Freelance portfolio and case study generation from past project deliverables

Introduction Building a professional portfolio as a freelancer means doing two jobs: delivering client work and then converting that work into polished case studies.

Most freelancers either skip this step entirely, leaving money on the table, or spend hours manually recreating project narratives across their website, LinkedIn, Behance, and email pitches. You photograph the deliverable, write the brief, record yourself explaining it, edit the video, and repeat this for every past project. It's tedious work that prevents you from taking new clients. The real frustration comes from the fact that you already have everything you need. The finished design files exist. The project details sit in your notes or emails. The only missing piece is a systematic way to transform raw deliverables into formatted case studies without touching them manually each time. This workflow automates that entire process. Starting from a single project summary and original file, it generates polished case study visuals, produces a video featuring a virtual presenter walking through your work, and writes persuasive copy tailored to different platforms. One input; multiple formatted outputs ready to post.

The Automated Workflow We'll use Make as the orchestration tool here because it handles file uploads cleanly and integrates smoothly with image and video generation APIs.

The workflow runs in five stages: capture the project brief, generate hero imagery, create a video walkthrough, write platform-specific copy, and deliver everything organised in a folder structure. Setting up the initial trigger The workflow starts when you submit a form with your project details. You'll use Make's HTTP module to receive the submission, which should include the project title, a brief description, your original design file (as an attachment or shareable link), and target platforms (website, LinkedIn, Twitter, Behance).

CopyPOST /api/projects
Content-Type: application/json { "project_title": "E-commerce Rebrand for LocalGoods", "project_brief": "Complete visual identity refresh including logo, colour palette, and website template redesign. Client saw 34% increase in conversion after launch.", "deliverable_url": "https://storage.example.com/localgoods-final-designs.zip", "target_platforms": ["website", "linkedin", "behance"], "freelancer_name": "Sarah Chen", "freelancer_role": "Brand Designer"
}

Make receives this data and passes it downstream. Store the deliverable URL and project details in a Make variable for later reference. Generating hero imagery with Adobe and Nsketch Next, the workflow needs to create a visually striking image representing your project. This is where you combine Adobe Photoshop AI and Nsketch AI. First, send the project brief to Claude Opus 4.6 to generate three image prompts tailored to the industry and style. Claude understands context better than simpler models and can craft descriptions that will produce cohesive, professional results.

CopyPOST /api/messages
Host: api.anthropic.com { "model": "claude-opus-4.6", "max_tokens": 500, "messages": [ { "role": "user", "content": "I completed a rebrand project for an e-commerce client called LocalGoods. The work included logo design, colour palette development, and website template design. Generate 3 distinct image prompts (100 words each) that showcase this design work in a professional, portfolio-ready style. Each prompt should be suitable for a case study hero image." } ]
}

Claude returns three prompts. Parse the response and feed the first prompt to Nsketch AI, which specialises in rapid generation with template-based workflows. Nsketch can generate the image in under 30 seconds.

CopyPOST /api/generate
Host: api.nsketch.ai/v1 { "prompt": "[Claude's first generated prompt]", "template": "portfolio_hero", "style": "professional", "dimensions": "1200x630"
}

Save the generated image. In parallel, you could optionally send the project file itself to Adobe Photoshop AI if you want to enhance or composite your original deliverable with branded elements. For most freelancers, the Nsketch output is sufficient and faster. Creating video with Hour One The video component is where this workflow truly saves time. Hour One accepts a text script and produces a video with a virtual presenter. You'll generate the script automatically using GPT-4o, then feed it to Hour One. Send the project details to GPT-4o to write a 90-second case study script. This model is fast and cost-effective for straightforward copywriting tasks.

CopyPOST /v1/chat/completions
Host: api.openai.com { "model": "gpt-4o", "messages": [ { "role": "user", "content": "Write a 90-second case study script for a freelancer to narrate. Project: E-commerce rebrand for LocalGoods. Key achievements: 34% conversion increase, new logo, colour palette, website template. Tone: professional and conversational. Include brief problem statement, solution overview, and results. Format as a script with speaker directions in brackets." } ], "max_tokens": 300
}

Take the script and send it to Hour One's API. Specify a virtual presenter avatar and a professional studio background.

CopyPOST /api/videos
Host: api.hourone.com { "script": "[GPT-4o generated script]", "avatar": "professional_presenter_female", "background": "studio_blue", "duration": "90s", "language": "en", "output_format": "mp4"
}

Hour One returns a video URL once processing completes (typically 2-5 minutes). Download and store this in your project folder. Writing platform-specific copy Different platforms need different copy lengths and tones. Use Claude Haiku 4.5 here because it's fast, cheaper than larger models, and completely adequate for platform-specific copy variations.

CopyPOST /api/messages
Host: api.anthropic.com { "model": "claude-haiku-4.5", "max_tokens": 1000, "messages": [ { "role": "user", "content": "Generate three versions of case study copy for these platforms: 1) Website (150 words, detailed narrative), 2) LinkedIn (50 words, engaging hook), 3) Behance (75 words, visual focus). Project: E-commerce rebrand for LocalGoods. Results: 34% conversion increase. Deliverables: logo, colour system, website template. Format as JSON with keys: website, linkedin, behance." } ]
}

Claude Haiku returns all three versions instantly. Parse the JSON and store each in a Make variable. Organizing outputs The final step assembles everything into a clean folder structure and generates an index document. Use Make's file handling to create a folder named after the project, then upload all assets: hero image, video file, a text document containing all copy variations, and metadata.

Copy/case-studies/LocalGoods-Rebrand/
├── hero-image.png
├── case-study-video.mp4
├── copy.txt (contains all platform versions)
├── metadata.json (project details, links, dates)
└── README.md (quick reference guide)

Within the README, include links to each generated asset and instructions on where to post each version.

The Manual Alternative If you prefer hands-on control or want to customise outputs beyond what the workflow provides, stop at any stage and review before the next step.

Make allows you to insert approval steps between modules. After Claude generates the image prompts, you could manually review them and select one, then continue the automation. After Hour One produces the video, you can review it before distribution. Alternatively, download the generated assets and use them as starting points. The hero image can go into Adobe Photoshop AI for further refinement. The video script can be rewritten before sending to Hour One. The copy can be edited for brand voice. The automation provides a rough draft; your judgement makes it exceptional.

Pro Tips Rate limiting and cost management Claude and GPT models have rate limits.

If you're automating this for multiple projects in quick succession, stagger requests using Make's delay modules. Add a 2-second pause between API calls to avoid hitting limits. Nsketch and Hour One are slower anyway, so they naturally create breathing room. Error handling for file uploads Always validate that your deliverable file actually exists before the workflow tries to process it. Add a Make HTTP module that checks the URL with a HEAD request. If it returns a 404, the workflow stops and sends you a notification rather than failing silently.

CopyGET /path/to/file
Head: true

Testing with a staging project Run the entire workflow once on a past project where you already know the desired output. This lets you catch configuration errors before automating new projects. Use Make's debug mode to inspect API responses at each step. Video generation takes time Hour One and Nsketch aren't instant. Budget 5-10 minutes for video rendering. Don't expect real-time results. If speed is critical, consider using Runway Gen-3 Alpha instead, which is faster though less presenter-focused. Cost per case study Calculate the exact expense before scaling. Claude Haiku costs roughly £0.02 per case study. GPT-4o costs £0.03 to £0.05 depending on prompt length. Nsketch and Hour One are flat-rate subscriptions. If you're generating five case studies per month, the automation pays for itself by recovering 10-15 hours of manual work.

Cost Breakdown