Check out what I made for a pitch deck - 50 images, under $100.

I've recently completed a project that required a substantial volume of high-quality, thematic imagery for a venture capital pitch deck. Given the constraints—both budgetary and stylistic—I opted to use Midjourney as the primary generation engine. The results were compelling, and I believe a methodological breakdown of the process, cost, and key learnings would be valuable for this community, particularly for those considering Midjourney for commercial or professional use-cases.

**Project Scope & Parameters:**
* **Objective:** Generate 50 unique, cohesive images depicting "the future of urban mobility."
* **Key Themes:** Autonomous vehicles, smart city infrastructure, vertical take-off and landing (VTOL) passenger drones, integrated public transit hubs.
* **Style Consistency:** A requirement for a semi-realistic, clean, slightly futuristic aesthetic across all images to maintain deck professionalism.
* **Budget Cap:** Strictly under $100 USD.
* **Tool:** Midjourney (via Discord, Standard Plan).

**Workflow & Prompt Engineering Strategy:**
Achieving consistency at scale is Midjourney's well-known challenge. I approached it with a structured prompt template, which I iteratively refined over approximately 100 preliminary generations (using Fast Hours). The final template followed this structure:

```
[subject description], [specific action/composition], [environmental context], [lighting and mood], [stylistic keywords] --ar 16:9 --style raw --s 250
```

A concrete example prompt:
```
A sleek, autonomous multi-passenger vehicle navigating a rain-slicked city street at dusk, surrounded by holographic navigation displays, cinematic lighting, neon reflections, futuristic --ar 16:9 --style raw --s 250
```

The `--style raw` parameter and a high `--s` (stylize) value of 250 were crucial. `--style raw` gives the model less opinionated, more literal interpretations, which aided in keeping the futuristic elements grounded. The 16:9 aspect ratio was a deck requirement.

**Cost & Time Analysis:**
* **Plan:** The Standard Plan ($30/month for 15 Fast Hours). I purchased one month.
* **Fast Hours Used:** The initial prompt development and testing consumed ~3.5 Fast Hours. The final 50 images, requiring multiple variations (`--v 5.2`) and occasional upscales of select grids, consumed ~4.1 Fast Hours.
* **Relaxed Mode:** After exhausting Fast Hours (7.6 total used), I generated the remaining batch variations using Relaxed Mode. This added time but no incremental cost.
* **Total Direct Cost:** $30 for the monthly subscription. Total project time was approximately 14 hours of active work, spread over three days (much of it waiting for Relaxed Mode generations).
* **Total Images Generated:** To get 50 final selects, I initiated approximately 380 jobs (grids of 4 images). The selection rate was thus around 13%.

**Key Findings & Pitfalls:**
* **Consistency via Seed & Parameters:** Using a fixed seed value (`--seed 12345`) was less effective than expected for character/vehicle consistency across scenes. I found greater success by standardizing the *stylistic tail* of the prompt (e.g., `cinematic lighting, neon reflections, futuristic, clean design, Sony FE 24mm f/1.4 GM`) and the parameters (`--style raw --s 250`).
* **Cost Control:** The major cost is time, not GPU minutes, if you plan to use Relaxed Mode. For non-rushed projects, the Standard Plan is sufficient. The critical move was doing intensive prompt crafting in a focused Fast Hours session.
* **Iterative Refinement is Non-Negotiable:** The first 20 prompts yielded only 2 usable images. Systematic adjustment of keywords (e.g., swapping "futuristic" for "neo-futurism" or specifying camera lens details) dramatically improved the hit rate.
* **Post-Processing Necessity:** Approximately 40% of the selected images required minor post-processing in Affinity Photo:
* Correcting minor anatomical or perspective oddities (e.g., weird wheel spokes).
* Adjusting color temperature to unify the palette across images.
* Removing an occasional unwanted artifact.

**Conclusion:**
This project demonstrates that Midjourney is a viable tool for producing commercial-grade assets at a very low cost-per-image (~$0.60 per final image, including subscription). However, the hidden cost is the significant investment in prompt engineering expertise and iterative refinement. The `--style raw` parameter was the single most important discovery for maintaining a professional, less "whimsical" aesthetic. For anyone attempting a similar batch project, I recommend allocating at least 50% of your timeline to the initial prompt development and testing phase. The payoff is a highly efficient generation phase where the majority of outputs are usable.