Imported from jeremylongshore/tons-of-skills-marketplace (skills/.curated/ideogram-cost-tuning/SKILL.md). Install upstream with npx skills add jeremylongshore/tons-of-skills-marketplace --skill ideogram-cost-tuning. Copyright stays with the author (MIT).

Ideogram Cost Tuning

Overview

Minimize Ideogram API spending by selecting the right model per task, caching identical prompts, batching images per call, and tracking credit burn rate. Ideogram bills per image generated at a flat rate that varies by model and rendering speed.

Pricing Reference

Prices approximate; check ideogram.ai/features/api-pricing for current rates.

Instructions

Step 1: Two-Phase Generation Workflow

// Draft with TURBO (cheap), finalize with V_2 (quality)
async function costEfficientGeneration(prompt: string, iterations = 5) {
  // Phase 1: Generate drafts cheaply
  const drafts = [];
  for (let i = 0; i < iterations; i++) {
    const result = await generateImage(prompt, { model: "V_2_TURBO" });
    drafts.push(result);
  }
  // Cost: 5 x $0.05 = $0.25

  // Phase 2: Pick best seed, regenerate at full quality
  const bestSeed = await selectBestDraft(drafts); // manual or automated
  const final = await generateImage(prompt, { model: "V_2", seed: bestSeed });
  // Cost: 1 x $0.08 = $0.08

  // Total: $0.33 instead of $0.40 (5 x V_2)
  return final;
}

Step 2: Batch Images Per Call

// Single API call for up to 4 images costs the same as 4 separate calls
// BUT saves latency (one round-trip instead of four)
async function generateVariations(prompt: string) {
  const response = await fetch("https://api.ideogram.ai/generate", {
    method: "POST",
    headers: {
      "Api-Key": process.env.IDEOGRAM_API_KEY!,
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      image_request: {
        prompt,
        model: "V_2_TURBO",
        num_images: 4, // 4 images in one call
        magic_prompt_option: "AUTO",
      },
    }),
  });

  const result = await response.json();
  return result.data; // 4 image objects
}

Step 3: Cache Identical Prompts

import { createHash } from "crypto";

const cache = new Map<string, { url: string; seed: number; cachedAt: number }>();
const CACHE_TTL_MS = 7 * 24 * 60 * 60 * 1000; // 7 days

function promptKey(prompt: string, style: string, model: string): string {
  return createHash("md5").update(`${prompt}:${style}:${model}`).digest("hex");
}

async function cachedGeneration(prompt: string, style = "AUTO", model = "V_2") {
  const key = promptKey(prompt, style, model);
  const cached = cache.get(key);

  if (cached && Date.now() - cached.cachedAt < CACHE_TTL_MS) {
    console.log("Cache hit -- saved one generation credit");
    return cached;
  }

  const result = await generateImage(prompt, { style_type: style, model });
  // Download and store locally before caching (URLs expire)
  const localPath = await downloadImage(result.data[0].url);
  cache.set(key, {
    url: localPath,
    seed: result.data[0].seed,
    cachedAt: Date.now(),
  });

  return cache.get(key);
}

Step 4: Budget Tracking

interface CostTracker {
  totalImages: number;
  totalCostUSD: number;
  byModel: Record<string, { count: number; cost: number }>;
  dailyBudgetUSD: number;
}

const tracker: CostTracker = {
  totalImages: 0,
  totalCostUSD: 0,
  byModel: {},
  dailyBudgetUSD: 10, // $10/day cap
};

const MODEL_COSTS: Record<string, number> = {
  V_2_TURBO: 0.05,
  V_2: 0.08,
  V_2A: 0.04,
  V_2A_TURBO: 0.025,
};

function trackGeneration(model: string, numImages: number) {
  const costPerImage = MODEL_COSTS[model] ?? 0.08;
  const cost = costPerImage * numImages;

  tracker.totalImages += numImages;
  tracker.totalCostUSD += cost;

  if (!tracker.byModel[model]) tracker.byModel[model] = { count: 0, cost: 0 };
  tracker.byModel[model].count += numImages;
  tracker.byModel[model].cost += cost;

  // Budget alert
  if (tracker.totalCostUSD > tracker.dailyBudgetUSD * 0.8) {
    console.warn(`Budget warning: $${tracker.totalCostUSD.toFixed(2)} of $${tracker.dailyBudgetUSD}/day`);
  }
  if (tracker.totalCostUSD > tracker.dailyBudgetUSD) {
    throw new Error(`Daily budget exceeded: $${tracker.totalCostUSD.toFixed(2)}`);
  }
}

function costReport() {
  console.log("=== Ideogram Cost Report ===");
  console.log(`Total images: ${tracker.totalImages}`);
  console.log(`Total cost: $${tracker.totalCostUSD.toFixed(2)}`);
  for (const [model, data] of Object.entries(tracker.byModel)) {
    console.log(`  ${model}: ${data.count} images, $${data.cost.toFixed(2)}`);
  }
}

Step 5: Billing Auto Top-Up Configuration

Ideogram Dashboard > Settings > API Beta > Billing:

Recommended settings:
  Top-up Balance: $20.00 (default)
  Minimum Threshold: $10.00 (default)

Conservative (small projects):
  Top-up Balance: $10.00
  Minimum Threshold: $5.00

Enterprise:
  Contact partnership@ideogram.ai for volume pricing
  1M+ images/month for custom rates

Cost Optimization Checklist

• Use V_2_TURBO for iteration, V_2 for final assets only
• Cache identical prompts (7-day TTL)
• Batch with num_images: 4 where possible
• Track daily spend with budget alerts
• Use V3 FLASH for UI previews and thumbnails
• Download images immediately (regeneration = double cost)
• Set conservative auto top-up limits

Error Handling

Output

• Two-phase generation workflow (draft then finalize)
• Prompt-based cache preventing duplicate charges
• Budget tracker with daily spending alerts
• Cost report by model version

Prerequisites

• Approved budget, baseline generation/quota metrics, synthetic fixture, and rollback revision for concurrency, cache, and destination policy.

Examples

integration=sandbox-generator; baseline=12000; deferred=120; deduped=900; rights=test-owned; output_retention=none; rollback=cost-r18 is a safe cost decision.

Resources

• Ideogram API Pricing
• API Billing Setup

Next Steps

For architecture patterns, see ideogram-reference-architecture.