WHAT THIS IS
This is an audio construction research project.. I’m studying how AI-generated and human-performed elements combine acoustically when they’re built into a finished song. I’m not judging creative quality, performance polish, lyrics, or whether it’s a hit. Creative direction is entirely yours. What I measure is the construction itself: how each layer sounds on its own, how they combine, and how mix and master processing changes the signal.
I ran a first round of this in May–June with producers who each delivered 4–6 songs and it went great. This round pays more, opens up to all genres, and asks for every layer as its own file.
HIRING MULTIPLE PRODUCERS
I’m bringing on 5–6 producers in parallel, each on their own contract. You’re not competing for one slot.. spread across producers, DAWs, mix chains, voices, instruments, and AI tools is exactly what the project needs.
TWO CONSTRUCTIONS — I NEED BOTH
Type A — AI foundation, you on top. An AI-generated instrumental bed (Suno, Mureka, ElevenLabs, Google Flow, Sonauto, or similar) with your real performance recorded over it.. your vocal, your guitar, your keys, whatever you play. This is what round one was.
Type B — your foundation, AI on top. Your own production as the base: your beat, your playing, your programming, no AI generator in the foundation. Then one AI element added on top.. an AI vocal, an AI harmony stack, a generated instrument stem dropped into your arrangement. The AI element has to be a real part of the song, not a buried sample.
Each producer delivers five songs: three Type A, one Type B, and one fully human song with no AI anywhere (not even mastering). If you can do more, tell me.. I’ll pay for them.
WHAT I NEED PER SONG
Three full-song WAV bounces of the same composition:
– State A (Raw): all layers together, no mix processing, no mastering
– State B (Mix): fully mixed, no mastering
– State C (Master): mixed and mastered (AI-assisted mastering like Ozone or LANDR is fine, just tell me which)
Plus every layer as its own file:
– The AI-generated element on its own, exactly as it came out of the tool (songname_ai_source.wav)
– Each layer you added, dry, one WAV per layer (songname_stem_leadvocal.wav, songname_stem_guitar.wav, etc.). Dry means dry.. no reverb, no processing printed.
Technical: WAV, 44.1 kHz / 24-bit minimum, straight from your DAW session, no lossy formats anywhere in the chain. 90 seconds minimum per state. Peaks under 0 dBFS, no clipping, vocal audible, sounds like real music production.
AI TOOLS
Any major generator with direct WAV download. If you use Suno it has to be v6 (v6, v6-wild, or v6-mini).. the older models are retired and I need current output. Tell me which one per song. For other tools, name the tool and the version. Free-tier Suno output isn’t accepted (no commercial rights).
GENRES — ALL OF THEM
Hip-hop and R&B are welcome again, but this round I specifically want spread: rock, pop, punk, indie, country, folk, singer-songwriter, electronic, house, techno, Latin, reggaeton, afrobeats, metal, jazz, soul, blues, worship, cinematic. Tell me your lane.
VERIFICATION — READ THIS BEFORE APPLYING
Every song gets checked before payment. For each one I need:
1. Screenshot or screen recording of the AI tool’s generation page showing the interface, your username, the timestamp, and the exact generation used
2. The prompt text you typed
3. The generation date
4. A screenshot of your DAW session showing the recorded audio tracks with waveforms and the session’s created/modified date
5. For the fully human song: the DAW screenshot plus a 5–10 second phone clip of you tracking a part
6. A short process note: DAW, mix chain for State B, mastering method for State C
7. Signed Contributor License Agreement (I provide it; required before first payment)
I run every delivered file through my own audio analysis instrument and compare it against everything previously delivered. AI vocals submitted as human, processed stems submitted as dry, and files re-sent from earlier rounds get identified and the song doesn’t get accepted. Saying it plainly so the people who apply are the ones who’ll actually deliver.. and if that’s you, this is easy money for good work.
PAYMENT
– $75 per accepted song.. five songs = $375
– One milestone per song, funded in order. The first song is your verification track.. once it’s accepted, the rest activate as you deliver
– Pay only for accepted songs (technical compliance and verification, not creative quality)
– More than five songs welcome at the same rate
TIMELINE
Aiming for all five by Friday, September 25.. if it runs a little late that’s okay, you’ve got the weekend. First song within 48 hours of hire so I can check the format early, then send songs as you finish them rather than all at once. If even the weekend doesn’t fit, give me your real dates in the proposal.. I’d rather know than guess.
IN YOUR PROPOSAL, PLEASE INCLUDE
– Which AI tool(s) you use and which version (Suno users: confirm v6)
– Your DAW and mix/master workflow
– What you perform yourself (vocals, which instruments)
– Your genre lane
– Confirmation you can deliver both constructions, all stems dry, and the full verification set
– Your realistic delivery dates
This is research dataset work, not work-for-hire creative production. You keep your copyright.. I’m paying for the files, the documentation, and the license to use them for research and model training. Thanks for reading, looking forward to hearing what you make.
Comments are closed.