ArcwellStack Documentation

ArcwellStack is a voice AI platform for lifelike speech, real-time agents, and enterprise-ready voice workflows. This guide covers core concepts, API usage, and the product modules you can build with.

At a glance

Voice Studio: TTS with emotion, style, and pacing controls
Speech-to-Text: batch and streaming transcription with diarization
Agents Platform: real-time voice agents with tools and memory
Dubbing Studio: multi-language alignment and voice replacement
Voice Library: curated voices and brand voice management
Voice Cloning: consented cloning with review and watermarking
Platform: workspaces, RBAC, audit logs, usage, and billing

Quick start

Create an account and workspace.
Generate an API key in Settings → API Keys.
Make your first TTS request.

curl -X POST https://api.arcwellstack.com/v1/voice/generate \
  -H "Authorization: Bearer $ARCWELLSTACK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"voice":"Nova","text":"Welcome to ArcwellStack","format":"wav"}'

To reduce perceived latency, use streaming generation for live experiences and conversational agents.

curl -N https://api.arcwellstack.com/v1/voice/stream \
  -H "Authorization: Bearer $ARCWELLSTACK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"voice":"Nova","text":"Streaming voice in real time"}'

Core concepts

Workspace A top-level container for teams, billing, and security policies.

Project An environment for a product or use case. Projects contain voices, agents, API keys, and usage limits.

Voice A TTS model configuration. Voices can be curated presets, brand voices, or consented clones.

Agent A real-time voice experience defined by a prompt, tools, and routing rules. Agents are attached to a project and can be versioned.

Job An asynchronous task such as a dubbing run, batch transcription, or a long-form synthesis.

SDKs

Official SDKs provide typed clients, streaming helpers, and retries.

pnpm add @arcwellstack/sdk

import { Arcwell } from '@arcwellstack/sdk';

const client = new Arcwell({
  apiKey: process.env.ARCWELLSTACK_API_KEY,
});

const audio = await client.voice.generate({
  voice: 'Nova',
  text: 'Build faster with ArcwellStack',
  format: 'wav',
});

Python SDK:

pip install arcwellstack

from arcwellstack import Arcwell

client = Arcwell(api_key=os.environ["ARCWELLSTACK_API_KEY"])

audio = client.voice.generate(
  voice="Nova",
  text="Build faster with ArcwellStack",
  format="wav",
)

Authentication

Use server-side API keys for backend calls.
Scope keys per project and rotate them regularly.
Never expose private keys in client apps.

Recommended environment variable:

ARCWELLSTACK_API_KEY=your_key_here

API basics

Base URL:

https://api.arcwellstack.com

Versioning:

All endpoints are under /v1/.
Backwards-incompatible changes ship in /v2/.

Standard headers:

Authorization: Bearer <api_key>
Content-Type: application/json
Idempotency-Key: <uuid> for create requests

Pagination:

limit and cursor are used for list endpoints.
Responses include next_cursor when more results are available.

Errors:

{
  "error": {
    "code": "rate_limit_exceeded",
    "message": "Too many requests",
    "request_id": "req_01H..."
  }
}

Voice Studio (Text-to-Speech)

Generate a voice sample

curl -X POST https://api.arcwellstack.com/v1/voice/generate \
  -H "Authorization: Bearer $ARCWELLSTACK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "voice":"Nova",
    "text":"Hello from ArcwellStack",
    "format":"wav",
    "style":"warm",
    "speed":1.0,
    "pitch":0.0
  }'

Supported controls

Style and emotion (warm, calm, energetic)
Speed, pitch, and pause controls
Pronunciation with SSML-style tags
Multi-speaker rendering for dialog

Best practices

Stream for live interactions and previews
Cache frequent prompts to reduce cost and latency
Use a fallback voice for high availability

Voice Library

Browse curated voices by language, gender, tone, and use case
Save favorites and build brand voice collections
Apply voice-level usage limits for teams

Voice Cloning

ArcwellStack supports consented cloning for brand or character voices.

Requirements:

Proof of consent for the source voice
Clean audio samples with minimal noise
Review workflow before production use

Safety:

Watermarking for cloned voices
Abuse detection and automated policy checks

Speech-to-Text (ASR)

Batch transcription

curl -X POST https://api.arcwellstack.com/v1/stt/transcribe \
  -H "Authorization: Bearer $ARCWELLSTACK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "url":"https://example.com/audio.wav",
    "language":"en",
    "diarization":true,
    "timestamps":true
  }'

Streaming transcription

Low-latency tokens for real-time captions
Optional endpointing for turn detection
Speaker labels for multi-party calls

Agents Platform

Build interactive voice agents that can listen, reason, and respond with natural speech.

Key features:

Tool calling and webhook integrations
Barge-in and turn-taking controls
Memory and context policies per agent
Multi-agent routing and escalation rules

Create an agent session

curl -X POST https://api.arcwellstack.com/v1/agents/sessions \
  -H "Authorization: Bearer $ARCWELLSTACK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "agent_id":"agent_123",
    "input_audio_url":"https://example.com/user.wav",
    "response_format":"stream"
  }'

Dubbing Studio

Dubbing lets you replace speech while preserving timing and intent.

Workflow:

Upload or link the source video/audio.
Select target languages and voices.
Run alignment and review the timing.
Export mixed audio or a full video package.

Webhooks

Subscribe to events for long-running jobs and usage updates.

Common events:

job.completed
job.failed
usage.threshold_reached
billing.invoice.created

Signature verification:

import { verifySignature } from '@arcwellstack/sdk/webhooks';

const isValid = verifySignature({
  payload: rawBody,
  signature: headers['arcwell-signature'],
  secret: process.env.ARCWELLSTACK_WEBHOOK_SECRET,
});

Usage, billing, and limits

Usage is tracked by characters, audio minutes, and agent sessions.
Projects can set monthly quotas and alert thresholds.
Overage pricing and annual discounts are available on paid plans.

Rate limits:

Standard plans enforce request-per-minute limits.
Enterprise plans can request higher concurrency and dedicated capacity.

Security and compliance

ArcwellStack is designed for production and regulated workflows.

Encryption in transit and at rest
Workspace-level RBAC and audit logs
Data retention controls and deletion requests
SSO and SCIM provisioning on Enterprise
Dedicated processing regions and private networking options

Reliability and SLAs

Multi-region infrastructure and automatic failover
Status and incident updates in real time
SLA options available for Enterprise accounts

Migration guide

Moving from another TTS or voice platform?

Map voice IDs to ArcwellStack presets or clone your brand voice
Start with batch workloads, then move to streaming
Validate latency, cost, and voice quality before cutover

Support

Community and email support for all accounts
Priority support and onboarding for Enterprise
Architecture reviews for large or regulated deployments

If you need help, contact support or request a solution design session.

Documentation

On this page