Ideogram Batch (500 rows) vs Batch Banana (1,000-row CSV): external batch hosts before Bubble's API Connector and BIAIG

Why this matters now

I keep running into the same snag: prompt catalog in a spreadsheet, then an external batch host row limit before anything reaches Bubble. Public docs give two common ceilings: Ideogram Batch Generation — up to 500 prompt rows (plus header), and Batch Banana Batch — up to 1,000 rows in CSV/TSV. This is not a “winner” story. It is a ceiling choice before the same next steps: normalize → Bubble Things → (if needed) API Connector for outbound HTTP. A closer look: The ~500-row bulk-tools ceiling in.

Limits and formats (Scout-verified docs only)

	Ideogram Batch	Batch Banana Batch
Prompt rows	up to 500 + header	up to 1,000
File	up to 1 MB; .csv / .xls / .xlsx / .ods	CSV/TSV up to 10 MB
Plan gating	Pro / Team / Enterprise	CSV bulk tied to Max on public pricing (~$100/mo on fetched pricing page)
Output	ZIP from Batch UI	Run All, per-run result columns, CSV export, bulk download
API surface	developer.ideogram.ai OpenAPI does not describe a batch-upload equivalent; Batch is a separate UI path	Multi-model via CSV columns (model, aspect ratio, images per prompt 1–8)

Ideogram: one row = one generation; num_images 1–4 per row — total images = sum of num_images, not row count. Batch does not support partner models (docs mention GPT Image, Nano Banana, etc.).

Batch Banana: required prompt column; optional model IDs and generation parameters in the sheet.

Practical inference (not vendor-documented): a 501–1,000-row catalog may fit one Batch Banana batch but needs ≥2 Ideogram batch files or pre-chunking in Sheets at 500.

Bubble handoff: three layers

1. External generation — Ideogram ZIP or Batch Banana CSV plus assets.
2. Normalization — Sheets/Excel: prompt, model metadata, image URL or file reference, external id, status. Chunk here to match your host (500 vs 1,000) if generation is still ahead of you.
3. Bubble ingress (manual.bubble.io + forum patterns):
   • Data → Import CSV — one data type per import, field mapping in the editor; good for catalog Things without code.
   • Data API bulk create — POST .../obj/{typename}/bulk, newline-delimited JSON, hard limit of 1,000 records per request, 4-minute timeout; requires Create via API privacy rules. Docs caution about capacity — test smaller chunks in production.
   • API Connector — generic outbound HTTP; common forum pattern: File Uploader → file URL → Connector call (CSV parsing often needs a plugin or external step). Same topic: Fozzels October 2026: Image Flow.

Batch Banana’s 1,000 rows are the same order of magnitude as Bubble bulk create’s 1,000, but neither vendor documents that pairing — do not assume an official integration. Same topic: On-Model Bulk Create (100 products.

Pre-chunk in Sheets vs mega-job inside Bubble

Pre-chunk in Sheets (500 / 1,000) when:

• you are still in generation and chose a lower ceiling (Ideogram 500);
• you want human QA per batch file before ZIP / Run All;
• you want one bulk create aligned with one external batch for simple auditing.

Situation	Why pre-chunk in Sheets (500 / 1,000)
Still in generation with a lower ceiling (e.g. Ideogram 500)	Matches a smaller batch size while generating
Human QA per batch file before ZIP / Run All	Review each batch file before downstream steps
One bulk create aligned with one external batch	Keeps auditing simple

Orchestrate inside Bubble (backend “Schedule API Workflow on a list”, batch container Thing — forum pattern) when:

• rows exceed 1,000 or logic is per-row (retries, rate limits, different APIs per row);
• images already have URLs and workflows only update Things via API Connector;
• you should not depend on a third-party batch UI for the whole pipeline.

BIAIG: the editorial brief names in-Bubble orchestration context; Scout found no public product spec (biaig.com shows “Launching Soon”). I treat BIAIG as an editorial/internal label for the in-app mega-job pattern — no parity claims to an undefined release. If your team uses BIAIG for a plugin or internal module, document your own definition separately.

Suggested workflow (no hype)

1. Pick a host by row ceiling and model scope (Ideogram Batch excludes partner models; Batch Banana emphasizes multi-model CSV with plan gating).
2. Export: Ideogram ZIP → map assets to URLs in a sheet; Batch Banana CSV export template → merge with prompts.
3. One “source of truth” table mapped to Bubble fields.
4. Ingress: ≤1,000 new Things — bulk API or CSV import; >1,000 — multiple bulk requests or split imports; live per-row image APIs — scheduled backend workflows, not one oversized bulk.
5. Before production: smaller chunks, watch the 4-minute timeout and app capacity (Bubble docs).

Risks (straightforward)

• Row limits are product caps, not proof of perfect Bubble alignment.
• API Connector / bulk handoff steps are inferred from Bubble docs and forum posts — not Ideogram or Batch Banana integration guides.
• Where to store ZIP assets (Bubble file fields vs CDN) remains an open question — validate field limits in your app.
• Plans and pricing change — verify before you ship.

Sources (Scout packet)

Ideogram Batch docs; Batch Banana batch/features/pricing; Bubble Data API bulk, CSV import, API Connector; Bubble Forum (CSV upload to API; parallel backend workflows).