Every route checked. One exact price.
Eden Gen runs each take on the best-value route and shows you the exact price before it starts. Failed takes cost nothing, and one balance works on every credit route.
- Routes checked
- 6
- Runs on
- WaveSpeed
- Top-up fee and failed takes
- $0
For freelancers testing models on real jobs.
- Most models on Eden Gen credit
- The rest with your own provider key
- Projects and spend caps
- Failed takes refunded to your balance
- We do not train on your prompts or files
- DPA included for every workspace
- Credit expires 6 months after purchase
For teams billing AI work to clients.
- Everything in Solo
- Bank transfer for top-ups
- Tax invoice when credit is purchased
- Per client usage statement
- Seat roles and per person caps
- Prompt, model and provider on every take
For agencies and broadcasters with volume.
- Everything in Studio
- Postpaid monthly invoice, NET 14
- Separate agreement and order form
- 99.5% uptime target
- Direct support for Agency
Set up with our team under a separate agreement.
Prices in USD, excluding VAT. Pay in USD, EUR, GBP or CZK. Prepaid credit is valid for 6 months and is not refundable.
Average across 5 models where another route costs less, plan fee included. Prices recorded 31 Oct 2026.
The model where the best route sits furthest under its maker's own API, same model and settings.
Model makers' own APIs, API platforms and your own keys, compared before every take.
It also knows who spent what, on which project.
Every take is filed to a project the moment it is made, by a person, an agent or an API key. Studios see spend per client and per teammate. Freelancers see what each job cost before they invoice it.
MatejProducer
KlaraArtist
JonasEditorMonthly limit per person and per key. Nobody can spend past it.
The app, the API and MCP agents all name a project or a tag. Budgets, exports and the invoice split follow from that, with no spreadsheet in between.
Against the priciest connected route for the same takes. Your workspace shows this number every month.
How the saving is measuredSee what you could save.
Pick a model, a length and how many takes you run. See what you pay on the route Eden Gen picks against the list price of the model maker's own API.
The Eden Gen row is what you pay for this batch on the selected plan, the price you confirm before each take; seats are billed separately and prices exclude VAT. The other row is the published list price of the model maker's own API for Seedance 2.5 at 720p, per second, recorded on 31 Oct 2026. List prices change often, and savings vary by model and setting: only models where a route costs less than the maker's API are listed here. How routing works
The things other platforms leave to support tickets.
How does billing work?
Solo and Studio prepay credit. Each top-up is a tax invoice: Czech VAT 21%, EU reverse charge when the VAT ID checks out in VIES (Czech VAT when it does not), or outside the scope of EU VAT. You can pay in USD, EUR, GBP or CZK; the balance stays in USD and the amount, VAT and exchange rate are shown before you pay. Usage of that credit is a statement, not a second invoice. Credit expires 6 months after purchase, oldest first, and is not refundable. Agency is postpaid: one monthly invoice, NET 14. Failed takes are refunded to your balance on every plan.
Do I pay for failed takes?
No. If no file lands, the price goes back to your balance automatically: provider timeout, content filter, a job stuck in the queue. You see the reason next to the take.
Can I cap what a project costs?
Yes. Every project has a hard cap. A take that would cross it is stopped before it runs, with the exact numbers, and you can raise the cap in one click.
What about the EU AI Act?
Every output file name includes the take ID. The take stores the prompt, model version, provider and time, and you can copy that record from the take. Labeling duties stay with you. We do not train on your prompts or files; providers process them under their API terms. The DPA at /dpa applies to every workspace.
Is there an API?
Yes. A REST API with an OpenAPI file, an MCP server for Claude Code, Codex and Cursor, tool schemas for DeepSeek and other function-calling models, and signed webhooks. Same prices, caps and refunds as the app. See the docs.
Where do the models run?
Through provider APIs only: the model makers' own (BytePlus ModelArk, OpenAI, Vertex AI, Kling, Alibaba, MiniMax, ElevenLabs, Recraft) and API platforms (Sunra, Sume, Higgsfield, OpenRouter, WaveSpeed, fal, Pika API Club, Replicate, DeepInfra). Eden Gen checks every route that is ready before a take and runs the best value. Some routes run only with your own provider key. We do not automate consumer accounts.