Speech And Conversational — hobby stack.
For each requirement below, pick the option that fits your build — recommended first, then free and cheaper alternatives — or skip what your project doesn't need. Tap the info icon next to any requirement to see why it matters.
Audio Object Storage (Recordings, Prompts, Datasets)
Where you keep files users upload or you serve — images, videos, documents — and how fast they reach visitors around the world.Pricing & free-tier limitsS3: Free 5GB 12mo, then $0.023/GB/mo, $0.0004/1k PUT, $0.0004/10k GET, egress $0.09/GB. R2: Free 10GB storage, 10M Class A, 10M Class B, ZERO egress, then $0.015/GB. B2: Free 10GB, then $0.006/GB/mo, egress $0.01/GB (3x free). Overage: S3 requests $0.005/1k.
Speech-to-Text (STT) API / Hosting
Where your code actually runs and serves requests. Picking the right host affects speed, scaling, and how much ops work you do.Pricing & free-tier limitsDeepgram: $200 free credits (~46k mins), then $0.0043/min pre-recorded, $0.0048/min streaming, $0.004/min Nova-2. Whisper API: $0.006/min. AssemblyAI: Free 5 hrs ($50), then $0.00025/sec ($0.015/min), $0.015/min extra speaker labels. Self-hosted: ~$0.0008/min on A10G ($0.75/hr).
Text-to-Speech (TTS) API / Hosting
Where your code actually runs and serves requests. Picking the right host affects speed, scaling, and how much ops work you do.Pricing & free-tier limitsElevenLabs: Free 10k chars/mo (~10 mins), Starter $5/mo 30k chars, Creator $22/mo 100k, Pro $99/mo 500k, $0.30/1k overage. AWS Polly: Free 12mo 5M chars, then $16/1M neural chars. Self-hosted Coqui: $0 GPU if local, A10G $0.75/hr = ~$0.001/1k chars. Cartesia: $0.015/1k chars ultra-low latency 40ms.
GPU Compute for Speech Inference (Whisper/TTS)
Where your code actually runs and serves requests. Picking the right host affects speed, scaling, and how much ops work you do.Pricing & free-tier limitsModal: Free $30/mo credits, A10G $0.000306/sec ($1.10/hr), A100 40GB $0.00064/sec ($2.30/hr), H100 $0.00122/sec ($4.39/hr) autoscale to 0. RunPod: Free $5 trial, A10G $0.38/hr, A100 $1.19/hr, H100 $2.29/hr spot. Vast.ai: A10G $0.26/hr, A100 $0.99/hr. AWS p4d: A100 $3.06/hr on-demand, H100 p5 $5.98/hr. Overages billed per-second.
Real-time Voice Infra (WebRTC, Voice Agents)
A piece of your stack you may or may not need, depending on scope. Pick the option that fits — or skip it if your project doesn’t require this capability yet.Pricing & free-tier limitsLiveKit Cloud: Free 50GB/mo + 50k mins, then $0.004/min audio SFU + $0.015/participant/min AI agent. Daily.co: Free 10k mins, then $0.004/min + $0.01/min transcription. Cloudflare Calls: Free 1k mins, then $0.01/min + $5/1k MAU. Agora: Free 10k mins, then $0.00099/min voice, $0.015/min AI agent. Self-hosted LiveKit: $5-20/mo VM cost.
Conversation State Management (Dialog Memory)
A piece of your stack you may or may not need, depending on scope. Pick the option that fits — or skip it if your project doesn’t require this capability yet.Pricing & free-tier limitsUpstash: Free 10k cmds/day, 256MB, then $0.20/100k cmds, $0.25/GB storage. Redis Cloud: Free 30MB, then Essentials $5/mo 250MB, $0.15/GB. ElastiCache Serverless: Free tier none, $0.125/GB-hr storage + $0.20/1M ECPUs. Self-hosted: $0 + $6/mo VPS.
Vector Database for Conversation Memory & RAG
Persistent storage for your app’s data — users, products, orders. The single most important architectural decision for most projects.Pricing & free-tier limitsPinecone: Free 100k vectors (2GB), then Starter $70/mo 2M vectors, $0.33/1M reads, $0.07/1k writes. Qdrant: Free 1GB cluster, then $0.25/GB + $0.05/M op, 25$/mo for 4GB. Supabase: Free 500MB pgvector, then $25/mo 8GB. Self-hosted Qdrant: $0 + $10/mo VM for 2M vectors.
LLM Inference for Dialogue Management / Agent Reasoning
Powers AI features — model access, embeddings, and inference. Costs scale with usage, so watch the meter.Pricing & free-tier limitsGroq: Free 14.4k tokens/sec limit, $0.59/M input, $0.79/M output (Llama 70B). Together: Free $25 credits, $0.88/M in/out Llama 70B. OpenAI GPT-4o: $0.0025/1k input, $0.01/1k output, Free $5 trial. Claude 3.5 Sonnet: $0.003/1k in, $0.015/1k out. Anyscale: $0.15/M input Llama 70B, $1/min overage. Overage: Groq 5k req/min limit.
Background Jobs for Async Transcription & TTS
A piece of your stack you may or may not need, depending on scope. Pick the option that fits — or skip it if your project doesn’t require this capability yet.Pricing & free-tier limitsInngest: Free 50k steps/mo, 30k events, then $20/mo 500k steps, $0.10/1k overage. Trigger.dev: Free 10k tasks, then $30/mo 100k, $0.20/1k overage. QStash: Free 500 msg/day, then $10/mo 40k msgs, $0.30/10k. AWS Step Functions: Free 4k transitions, then $0.025/1k transitions.
Queue for Audio Processing Jobs (Durable)
Where you keep files users upload or you serve — images, videos, documents — and how fast they reach visitors around the world.Pricing & free-tier limitsSQS: Free 1M req/mo, then $0.40/1M req, FIFO $0.50/1M. Cloudflare Queues: Free 1M ops, then $0.50/M ops. QStash: Free 500/day, $10/mo 40k. Confluent Kafka: Free $400 trial, then $0.20/GB in/out, $0.001/vCPU-hr. RabbitMQ Cloud: Free 100 queues, then Little Lemur $19/mo.
API Gateway + Rate Limiting for Voice API
Controls and protects your APIs — quotas, abuse prevention, and firewalls. Important once you have real traffic or many clients.Pricing & free-tier limitsCloudflare Workers: Free 100k req/day, 10ms CPU, then $5/mo 10M req inclusive, $0.30/M overage. Upstash Rate Limit: Free 10k/day, $0.20/100k. AWS API Gateway: Free 12mo 1M calls, then $3.50/M REST, $1/M HTTP API. Kong Konnect: $250/mo platform + $0.02/1k req.
How this hobby speech and conversational checklist works.
Each requirement below is something a hobby speech and conversational build typically needs. Pick one of the four researched options — recommended, free, cheaper or paid — add your own with "Other", or skip the requirement if your project doesn't need it. Nothing is mandatory; the plan on the right tracks what you've decided so nothing gets forgotten.
Your picks are saved in this browser automatically, so you can come back anytime. Options are researched per build level and refreshed as vendors change their plans — always verify details on the provider's page before committing.