On-Model Bulk Create (100 products per run): on-model PDPs in Bubble alongside BIAIG flat bulk
A decision matrix for Bubble fashion catalogs: when to run On-Model batch (up to 100 tags per run) outside the Bubble UI, when to keep flat/lifestyle bulk on BIAIG orchestration routes, and how to keep one SKU record as source of truth.
Why now
The digest from 15.10 called out On-Model Bulk Create — up to 100 products per run. I keep running into the same fork for apparel and marketplace builds: on-model PDP tooling (flat→model, model swap, packshot, recolor) versus flat generative bulk I can orchestrate from Bubble.
I did not write this to pick a winner. I needed a workflow split—Bubble holds catalog records; image production runs on separate lanes with different caps and QA.
On-Model Bulk Create (documentation facts)
On-Model’s in-app Bulk Create wizard uses one shared setup per run, then one job per product tag, capped at 100 tags per run. Each tag should map to a SKU-style code, accept 1–10 input images, and carry the tag on results for gallery filtering. Supported job types:
- Flat to Model
- Model Swap
- Create Packshot
- Garment Recolor
For Garment Recolor, I plan around the fact that one “job” is often a product × color pair. The October 2026 release example—20 products × 5 colors = 100 jobs—is not the same as “100 products” for every colorway mix.
The October 2026 update states Bulk Create covers all four tools and aligns the in-app path with API batch processing. The separate REST API (v2.api.piktid.com, webhooks on Pro+) targets PIM/DAM automation rather than manual work inside Bubble pages.
What “BIAIG routes” means here
BIAIG is an editorial label in the brief—not a verified public product name. In this article it means Bubble-side AI image generation orchestration: API Connector, backend workflows (including recursive/scheduled runs), optional external bulk actors (e.g. prompt- or CSV-driven flat generation), and persisting assets after QA.
OpenAI sits in a different lane: consumer ChatGPT Shopping “Try on” (selfie + listing) is not a merchant on-model bulk pipeline; developer Images API examples fit prompt-driven flat/lifestyle/try-on-style outputs from Bubble, not On-Model’s garment-geometry batch tools (Flat to Model, Model Swap, cross-view recolor). A closer look: ChatGPT virtual try-on and shopping.
Bubble: source of truth vs scale
Source of truth: your Product or Variant in Bubble with a stable SKU or unique id.
Bulk data: Bubble Data API bulk create accepts newline-delimited JSON, up to about 1,000 Things per request, with a ~4-minute timeout. That scale is not the same as On-Model’s 100 tags/jobs per image run. I treat two batch layers—database ingest and image production—as separate problems now; I used to blur them.
| Name | What it is for | Scale / limits | What it is not for |
|---|---|---|---|
| Bubble Data API bulk create | Database ingest (source of truth) | Newline-delimited JSON; ~1,000 Things per request; ~4-minute timeout | On-Model image batch size (~100 tags/jobs per image run) |
| On-Model image run | Image production | ~100 tags/jobs per image run | Bubble Data API bulk create scale (~1,000 Things per request) |
Community pattern for large AI catalogs (Bubble forum): parse product lists with an LLM into structured JSON, drive recursive backend workflows, track a batch record for UI progress, and store generated image URLs on image/file fields.
Operational fix: external AI URLs are often short-lived. Documented practice: type API Connector responses as image and use “:saved to Bubble Storage” so PDPs do not break when vendor CDNs expire.
Decision matrix
| Scenario | Lane | Rationale |
|---|---|---|
| Fashion drop needs consistent on-model identity, multi-view flat→model, model swap, or cross-view garment recolor | On-Model (Bulk Create in-app or API + webhooks) outside Bubble UI | Fashion-specific batch tools and the ≤100 tags/jobs per run cap match apparel PDP production |
| Non-apparel catalog, text/SKU metadata, white-background or lifestyle heroes | BIAIG routes from Bubble (workflows + generative APIs) | No need for a fashion geometry pipeline when inputs are metadata or generic product shots |
| >100 SKUs or >100 colorway jobs in one drop | Split runs (chain Bulk Create) or API-only automation into Bubble backend | One run cannot satisfy both Data API bulk (1000) and On-Model image bulk (100) assumptions |
| Merchandisers need in-app progress | Bubble batch Thing + status fields; vendor jobs complete via webhook | Progress UI is your data model, not On-Model’s wizard |
| Before publish | Always | Treat vendor quality claims as marketing until validated on representative SKUs |
Minimum integration contract
- Bubble SKU/id = On-Model tag = correlation field in webhook/job metadata (public webhook schema detail was not fully captured in research—store a correlation id on your batch record).
- After QA, ingest into Bubble Storage—do not rely on vendor CDN URLs alone in image fields.
- For collections >100, document runbooks (e.g. 400 SKUs → four chained Bulk Create runs or an API job queue).
- Licensing and model release for custom identities (On-Model uploads vs brand-owned GPT edits) needs legal review beyond marketing pages.
What I take from this
- On-Model Bulk Create (100/run) fits fashion on-model PDP batch when tags are SKUs and jobs are Flat to Model, Swap, Packshot, or Recolor.
- BIAIG routes fit flat/lifestyle/generative bulk orchestrated from Bubble when catalog logic and storage stay in your app.
- Keep the Bubble product record as the single source of truth; treat image vendors as production pipes with different caps and QA.
Sources (Scout packet)
On-Model Bulk Create blog; October 2026 release; Batch Processing and API Integration; Bubble Data API bulk and Files docs; forum threads (AI catalog generator, Bubble Storage upload); OpenAI Shopping help and Images cookbook (adjacent); Apify bulk AI actor example; Bria Skills (enterprise generative—distinct from the BIAIG editorial acronym). Same topic: Fozzels October 2026: Image Flow.
Open questions
Internal definition of BIAIG; verification of digest 15.10; full On-Model webhook field map for SKU correlation; Bubble workload when combining Data API bulk with parallel AI image workflows; preferred pattern for >100 fashion SKUs—chained in-app Bulk Create vs API-only.