A route-first Claude Code setup guide for OpenRouter and DeepSeek: env vars, model names, /status, provider proof, first-error fixes, and free/unlimited stop rules.
Sora web/app has ended and the API has a scheduled removal date. Use this route board to choose Veo 3.1, Runway, Kling, Seedance, Pika, Luma, PixVerse, HeyGen, or a provider route without mixing free trials and official APIs.
The cheapest LLM API model depends on input tokens, output tokens, cache hits, Batch/Flex terms, retries, and whether the output passes your workload.
A dated contract map for free AI API tiers: recurring free routes, starter credits, cheap paid APIs, gateway free variants, and image API testing lanes without mixing provider and platform claims.
Sora app/web is gone, the API still has per-second prices until its scheduled shutdown, and ChatGPT Pro is not a Sora-only buy.
A current Grok API cost worksheet with official xAI rows, free-tier caveats, and realistic examples for chat, documents, coding, and research agents.
If you already have the image, pick the route before the brand. Use Runway for the safest creator default, Adobe Firefly when brand-safe editing matters, Gemini Omni or Veo 3.1 for Google's paid app and API routes, Kling for multi-shot or element control, and Pika or Vidu for quick effects tests. Treat Sora as a sunset caveat, not a normal current recommendation.
DeepSeek V4 Flash is the lowest verified official paid token floor checked here, but the cheapest practical LLM API provider depends on route owner, quality threshold, retries, latency, and gateway fees.
A Nano Banana API key is a Gemini API key managed through Google AI Studio. Create the key there, but keep it separate from billing, project limits, free-tier claims, and provider gateway keys before you call Nano Banana 2, Lite, or Pro.
The current official Nano Banana image API has no free image-output tier for Lite, Nano Banana 2, or Pro. Lite is the lowest official 1K lane, while laozhang.ai's $0.025 Gemini Flash / Nano Banana route is a separate provider contract, not a Google price row.
Treat Plus-to-Business as a workspace data migration. Inventory chats, memories, GPTs, Projects, and files before merging, then verify access immediately after.
Scheduled Tasks work best as scheduled prompts with verified ChatGPT tools/apps: put critical facts in the task prompt, treat Memory as conditional, and use another route for files, webhooks, or deterministic jobs.
You can create a Gemini API key for free in AI Studio, but that key is a credential, not a separate quota bucket. Check pricing for model eligibility and AI Studio for the live project limit before trusting 20 RPD or unlimited-key claims.
Start new Google image work on Nano Banana 2, escalate to Nano Banana Pro only when the job requires it, and keep Imagen 4 only as a temporary legacy route where your surface still allows it.
A practical Nano Banana prompt workbook with route cues, a test card, and 50 templates for product shots, ads, edits, characters, text visuals, and troubleshooting.
Build LinkedIn, resume, team bio, founder, creative, and speaker portraits with Nano Banana prompt cards that keep identity, consent, and professional fit in check.
Choose the logo job, fill the brand brief, paste a template, test 3-5 outputs with the same prompt, and repair only the weak dimension before handoff.
A practical route guide for making album cover art that still reads, exports cleanly, and survives basic release checks.
Pick the right coat of arms maker for artwork, fantasy, AI concepts, or heraldry practice, then verify rights before public, family, or official use.
A practical guide to drawing tarot cards online, choosing spreads and reversals, using AI safely, and keeping custom tarot art separate from readings.
A route-first checklist for unexpected ChatGPT charges: identify the billing owner, gather proof, redact payment data, cancel renewal separately, and request a refund through the right channel.
A branch-first Claude KYC guide that explains whether you are seeing phone verification, government ID, payment verification, finance KYC agent material, or a location, policy, age, or unsafe proxy-verification issue.
Use the route owner, response headers, credential path, and same-route verification to recover from Claude rate-limit errors without making the signal worse.
Claude Fable 5 is a wait-and-recheck branch while GLM-5.2 is a documented Z.AI Coding Plan route to test. Start with current access, then run one same-task trial before changing a default model.