flux-1.1-pro

`flux-1.1-pro` is Black Forest Labs' FLUX 1.1 Pro — a fast, production-grade text-to-image model predating the FLUX.2 line but still widely used for reliable commercial generation (docs.bfl.ai). On CallMissed, use `"model": "flux-1.1-pro"` on `/v1/images/generations`. The id matches BFL's `flux-pro-1.1` endpoint naming on their API and Azure's `FLUX-1.1-pro` Foundry listing.

BFL pricing lists FLUX 1.1 Pro at about four credits ($0.04) per image on the direct API (docs.bfl.ai/quick_start/pricing). CallMissed charges $0.05 per image — competitive for high-volume ad creative and catalog photography. Azure documents parameters such as width, height, seed, safety_tolerance, output_format (PNG/JPG), and prompt_unsampling on Foundry deployments (learn.microsoft.com/azure/ai-foundry/foundry-models/concepts/models-sold-directly-by-azure).

Strengths: excellent prompt adherence for its generation, fast turnaround relative to heavier diffusion stacks, and predictable behavior for ecommerce templates ("product on white background", "lifestyle scene with mug"). Teams use it when FLUX.2 Pro quality is unnecessary but Cloudflare `flux-2-dev` is too experimental.

Compare to `flux-2-pro` when you need maximum detail, multi-reference editing, or the newest FLUX.2 capabilities. Compare to `flux-2-klein-9b` when you need the absolute lowest cost per image on CallMissed. Compare to `nano-banana-2` when you want Gemini-native image generation instead of BFL diffusion.

Integration: same OpenAI-compatible image endpoint as other CallMissed models; set explicit sizes when supported, iterate seeds for variant exploration, and log prompts for brand compliance. Batch overnight jobs for marketing asset packs.

Limitations: below FLUX.2 Pro on hardest compositional prompts, no magic fix for human hands/faces (iterate or inpaint externally), and safety_tolerance settings may blank some prompts — adjust creatively rather than fighting filters. Always disclose AI-generated imagery per your jurisdiction's marketing rules.

Production catalog workflow: ecommerce teams generate hundreds of SKUs — script prompts from CSV (`product`, `background color`, `angle`), call `flux-1.1-pro` with deterministic seeds for reproducibility, QA sample 5% manually.

Foundry parameters: Azure exposes width/height, seed, safety_tolerance, output_format — when CallMissed exposes size in OpenAI-compatible shape, map your aspect ratio needs accordingly.

Speed vs FLUX.2: 1.1 Pro often returns faster — useful for interactive ad builders where users iterate prompts live.

Cost at 10k images/month: 10k × $0.05 ≈ $500 — model in finance before promising unlimited gen tiers to customers.

Legal: trademarked logos in prompts may create risky outputs — block prompt terms server-side.

When to retire 1.1 Pro internally: if FLUX.2 Pro wins all blind evals and price delta is acceptable, migrate; keep 1.1 for bulk drafts only.

Integration testing: assert HTTP 200 and non-empty image payload in CI; FLUX outages should trigger fallback to `flux-2-dev` or `nano-banana-2` if your product allows.

Historical placement: FLUX 1.1 Pro predates FLUX.2 but remains in production for teams with tuned prompt libraries — migrating thousands of prompts to FLUX.2 requires re-validation; 1.1 Pro avoids regression risk during transition.

BFL API naming: direct BFL endpoint `flux-pro-1.1` corresponds to customer id `flux-1.1-pro` on CallMissed — document mapping for teams using both BFL native and CallMissed unified billing.

Parameter tuning: Azure Foundry documents `prompt_upsampling` and `safety_tolerance` — higher tolerance may allow edgier fashion photography prompts; lower for family-friendly apps.

Marketplace sellers: generate consistent white-background SKU images with seeded prompts — marketplace algorithms favor uniform catalog aesthetics.

Print workflow: 1024² suffices for web; print may need external upscaler (Topaz, etc.) — budget upscaler licensing separately from API costs.

Content moderation: log prompts for UGC apps; block disallowed terms before they hit BFL filters — reduces user frustration from empty responses.

Seasonal load: Black Friday campaigns spike image gen — pre-warm worker pools and set concurrency limits so one tenant cannot exhaust shared capacity on shared hosting (implement your own queues).

End-of-life planning: when FLUX.2 fully replaces 1.1 in your evals, migrate prompts incrementally by category — do not big-bang switch without A/B revenue impact analysis.

Design system integration: store prompt templates in git (`prompts/flux-1-1-pro/product-hero.txt`) — designers PR prompt changes like code, enabling reproducible brand campaigns. Track cost per accepted image in analytics so finance can compare FLUX 1.1 Pro against FLUX.2 Pro uplift before committing to migration. Merchandising teams should A/B test conversion rate on FLUX 1.1 vs photographer shots quarterly — revenue impact beats subjective aesthetic debates. Prompt engineers should log negative prompts that reduce artifact hands and duplicate limbs; FLUX 1.1 responds well to explicit "single subject, centered" framing for catalog shots. Affiliate marketers generating ad creatives at scale should rotate seeds daily and archive metadata so winning creatives can be reproduced for compliance audits months later.