Private beta — join the waitlist
The control plane for generative media.
One API above sixteen providers — fal, Replicate, OpenAI, Google, Black Forest Labs and more — with cross-provider failover, byte-exact custody of every output, and a C2PA-marked audit trail. Like OpenRouter, but for media pipelines that have to survive an audit.
Don't stop using fal — stop depending on fal.
The product demoing itself: one request surviving a provider failure, exactly as the ledger records it. Every visual on this page is a real pipeline output, receipt attached.
10 verified provider integrations — every one proven with a real job
fal
Replicate
Runware
ElevenLabs
EvoLink
Kie.ai
PiAPI
LaoZhang
OpenAI
Black Forest LabsThe problem
Provider infrastructure is built to serve inference — not to be depended on.
Your outputs vanish
Providers delete generated files aggressively — Replicate after about an hour, fal after about seven days. If a provider URL is your only copy, you have already lost the master.
Models retire under you
Vendors sunset models within months of launch. A pipeline pinned to one model inherits that model's lifespan — usually without notice you can act on.
429s arrive with your traffic
Capacity errors spike exactly when you launch. A single-provider pipeline turns someone else's overloaded GPUs into your incident.
Aggregators fail over within themselves
Every aggregator retries inside its own catalog. When the aggregator itself degrades, every model behind it degrades with it. Nobody fails over across companies.
What Wakemark does
Route it. Keep it. Prove it.
Route & survive
Capability-based routing across sixteen provider integrations — different companies, not one catalog. On a 429, a 5xx, or a timeout, the job fails over automatically and your endpoint keeps answering. Retries never fire on your own bad input, so failover never double-spends.
Keep every byte
Every output is streamed into storage and SHA-256-verified before provider URLs expire — byte-exact, never transcoded. Platform bucket by default; export to a bucket you own is on the near-term roadmap.
Prove everything
C2PA Content Credentials on outputs, an append-only event ledger of every provider interaction, and a lineage graph from prompt to published asset. An audit surface no aggregator offers.
Own your masters.
Generated media is an asset your organization paid for. Wakemark treats it that way: hash-verified custody, under your control, for as long as you need it — not until a provider's cleanup job runs.
EU AI Act · Article 50
The transparency deadline is dated. Your evidence should be too.
Article 50 requires outputs of generative AI systems to be marked in a machine-readable format and detectable as artificially generated. The Commission's Code of Practice on marking and labelling (final version published 10 June 2026) names C2PA Content Credentials as an exemplar mechanism — the same standard Wakemark stamps on outputs today.
2 Aug 2026
EU AI Act Article 50 transparency obligations apply.
2 Dec 2026
Deadline for machine-readable marking for systems already on the market.
€15M / 3%
Maximum penalties — €15M or 3% of worldwide annual turnover.
Designed to support Article 50 workflows
- Machine-readable marking by default — C2PA Content Credentials attached to outputs as they move through the pipeline.
- An append-only event ledger recording every generation: provider, model, parameters, timestamps, raw responses.
- Byte-exact archived outputs whose hashes match what the ledger says was produced.
- A lineage graph connecting each published asset back to the jobs and inputs that made it.
Wakemark is infrastructure, not legal advice. Article 50 obligations sit with the organizations that provide and deploy AI systems — how you meet them is a decision for you and your counsel. Wakemark gives you the marking, custody, and audit trail to build on.
For developers
One request. Sixteen providers. No babysitting.
curl -X POST https://api-staging.wakemark.ai/v1/jobs \
-H "Authorization: Bearer wk_live_…" \
-d '{
"capability": "text-to-image",
"input": { "prompt": "product hero shot, studio light" },
"routing": {
"providers": ["fal", "replicate", "runware"]
},
"idempotency_key": "launch-hero-001"
}'
# fal returns 429 at your launch spike → the job fails over
# to replicate; the ledger records both attempts.
{
"id": "job_01JZKQ6M3R…",
"status": "succeeded",
"provider": "replicate",
"assets": [{ "sha256": "9f2c…", "c2pa": true }]
}Cross-provider failover
Retries route across companies, not within one catalog — and only on provider-side failures, never on your own bad input.
Byte-exact custody
Outputs are streamed to storage and SHA-256-verified. Provider URL expiry stops being your problem.
Hard spend caps
Caps are enforced before submit. A job that would cross the cap is rejected, not billed.
C2PA passthrough
Content Credentials are stamped and preserved through the pipeline, sidecars included.
Passthrough mode
Provider-specific parameters are forwarded untouched. The job still lands in the ledger; the output is still archived.
From the pipeline
Real outputs, with their receipts attached.
Everything below was generated through Wakemark and archived byte-exactly — hover to play. Each tile lists the job that produced it, the provider that served it, and the sha256 the ledger recorded. Our gallery carries Content Credentials.
Pricing
Simple plans, billed by seat — never a markup on routed spend.
Launch pricing — subject to change before general availability.
Free
$0
for solo builders
- Bring your own provider keys (BYOK)
- Capability routing + cross-provider failover
- Byte-exact output custody
- Audit ledger & lineage graph
- Community support
Team
$49–99
per month
- Everything in Free
- Seats for your team
- Hard spend caps & usage controls
- Higher concurrency
- Priority support
Enterprise
Contact us
custom terms
- Everything in Team
- Export custody to your own bucket (roadmap)
- SSO
- DPA & subprocessor list
- Dedicated support
FAQ
Honest answers
Is this just another aggregator?
No. Aggregators sell access to many models behind one API and fail over only within their own catalog. Wakemark is the layer above providers and aggregators: it routes across different companies, keeps byte-exact custody of the outputs, and maintains the audit trail. Keep your fal account — Wakemark makes it survivable.
Do you resell provider credits?
No. Wakemark is BYOK: you bring your own provider keys, and providers bill you directly at their own prices — no markup on routed spend. Keys are stored envelope-encrypted (a per-secret AES-GCM data key, wrapped by a KMS root key and bound to your tenant) and exist in plaintext only in memory, at the moment a request is signed.
Is the media on this page real?
Yes — and that is the point. Every loop and still on this site was generated through Wakemark's own API against real providers, archived byte-exactly, and is listed with the job id that produced it and the sha256 the ledger recorded. No stock footage anywhere. Our gallery carries Content Credentials.
What happens to my bytes?
Every output is streamed into storage and verified against its SHA-256 hash — byte-exact, never transcoded, never “optimized”. Outputs live in a private bucket served from a dedicated media domain; export to a bucket you own is on the near-term roadmap. Provider URLs expire in hours or days. Your masters don't.
When does Article 50 apply?
Article 50 of the EU AI Act applies from 2 August 2026. Systems already on the market before that date have until 2 December 2026 to implement machine-readable marking. The Commission's Code of Practice (final version, 10 June 2026) names C2PA Content Credentials as an exemplar marking mechanism. Penalties run up to €15M or 3% of worldwide turnover.
Does using Wakemark make me compliant?
No tool can make you compliant, and you should be suspicious of any that claims to. Wakemark is designed to support Article 50 workflows — machine-readable C2PA marking, a verifiable audit ledger, durable custody — but the legal obligations sit with your organization and depend on how you deploy. Bring your counsel; we bring the evidence trail.
I need a provider-specific feature your API doesn't model.
Use passthrough mode: provider-specific parameters are forwarded untouched, while the job still flows through the ledger, the output is still persisted, and C2PA marking still happens. You get the provider's exotic features and keep the audit trail.
Publishing AI-generated media into the EU?
The deadlines are dated and the wave is here. Join the waitlist for the private beta — we onboard in small batches, in order.