Skip to main content
MiniMax builds its own language and speech models, reached through one account and one key. In SentiVue Talk it drives the language model and the voice.

​
What you need

A Group ID is required whenever this connection drives the voice. Saving a configuration that uses MiniMax for voice without one is rejected. If the connection only drives the language model, leave it blank.

​
Getting your credentials

Both credentials live on the same page of your MiniMax account, so one visit gets you everything.
  1. Sign in at platform.minimax.io and open User Center → Basic Information.
  2. Copy the GroupID shown on that page. It identifies your account and scopes every speech request to it.
  3. Open the Interface Key tab on the same page and create a new secret key.
  4. Copy the key and store it somewhere safe before leaving the page. MiniMax keys are long tokens, so take the whole value rather than what fits on screen.
MiniMax runs two separate platforms: the global one at platform.minimax.io and a mainland China one at platform.minimaxi.com. They are separate accounts with separate keys, and a key issued on one does not authenticate on the other. SentiVue Talk points at the global platform by default. Sign in to the console matching the host you intend to use — if your key is not listed at platform.minimax.io, it belongs to the other region.

​
Adding the connection

  1. Go to Connections and choose MiniMax.
  2. Paste the API key.
  3. Enter the Group ID if this connection will drive the voice.
  4. Choose which services this connection drives — language model, voice, or both — and save.
Place a test call after connecting. SentiVue Talk checks that a Group ID is present before saving a voice configuration, but a wrong or expired key is not caught at that point — it surfaces on the first live call.

​
What you can select afterwards

Choose your models on Model Configurations. Language model — browse the current names on MiniMax’s models overview. Voice — pick a voice by its voice ID, from the system voice ID list or one you created. Speech models come in hd and turbo pairs; start on turbo for calls.

​
Using MiniMax for only one service

The connection does not have to drive both roles. Turn off the one you do not want, and SentiVue Talk uses your other connections for it — MiniMax for the language model with voice from ElevenLabs, or MiniMax voices in front of a different language model. Change this at any time on the connection. If you turn the voice off, the Group ID is no longer needed. If you later turn it back on, add one before saving.

​
Custom hosts

Each service has its own base URL, and the two defaults are different: The voice default points at a deeper path than the language model one. If you override it, keep the full path rather than trimming it back to /v1. Set these if you reach MiniMax through a proxy or gateway, or if your account is on the mainland China platform — in that case swap the host for api.minimaxi.com on both services, keeping the rest of each path unchanged. Note the extra i: the two hostnames differ by a single character.

​
Completing the pipeline

MiniMax covers the language model and the voice, but has no transcription. You also need: That is a separate connection.

​
Troubleshooting

A Group ID is required whenever the connection drives the voice. Copy it from User Center → Basic Information in the MiniMax console — it sits on the same page as your API key. Alternatively, turn the voice off and use another provider for it.
The key is most likely from the wrong MiniMax region. Global keys work only against api.minimax.io and mainland China keys only against api.minimaxi.com; the error reads as an invalid key even when the key itself is valid. Sign in to both consoles and check which one lists your key, then match the base URLs to it. Also confirm you pasted the whole key — these are long and easy to truncate.
Enter the voice ID, not the display name. Check the exact value against MiniMax’s system voice ID list. For a cloned or designed voice, use the ID you assigned at creation and confirm the voice still exists on your account — MiniMax removes a custom voice that has never been used within its retention window.
Move to a -highspeed language model variant and a turbo speech model. If it is already on both, check your account balance — MiniMax is pay-as-you-go, and requests fail or throttle once it runs out.