jackwener/OpenCLI · error · ArgumentError

Estimated GPU cost ${estimatedMinutes} minutes exceeds --max

Error message

Estimated GPU cost ${estimatedMinutes} minutes exceeds --max-minutes ${max}

What it means

assertBudget enforces a user-supplied ceiling (--max-minutes, default 2) on the estimated GPU minutes for a generation. This ArgumentError is thrown when estimatedMinutes (per-job cost x repeat, adjusted for turbo/hd/omni/draft/relax) exceeds that ceiling, so a job that would be unexpectedly expensive is blocked before submission.

Source

Thrown at clis/midjourney/capabilities.js:279

  if (speed === 'relax') perJobMinutes = 0;
  const estimatedMinutes = Number((perJobMinutes * repeat).toFixed(2));

  return {
    requestedModel,
    effectiveModel: selectedModel,
    routingReason,
    resolution,
    speed,
    repeat,
    estimatedMinutes,
  };
}

export function assertBudget(account, estimatedMinutes, maxMinutes, reserveMinutes, creditsToMinutes) {
  const max = normalizeNonNegativeNumber(maxMinutes, 2, '--max-minutes');
  const reserve = normalizeNonNegativeNumber(reserveMinutes, 0, '--reserve-minutes');
  if (estimatedMinutes > max) {
    throw new ArgumentError(`Estimated GPU cost ${estimatedMinutes} minutes exceeds --max-minutes ${max}`);
  }
  const remainingCredits = Number(account?.total_credits ?? account?.credits_total);
  const remainingMinutes = creditsToMinutes(remainingCredits);
  if (Number.isFinite(remainingMinutes) && remainingMinutes - estimatedMinutes < reserve) {
    throw new CommandExecutionError(
      `Midjourney budget guard would leave ${Number((remainingMinutes - estimatedMinutes).toFixed(2))} minutes, below the ${reserve}-minute reserve`,
    );
  }
  return { maxMinutes: max, reserveMinutes: reserve, remainingMinutes };
}

export function estimateAction(
  operation,
  { videoResolution = 'sd', batchSize = 1, sourceIsVideo = false, sourceUsesOmni = false } = {},
) {
  if (!ACTION_CHOICES.includes(operation)) {
    throw new ArgumentError(`operation must be one of: ${ACTION_CHOICES.join(', ')}`);
  }

View on GitHub (pinned to 49907e53dc)

Solutions

Raise --max-minutes to at least the estimated cost shown in the error (e.g. --max-minutes 8).
Reduce repeat count, or drop --hd/--turbo/--oref to lower the per-job GPU cost.
Use relax speed (0 GPU minutes) if latency is acceptable and the estimate must fit a tiny budget.
Call the plan/estimate API first and set --max-minutes from estimatedMinutes programmatically.

Example fix

// before
midjourney generate 'a cat --hd' --repeat 4  // default --max-minutes 2
// after
midjourney generate 'a cat --hd' --repeat 4 --max-minutes 8

Defensive patterns

Strategy: validation

Validate before calling

const planResult = await midjourney.plan({ prompt, repeat, resolution, speed });
const maxMinutes = Number(process.env.MAX_MINUTES ?? 2);
if (planResult.estimatedMinutes > maxMinutes) {
  throw new Error(`estimate ${planResult.estimatedMinutes} min exceeds budget ${maxMinutes}; raise --max-minutes or reduce repeat`);
}

Type guard

function fitsBudget(planResult, maxMinutes) {
  return typeof planResult?.estimatedMinutes === 'number' && planResult.estimatedMinutes <= maxMinutes;
}

Try / catch

try {
  return await generate(args);
} catch (err) {
  if (err instanceof ArgumentError && err.message.startsWith('Estimated GPU cost')) {
    const est = Number(err.message.match(/cost ([\d.]+)/)[1]);
    const max = Number(err.message.match(/--max-minutes ([\d.]+)/)[1]);
    return generate({ ...args, maxMinutes: Math.ceil(est) });
  }
  throw err;
}

Prevention

Call the plan/estimate API before generating and set --max-minutes from the estimate.
Remember turbo doubles, hd/omni raise per-job cost; re-budget when changing these flags.
Don't leave --max-minutes at the default 2 when running repeat batches.
Set --max-minutes explicitly in CI/scripts rather than relying on defaults.

When it happens

Trigger: Calling the generate flow with estimatedMinutes > maxMinutes: e.g. --repeat 4 with hd resolution (or turbo, or omni reference) while --max-minutes is left at the default 2, or passing a non-negative but too-small --max-minutes.

Common situations: Forgetting to raise --max-minutes when scaling up repeat or switching to HD/Turbo/Omni; keeping the default 2-minute budget after moving from relax to fast jobs; batch scripts with fixed budgets hitting costlier models.

Related errors

AI-assisted analysis of jackwener/OpenCLI@49907e53dc (2026-08-29). Data as JSON: /api/errors/a286724e11bd84a0. Report an issue: GitHub.