Deploy and Host self hosted Cap AI Transcripts (Open-Source Loom Alternative) on Railway

This template runs Cap, the open-source screen recorder, with transcripts and AI titles, summaries and chapters switched on. Add provider keys on cap-web and your recordings stay in your own bucket. Built for small teams leaving Loom.

About Hosting Cap AI Transcripts open-source software on Railway (self hosted Cap AI transcripts template)

This template runs the official Cap stack: pinned cap-web, matching media-server, MySQL 8 and a private Railway Bucket. Recordings and metadata stay in your bucket and database.

Only audio and transcript text reach a provider, and only once you add a key. Railway handles TLS, private networking, health checks and the bucket CORS rule on every boot. Migrations run on first boot, so the login code shows up in the cap-web logs until you add Resend.

Why Deploy Cap AI Transcripts, the Loom alternative on Railway (Railway Free Trial)

Loom bundles AI features into paid per-seat plans, and your recordings live on its servers. Here you add ASSEMBLY_API_KEY and a Groq or OpenAI key to cap-web, and the transcript and summary run on your own Cap while the video stays in your bucket. The $5 Railway trial covers the first stretch while you test keys.

Railway is a singular platform to deploy your infrastructure stack. Railway will host your infrastructure so you don't have to deal with configuration, while allowing you to vertically and horizontally scale it.

By deploying Cap AI Transcripts on Railway, you are one step closer to supporting a complete full-stack application with minimal burden. Host your servers, databases, AI agents, and more on Railway.

Railway vs Other Hosting Providers and VPS for Cap AI Transcripts self hosting

Railway manages the four Cap services; a VPS makes you wire them yourself.

Provider	What You Get with Railway	What You Get with the Other Provider
DigitalOcean	Managed deploy, private networking, and a bucket for recordings.	Droplets with Spaces, firewall, TLS, and backups you configure.
AWS	Managed MySQL and S3 storage; only cap-web is public.	EC2, RDS, S3, VPC, IAM, and certificates you own.
Hetzner	Small-team hosting at about 850 MB idle RAM, plus bucket storage.	Cloud VPS, but you run Docker, MySQL, MinIO, and CORS.

Either way, recordings stay in your bucket and AI keys go on cap-web.

Common Use Cases for hosted Cap AI Transcripts

Searchable support transcripts: With ASSEMBLY_API_KEY set, every ticket walkthrough gets a transcript, so agents search past replies instead of re-recording the same fix.
Chapters for long walkthroughs: A 20-minute onboarding call becomes labeled jumps, which beats a summary when someone only needs step four.
AI titles for a readable library: Generation providers name recordings, so a folder of screen-2026-09-14 files turns into something people scan.
Captions for accessibility: Transcript text becomes captions on the share page, covering deaf viewers and anyone watching muted at a desk.
Summaries pasted into tickets: Drop the summary into Jira or Linear so the thread has context without asking anyone to watch six minutes.

Dependencies for Cap AI Transcripts Docker hosted on Railway

The template starts cap-web, media-server, MySQL 8.0.46 and a private Railway Bucket, but sets no AI keys. Auto-detection prefers Groq, then OpenAI, then Anthropic for generation, and Anthropic, then OpenAI, then Groq for chat.

Deployment Dependencies for Managed Cap AI Transcripts Service (Screen Recording)

cap-web is the only public service, on port 3000 with health check /api/status and a volume at /app/apps/web/.workflow-data. media-server stays private on port 3456.

Implementation Details for Cap AI Transcripts (Using Cap official docker image)

Both images are pinned by digest from ghcr.io; the bucket CORS rule is re-applied to WEB_URL on every boot. Add keys as variables, redeploy cap-web, and remove a key to turn that feature off.

How does Cap AI Transcripts compare against other async video tools with AI summaries

Cap vs Loom (Loom Alternative)

Where AI runs: Loom hosts its AI and gates it to paid plans. Cap calls the provider key you set on cap-web.
Storage: recordings sit in your Railway bucket, metadata in MySQL. Loom keeps both.
Setup: you add an AssemblyAI key plus one of Groq, OpenAI or Anthropic. Loom wins on zero setup.

Cap vs Tella (Tella Alternative)

Transcripts: Tella builds transcription in. Cap needs ASSEMBLY_API_KEY on cap-web first.
Provider choice: you pick Groq, OpenAI or Anthropic and swap later. Tella chooses for you.
Polish: Tella's editor is smoother out of the box. Cap wins on ownership and data location.

Cap vs Vidyard (Vidyard Alternative)

Focus: Vidyard leans into sales outreach and viewer analytics. Cap is a recorder you host.
AI outputs: titles, summaries and chapters come from your provider key, billed to you.
Compliance: media stays in your bucket, but audio and text go to your AI provider. Vidyard wins on sales workflows.

Cap vs Zight (Zight Alternative)

Scope: Zight covers screenshots, GIFs and clips. Cap records screens with transcripts and AI summaries.
Search: once AssemblyAI returns text, share pages carry searchable transcripts.
Lock-in: Cap is AGPL-3.0 and self hosted, so your data leaves with you. Zight wins on quick capture.

How to use Cap AI Transcripts (the OSS open-source screen recorder)?

Open your Railway domain, enter any email and paste the 6-digit code from the cap-web logs.
Get an AssemblyAI key for transcripts, then add ASSEMBLY_API_KEY on cap-web.
For AI titles, summaries and chapters add GROQ_API_KEY, OPENAI_API_KEY or ANTHROPIC_API_KEY. Without AI_PROVIDER, generation prefers Groq, then OpenAI, then Anthropic.
Redeploy cap-web to load the new keys.

ASSEMBLY_API_KEY=your-key
GROQ_API_KEY=your-key
OPENAI_API_KEY=your-key
ANTHROPIC_API_KEY=your-key

In Cap Desktop set the Cap Server URL to your domain, sign in, record a short clip.
Open the share page and wait for the transcript, then the summary.
Remove a key and redeploy to switch that feature off.

How to self host Cap AI Transcripts on other VPS Services (Cap self hosting guide)

Run upstream's docker-compose.yml yourself. You supply the host, AI keys and backups.

Clone the Repository

Clone the repo; docker-compose.yml sits at the root.

git clone https://github.com/CapSoftware/Cap.git
cd Cap

Install Dependencies

Install Docker with Compose and enough RAM for cap-web, media-server, MySQL and MinIO. About 850 MB sits idle before recordings.

Configure Environment Variables

Set WEB_URL to https://your-domain, generate NEXTAUTH_SECRET and DATABASE_ENCRYPTION_KEY, and replace upstream's default S3 credentials. Add ASSEMBLY_API_KEY for transcripts and GROQ_API_KEY or OPENAI_API_KEY for titles and summaries. Use the newer AI_PROVIDER options only if your image supports them.

Start the Cap AI Transcripts Application

Run docker compose up -d, check /api/status returns 200, then point Cap Desktop at your domain. Read the login code from cap-web logs until Resend is set.

Railway runs the same stack and handles the bucket, private networking, CORS and health checks for you.

Official Pricing of Cap AI Transcripts (Cap pricing)

Cap itself is free to self host, and self-hosted Cap treats every user as Pro: no 5-minute limit, no billing. Your AI provider bills you separately for transcripts and summaries. The template adds no markup and sets no AI keys; AssemblyAI, Groq, OpenAI or Anthropic bill you directly.

Cap AI Transcripts cloud vs self hosted comparison (Pricing, features, costs, and more)

Cap cloud is run by the Cap team. This template is the same app on your Railway account with your bucket, your MySQL and your AI keys, so the bill splits between infra and the provider you pick.

Monthly cost of self hosting Cap AI Transcripts on Railway

Expect about $10-20/mo for a small team at idle around 850 MB RAM, plus bucket storage at $0.015/GB-month with free egress, plus provider charges on your own keys. Railway starts with a $5 trial.

System Requirements for Hosting Cap AI Transcripts on a VPS

Budget 2 GB RAM of headroom: idle is about 490 MB for cap-web, 220 MB for MySQL and 130 MB for media-server, and media-server grows during long recordings. You need Docker, MySQL, an S3 bucket and replaced secrets; upstream compose ships known defaults.

Frequently Asked Questions (FAQs)

Which key turns on transcripts?

ASSEMBLY_API_KEY on cap-web. Add it, redeploy, and Cap transcribes recordings after upload. No key, no transcript, and no AI summary either.

Which provider writes my summaries?

Without AI_PROVIDER, Cap auto-detects from the keys present. Generation features like titles, summaries and chapters try Groq, then OpenAI, then Anthropic. Chat prefers Anthropic, then OpenAI, then Groq.

Does audio leave my server?

Yes. Recordings stay in your Railway Bucket and metadata in MySQL, but the audio and the transcript text go to whichever provider holds the key you added.

Can I point it at a local model?

Not on this pinned 2026-09-07 build. Newer upstream docs add AI_PROVIDER=openai-compatible with AI_BASE_URL, AI_MODEL and AI_API_KEY, so check the image first.

What does the AI side cost?

Cap is free to self host. You pay Railway, about $10-20/mo for a small team, plus provider usage from AssemblyAI and your AI vendor. The $5 trial covers the first look.

How do I turn AI off?

Remove the keys from cap-web and redeploy. Transcripts and generated titles, summaries and chapters stop; recording, upload and playback keep working as before.