From zero to a working 5 to 10 minute personalized audio in 3 to 4 focused days
Goal: produce one working audio track based on Jeff's own birth data and Gene Keys sequence. The track should be magical enough to make someone who listens to it on a phone go "holy shit, this was made for me." This single audio becomes the proof point for Sunday with Josh, Thursday with Colin, and Omaha conversations.
Constraint: manual build, no agent automation yet. This is a one-off proof of concept using ElevenLabs and Suno directly. The agent architecture comes later. Right now, you are the agent.
Timeline: 3 to 4 days of focused work, ideally 4 to 6 hours per day. Doable between now and Sunday if you start tomorrow.
Total monthly cost if you sign up for everything fresh: $52 plus the Claude Pro you already have. Cancellable after the demo if you want to pause subscriptions until the funded build phase.
Create a new file called jeff_chart_data.md in the AFFIRMOLOGY folder. Include:
You said you have most of this somewhere already. Just consolidate it into one file so Claude can ingest it cleanly.
Open a fresh Claude conversation. Paste the data file. Then use this prompt:
Claude will produce a script. Read it. Adjust as needed (you know yourself best). Save the final version as jeff_demo_script_v1.md in the AFFIRMOLOGY folder.
Sign up at elevenlabs.io. Subscribe to Creator tier ($22/mo, 100K characters). Go to Voice Library.
Recommended pre-trained voices for spiritual/affirmation content:
Generate a 30-second test with each candidate voice using your first paragraph. Listen back. Pick the one that lands.
In ElevenLabs, paste your script in chunks (paragraph by paragraph works best for control). Adjust the Stability slider to around 50 (balanced) and the Style slider to around 30 (some emotion but not too dramatic).
For pause indicators in the script, use SSML break tags:
<break time="3s" />
Or just paste two periods and a space ".. " which ElevenLabs interprets as a natural pause.
Download each chunk as MP3. Stitch them together later, or generate the whole script at once if it fits in one request.
Sign up at suno.com. Subscribe to Pro ($10/mo for commercial license).
Use these prompt variations to generate 4 to 6 music options:
Pick your favorite, download as WAV or high-quality MP3.
ffmpeg -i voice.mp3 -i music.mp3 \ -filter_complex \ "[0:a]volume=1.0[a0]; \ [1:a]volume=0.20[a1]; \ [a0][a1]amix=inputs=2:duration=longest" \ -c:a libmp3lame -b:a 192k \ affirmology_demo_v1.mp3
Wait until hour 2 or 3 of the meeting. Once you have shown him the deck and the term sheet and he is engaged: "I built a demo of the product. I want you to listen to 5 minutes of it on headphones right now." Hand him your phone with the audio queued up. Watch his face. Do not talk. Let the audio do the work.
Tell him over Zoom: "I built a working demo of the product using my own data. Want me to send you the file to listen to before our next conversation?" Send it after the call. Follow up with a question: "What did you think when you listened?"
Same approach as Josh. Phone with headphones. Personal data. Live listening. Reaction.
If the demo is good enough by September, set up a booth with 4 to 6 headphone sets. Each visitor enters their birth data on a tablet, hits "generate sample," gets a 90-second personalized clip queued up automatically. This is the activation flow. The 5-minute Sunday demo is the proof that the activation flow is possible.
Once the manual demo works and gets reactions, the agent automation work becomes meaningful instead of speculative. You have proven the product end to end. The next phase (post-funding) is wrapping the manual steps in agent orchestration:
This is the architecture Claude Code can help you build over a 30 to 60 day sprint once the round closes. The manual demo proves the pieces work. The agents just make it scalable.