# MAI Voice 2 Flash on Eden Gen

> MAI Voice 2 Flash is the low-latency variant of Microsoft's text-to-speech model, built for real-time voice agents at a lower price per character.

- Model id: `mai-voice-2-flash`
- Kind: audio
- Maker: Microsoft
- Best for: audio
- Use for: MAI Voice 2 Flash is the low-latency variant of Microsoft's text-to-speech model, built for real-time voice agents at a lower price per character.
- Speed: standard
- Features: Text to audio

## Pricing

- standard: $16.20 per 1M characters

Example: a 1,000 character line at standard costs $0.016.
Prices are for the Solo plan on the best connected route today. POST /v1/estimate returns the exact price for your workspace before a take; that is the price charged.
Failed or blocked takes are refunded automatically. Prepaid credit expires 6 months after purchase and is not refundable.

## Parameters

- `resolution`: "standard" (default "standard")
- `aspect_ratio`: "1:1"
- `count`: 1, 2 or 4 takes per run
- `references`: no references. Each item is `{ "asset_id" | "url" | "take_id" | "element_id", "role": "reference" }`.
- `project_id`: the project the cost lands on (defaults to the key's project, then the newest one).
- `webhook_url`, `metadata`: optional.
- `tags`: up to 10 labels for usage reports, e.g. `{ "client": "acme", "campaign": "q4" }` (or header `X-Takes-Tags: client=acme,campaign=q4`).

Unsupported values return HTTP 400 with `param` and `supported` in the error, so you can correct and resend.

## Call it

```bash
curl -X POST https://edengen.ai/v1/takes \
  -H "Authorization: Bearer $TAKES_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -H "Prefer: wait=60" \
  -d '{"model":"mai-voice-2-flash","prompt":"A calm narrator reads the line, close and unhurried","resolution":"standard","aspect_ratio":"1:1"}'
```

Check the price first with `POST https://edengen.ai/v1/estimate` (same body). Poll `GET https://edengen.ai/v1/takes/{id}` or pass `webhook_url`.
Never resend a create to check progress: a new request is a new paid take unless it carries the same Idempotency-Key.

## More

- All models: https://edengen.ai/llms.txt
- OpenAPI: https://edengen.ai/v1/openapi.json
- MCP server (OAuth, no key needed): https://edengen.ai/mcp