Provider Async

Async

Mode: audio · Models: 1

Vendor: Async AI · Official API docs: docs.picsart.io

Async AI provides Async Flash v1.0, a low-latency text-to-speech model with a large voice library (100+ named voices) and fine-grained control over the audio container, sample rate, encoding, and bit rate. It is exposed through Picsart's generation platform.

Models

id	Name	Input type
async-flash-v1	Async Flash v1.0	tts

CLI

# basic text-to-speech (default voice, mp3)
gen-ai generate -m async-flash-v1 \
  -p "Welcome to Picsart AI Playground." -s

# pick a voice and a WAV container at a higher sample rate
gen-ai generate -m async-flash-v1 \
  -p "Your export is ready." \
  --voice f493c663-b272-493e-8b78-72d2262a2a8d \
  --container wav --sample-rate 48000

Run gen-ai models info async-flash-v1 --json for the full voice list (100+ voices).

MCP

{ "name": "picsart_generate",
  "arguments": {
    "model": "async-flash-v1",
    "prompt": "Welcome to Picsart AI Playground.",
    "voiceId": "cca0e076-94b9-4c6d-86b7-546168f11174",
    "container": "mp3"
  } }

Parameters

Full parameter surface, sourced from gen-ai models info <id> --json. CLI flags show the primary short form; the canonical --kebab-case long form always works too.

async-flash-v1 — Async Flash v1.0

Try async-flash-v1 in Playground ↗

Input type: tts

Param	CLI flag	Type	Values
prompt	-p	text	required
voiceId	--voice	enum	100+ voices — e.g. cca0e076-… (Jennie, default) · f493c663-… (Stella) · 317bf805-… (Max) · a6268eaf-… (Elara) · 7a3ef29d-… (Grace). Run gen-ai models info async-flash-v1 --json for the full list
container	--container	enum	mp3 · wav · raw (default mp3)
sampleRate	--sample-rate	range	8000–48000 (default 24000)
encoding	--encoding	enum	pcm_s16le · pcm_f32le (default pcm_s16le)
bitRate	--bit-rate	range	32000–320000 (default 192000)

Notes: encoding and bitRate apply to raw/PCM and compressed containers respectively. Voice IDs are opaque UUIDs — use the --json model info to map IDs to names.

Pricing

gen-ai pricing async-flash-v1

Text-to-speech is priced per generation; cost is resolved per modelId via the backend /options call.