limits: quality target, cost sensitivity, number of variants, duration, aspect ratios, deadline.
• Continuity requirements: product identity, character face, scene layout, voice, color grade.
• Model preference: ask the user only when quality, speed, cost, or audio tradeoffs are not clear from the brief.

Core workflow

1. Write a short pipeline graph before running anything.

   input assets -> planner -> generation nodes -> utility nodes -> QA -> final outputs
   

2. Resolve endpoints for each role. Check known endpoint IDs first.

   vg generate models --endpoint_id openai/gpt-image-2 --json
   vg generate models --endpoint_id fal-ai/nano-banana-pro/edit --json
   vg generate models --endpoint_id bytedance/seedance-2.0/image-to-video --json
   vg generate models --endpoint_id xai/grok-imagine-video/image-to-video --json
   vg generate models --endpoint_id veed/fabric-1.0 --json
   

   Use text search only as fallback discovery for roles not covered by model-catalog or the utility reference:

   vg generate models "image generation product photography" --json
   vg generate models "image editing reference preservation" --json
   vg generate models "image to video" --json
   vg generate models "subtitle video utility" --json
   vg generate docs "workflow utility endpoints" --json
   

3. Inspect every endpoint before use.

   vg generate schema <endpoint_id> --json
   vg generate pricing <endpoint_id> --json
   

4. Upload local files once and reuse returned URLs.

   vg generate upload ./input.png --json
   vg generate upload ./voiceover.wav --json
   

5. Run each node with JSON output. Use async for slow generation.

   vg generate run <endpoint_id> --<field> "<value>" --json
   vg generate run <endpoint_id> --<field> "<value>" --async --json
   vg generate status <endpoint_id> <request_id> --download "./outputs/workflow/{request_id}_{index}.{ext}" --json
   

6. For downstream nodes, pass the media URL from the previous result when it is available. If you only have a local file path, upload it first.

7. Download final assets with templates that cannot collide.

   --download "./outputs/workflow/{request_id}_{index}.{ext}"
   

8. Return a compact manifest.

   {
     "goal": "short deliverable description",
     "nodes": [
       {
         "id": "shot_01",
         "role": "image_to_video",
         "endpoint_id": "...",
         "request_id": "...",
         "input_urls": ["..."],
         "output_urls": ["..."],
         "downloaded_files": ["..."],
         "notes": "continuity or defect notes"
       }
     ],
     "final_files": ["..."]
   }
   

Pipeline rules

• Keep one node responsible for one clear transformation.
• Fan out independent generation, crop, upscale, subtitle, or variation nodes.
• Keep sequential chains only when node B needs node A output.
• For consistency, prefer reference/edit or image-to-video over independent text-only generations.
• For default creative model choices, follow model-catalog unless the user names a model.
• Use utility endpoints for deterministic work: crop, resize, grid, composite, audio merge, subtitle, speed change, compression.
• Record endpoint, schema-relevant parameters, request ID, and output path for every node.
• If a 422 error occurs, read validation_errors, inspect schema again, then fix the exact field.

Quality gate

Before returning, verify:

• The pipeline graph matches the requested deliverable.
• No generation model was chosen from memory alone.
• All local source files were uploaded before use.
• Final files were saved through --download.
• Utility endpoints used exact schema fields.
• Continuity anchors were repeated where identity or product fidelity matters.
• Each node output is either accepted, retried, or marked with a defect.

If the workflow becomes too complex, stop expanding and ask the user to choose between faster iteration, higher fidelity, or broader variation.