> This page is for Model APIs.

> Sarvam Voice Agents (Samvaad) is a voice AI agent platform for Indian languages: build, test and deploy AI calling agents on your own phone numbers, from ₹3 a minute. Start here: https://docs.sarvam.ai/voice-agents/why-voice-agents.
> For clean Markdown of any page, append `.md` to the page URL.
> For a complete documentation index, see https://docs.sarvam.ai/llms.txt. For a section index, append `/llms.txt` to any section URL (for example https://docs.sarvam.ai/voice-agents/llms.txt).
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.sarvam.ai/_mcp/server.

# Limitations

> What Dictate does not do and its limits, so you can decide before you build.

Read this page before you build on it.

## What Dictate does not do

| Not offered                                                                                       | What to do instead                                                         |
| ------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------- |
| **Subtitle files (SRT, VTT)**                                                                     | Build them from section times in your own code, or use Speech-to-Text      |
| **Browser tokens.** There are no short-lived keys. A browser WebSocket cannot send the key header | Open the socket from your server                                           |
| **A `translate` mode**                                                                            | Ask for the translation in `style.rules`                                   |
| **Memory between calls or turns.** There is no field for earlier turns                            | Add names to `vocabulary` as they are said (`config.update` on the socket) |
| **A recording over 1 hour**                                                                       | Split the file, and send each part as a job                                |

## Limits

| Limit                                | Value                                                                                                                                   |
| ------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------- |
| Clip length and size                 | 30 s, up to 10,000,000 bytes                                                                                                            |
| Job length                           | 1 hour. An upload under about 100 MB. Use `audio_url` for more                                                                          |
| Job results are kept                 | 7 days. Delete a job when you have the result                                                                                           |
| Live session                         | 3 hours. Closes after 60 s with no message from you. Each message up to 1 MiB, and at most 1 s of audio (0.5 s with `stream_type=fast`) |
| `vocabulary`                         | Up to 1,000 entries                                                                                                                     |
| `context.reference` / `context.text` | 50,000 / 2,000 characters                                                                                                               |
| `style.rules`                        | 2,000 characters                                                                                                                        |
| `output_rules`                       | 4,000 characters                                                                                                                        |
| Other form fields                    | At most 1 MiB each                                                                                                                      |
| Rate limits                          | Per account, the same as Speech-to-Text. See the [pricing page](/api/getting-started/pricing#dictate-limits)                            |