suede-ai

A Suede Labs AI project · Built by Jason Colapietro

Python SDK for Suede Labs AI — x402 pay-per-call music generation, stem splitting, MIDI transcription, mastering, lyric sync, guitar rig tools, and IP-registry lookups. 22 endpoints settled in USDC on Base. No API keys. No subscriptions. Sign an EIP-3009 authorization and call the endpoint.

The SDK wraps the 402-challenge / sign / retry loop so your agent code spends its time writing creative prompts, not encoding typed data.

Install

pip install suede-ai

Requires Python 3.10+. Pulls in httpx, eth-account, and pydantic.

60-second quickstart

from suede_ai import SuedeClient

# Funded EOA on Base with USDC. Treat this like any other secret.
PRIVATE_KEY = "0x..."

with SuedeClient(wallet_private_key=PRIVATE_KEY) as suede:
    track = suede.create_music(
        prompt="lo-fi rainy afternoon, vinyl crackle, soft piano",
        duration_seconds=30,
    )
    print(track["assetUrl"])      # https://cdn.suedeai.ai/audio/trk_...mp3
    print(track["provenance"])    # {"fingerprint": "0x..."} — on-chain attestation

The first call returns 402 with the x402 challenge. The SDK signs an EIP-3009 transferWithAuthorization for USDC on Base, replays with X-PAYMENT, and returns the JSON body. You never touch the typed data.

The 22 endpoints

Method	Endpoint	Price (USDC)	What it does
create_music	POST /create-music	0.20	Rights-aware music generation
agent_generate	POST /agent/generate	0.20	Agent-facing music output (same payload)
agent_video	POST /agent/video	1.50	Short music-video clip generation
extend	POST /v1/extend	0.40	Continue an existing Suede track
cover	POST /v1/cover	0.40	Stylistic re-imagining of a track
voice_cover	POST /v1/vox	0.40	Replace lead vocal with a Suede voice
continue_track	POST /v1/continue	0.40	Extend an uploaded audio file
stems_pro	POST /v1/stems-pro	0.40	4-stem split: vocals / drums / bass / other
stems_basic	POST /v1/stems	0.20	2-stem split: vocals + instrumental
vox	POST /v1/acapella	0.20	Isolate the vocal stem
midi	POST /v1/midi	0.10	Transcribe audio to MIDI
wav_master	POST /v1/mastering	0.10	High-quality WAV master
lyric_sync	POST /v1/lyric-sync	0.10	Timestamped lyrics for a track
lyrics	POST /v1/lyrics	0.04	Generate fresh song lyrics from a prompt
style_coach	POST /v1/style-coach	0.02	Expand short tags into a prompt-ready style brief
rights_lookup	GET /v1/rights/{assetHash}	0.005	Suede Registry attestation lookup (owner / IP / NFT)
analyze	POST /v1/analyze	0.003	BPM / key / mode / energy / danceability
prompt_analyze	POST /v1/prompt-analyze	0.003	Extract genre, mood, instrumentation from a prompt
chain_chat	POST /v1/chain-chat	0.02	Plain-language Q&A about on-chain rights / royalties
rig_analyze	POST /v1/rig/analyze	0.10	Infer guitar signal chain from audio
rig_oracle	POST /v1/rig/oracle	0.10	Recommend a full guitar rig for a target tone
rig_roast	POST /v1/rig/roast	0.05	Roast a gear list for laughs

Prices are sourced from the live manifest at the time of writing and are enforced server-side.

How payment works

1. Client calls POST /create-music.
2. Server returns 402 Payment Required with a JSON accepts array. Each entry declares scheme (exact), network (eip155:8453), asset (USDC on Base), payTo address, and maxAmountRequired.
3. SDK builds an EIP-3009 TransferWithAuthorization typed message and signs it with the configured wallet.
4. SDK base64-encodes the payment payload and replays the original request with the X-PAYMENT header.
5. Suede's facilitator settles on-chain; the response body returns asset URL + on-chain provenance.

You can inspect the live manifest yourself:

curl https://app.suedeai.ai/.well-known/x402.json | jq

Advanced

Reuse an httpx.Client

import httpx
from suede_ai import SuedeClient

http = httpx.Client(http2=True, timeout=120.0)
suede = SuedeClient(wallet_private_key=PRIVATE_KEY, http_client=http)

Direct call to any endpoint

result = suede.request("POST", "/v1/style-coach", json={"tags": "lofi, rainy"})

Inspect the live manifest

manifest = suede.manifest()  # free — no payment required

Guitar rig tools

# What signal chain is this guitar recording running?
chain = suede.rig_analyze(audio_url="https://cdn.example.com/riff.mp3")

# Recommend a rig for a target tone
rig = suede.rig_oracle(goal="warm blues crunch", genre="blues", budget_usd=1500)

# Roast your pedalboard
roast = suede.rig_roast(goal="tight metal", gear=["Boss DS-1", "Line 6 Spider"])

Rights and on-chain tools

# Q&A about on-chain rights for a specific asset
answer = suede.chain_chat(
    question="Who owns this track and what licenses are active?",
    asset_hash="0xabc123...",
)

# Analyze what genre/mood a text prompt implies
analysis = suede.prompt_analyze(prompt="dark cinematic orchestral tension build")

Roadmap

• Async client (AsyncSuedeClient)
• Pydantic response models per endpoint
• examples/ folder: LangChain tool wrappers, CrewAI tasks, agentcash adapter

License

MIT. See LICENSE.

---

Built by Jason Colapietro · Suede Labs AI · x402 manifest