Skip to main content
Arcana is retired on August 15, 2026 at 12:00 UTC. From that point, cloud requests that specify an Arcana model are served by Coda. Migrate before then so you can test the change on your own schedule rather than discovering it in production.
Plan your cutover against 12:00 UTC rather than against a local calendar date. If you operate outside the Americas, the switch lands in the middle of your working day on August 15, not at the end of it.

​
Does this affect you

You are affected if any request, config file, or environment variable in any environment sets modelId to arcana, arcanav2, or arcanav3. You are not affected by omitting modelId. Requests without it are served by Mist v3, which is unchanged. To check your own usage, search your code and configuration for those three strings, and repeat the search for every deployed environment rather than only your local checkout. There is no response header naming the model that served a request, so confirm by inventory and configuration review, not by inspecting a response.

​
Where each language goes

​
What changes when you switch to Coda

The API contract is unchanged, so modelId: coda is the only edit most integrations need. These differences are not errors and will not appear as failures:
  • Your speaker may not exist on Coda. Voice availability differs per model. Check yours in the voice catalog before switching
  • Word timestamps cover English and Spanish only. If you rely on them for interruption handling in another language, no timestamps event arrives and no error is raised
  • Audio will not be identical. Coda is a different model, so prosody and pacing differ even for a speaker present on both

​
After the cutoff

If you have not migrated, requests naming an Arcana model continue to succeed but are served by Coda. You will not receive an error. The audible difference is the signal that migration happened without you.

​
Get help

If you have any questions or concerns about migrating, contact support@rime.ai before August 15. Include your account, the model IDs you use, and the languages you serve, and someone will work through it with you.