SoraClip Backend Platform and Cost Comparison
Updated: September 29, 2026.
Estimate basis: Singapore, US dollars, public pay-as-you-go prices. Taxes, exchange-rate differences, and enterprise discounts are excluded.
1. Purpose
Evaluate Render and Microsoft Azure for hosting SoraClip's usage and cost management services. This report covers required services and monthly fees to support platform selection and operating budgets.
This report compares the running costs of the backend that records usage and calculates costs, including authorization, the database, persistent connections, scheduled jobs, and monitoring. It excludes model fees for translation and the artificial intelligence assistant. Both reports use the same operating assumptions; feature usage costs are covered in Usage Management and Cost Estimates.
2. Implementation scope
| Work item | Description |
|---|---|
| Usage authorization | Verify the app user's login status and eligibility, and issue short-lived temporary credentials; the backend holds production service keys |
| Usage records | Store translation and assistant usage for each user |
| Cost statistics | Calculate costs using provider rates and summarize them by user, feature, and month |
| Reconciliation | Regularly reconcile provider usage and charges, and update items awaiting confirmation |
| Subscription integration | Reserve quota and billing-cycle information for future subscription plans |
Payment, refund, and subscription billing flows are outside this phase. The app connects directly to speech and model providers; the backend does not relay or store complete audio.
3. Required services
| Service category | Purpose | Render configuration | Microsoft Azure configuration |
|---|---|---|---|
| Application service | Handle app authorization requests and continuously receive model usage events | Web Service | App Service |
| Database service | Store users, usage, rates, costs, and quotas | Render Postgres | Azure Database for PostgreSQL Flexible Server |
| Scheduled jobs | Regularly retrieve usage, reconcile charges, and update statistics | Cron Job | WebJobs |
| Monitoring service | Track service health, runtime errors, and resource usage, and send alerts | Built-in monitoring, logs, and event notifications | Azure Monitor, Application Insights |
PostgreSQL is a relational database for storing and querying structured data. Cron Job and WebJobs can run backend tasks on a schedule; Application Insights tracks application performance and errors.
The application service continuously receives OpenAI usage through a sideband connection, a separate backend connection. Azure's scheduled jobs also share the application service's resources. Render service types, Azure App Service, Azure WebJobs
4. Platform comparison
4.1 Management and operation
| Comparison item | Render | Microsoft Azure |
|---|---|---|
| Host maintenance | The platform manages hosts, operating systems, and supported runtimes | The platform manages hosts, operating systems, and supported runtimes |
| Service management | Applications, databases, scheduled jobs, and deployments are managed in one workspace | Applications, databases, monitoring, and permissions are configured as separate resources and can fit into the company's existing Azure management structure |
| Continuous operation | A paid Web Service does not sleep when idle | Enable Always On in App Service to keep the application running |
| Team management | This estimate uses a Pro workspace for two or more administrators | Resource permissions are assigned by role; basic permission management has no additional seat fee |
| Runtime logs | Pro includes 14 days of log retention | Standard retention is 31 days for Analytics Logs and 90 days for Application Insights data |
| Database recovery | Paid databases on Pro support recovery to a point within the past 7 days | Automatic backups are retained for 7 days by default, configurable up to 35 days |
| Alerts | The platform provides service, deployment, and scheduled-job event notifications | Configure monitoring rules and notifications through Azure Monitor |
| Deployment region for this estimate | Singapore; application and database in the same region | Singapore; application and database in the same region |
Point-in-time recovery restores a database to a specified time within the backup retention period. The available recovery window and long-term archiving are separate service requirements.
Sources: Render shared responsibility, workspace plans, database backups, Azure shared responsibility, permission management, monitoring pricing and retention, database backups.
4.2 Idle sleep
Idle sleep stops an application after a period without requests to save computing resources. The application restarts when the next request arrives. This startup delay affects temporary credential issuance and can delay the start of translation or assistant features.
Render's free Web Service sleeps after 15 consecutive minutes without inbound traffic and typically takes about 1 minute to restart. This estimate therefore uses paid compute. Azure uses a plan that supports Always On, with the setting enabled to prevent the application from unloading when idle. Render free service rules, Azure Always On
Continuous-operation settings prevent idle sleep. Brief interruptions during deployments, maintenance, and incidents still require service recovery mechanisms.
5. Service usage and monthly cost estimates
5.1 Shared assumptions
This section compares 100, 500, and 1,000 monthly active users with the same usage per person: 30 translation sessions per month at 10 minutes each, and 10 Agent conversations at 5 minutes each, with 10 responses per conversation. These assumptions match Usage Management and Cost Estimates. The figures below describe backend workload only and exclude translation and model fees.
| Item | 100 users | 500 users | 1,000 users |
|---|---|---|---|
| Translation sessions per month | 3,000 | 15,000 | 30,000 |
| Agent conversations per month | 1,000 | 5,000 | 10,000 |
| Feature authorization requests per month | 4,000 | 20,000 | 40,000 |
| Usage records per month | 14,000 | 70,000 | 140,000 |
| Total sideband connection hours per month | 83.33 | 416.67 | 833.33 |
| Peak simultaneous sideband connections | 10 | 50 | 100 |
| Outbound transfer allowance per month | 5 GB | 25 GB | 50 GB |
| Runtime log allowance per month | 1 GB | 5 GB | 10 GB |
| Estimated new usage records and indexes per month | 0.056 GB | 0.28 GB | 0.56 GB |
| Estimated 12-month data plus allowance for other data | 1.672 GB | 4.36 GB | 7.72 GB |
All three scenarios use 2 company administrators, a single region, 1 application instance, and 1 database, without high-availability redundancy. The application and database run throughout the month; hourly charges assume 730 hours per month. Render databases are configured with 10 GB, and Azure databases with 32 GiB. Scheduled jobs run once an hour with 5 minutes reserved per run, totaling 3,600 minutes over a 30-day month.
Authorization is estimated at one request per feature activation. Each user generates 40 Soniox usage records per month (30 translation sessions and speech recognition for 10 Agent conversations) and 100 OpenAI response settlement records, totaling 140 records. Records contain usage and cost metadata only, without complete transcripts, audio, or individual streaming chunks. Additional volume from retries, reconnections, extra settlements, and credential renewals will be adjusted using actual records after launch.
Storage is provisionally estimated at 4,000 bytes per usage record including indexes, accumulated over 12 months, plus 1 GB for accounts, authorizations, rates, and summaries. For example, the 1,000-user scenario is 140,000 records × 4,000 bytes × 12 months + 1 GB = 7.72 GB. The 12-month period is a capacity budgeting basis, not a decision to delete data after one year. The retention policy and subsequent capacity still need to be confirmed.
Connection hours add together the duration of every connection; they are neither server runtime hours nor an additional billing unit. Sideband remains connected during Agent conversations. Translation connects directly to Soniox and does not use sideband, so longer translation sessions do not directly add sideband hours.
Traffic and log volumes are estimated allowances, not measured usage or fixed quotas per user. Usage frequency, duration, and data volumes will be adjusted using records after launch. Peak connection capacity in each scenario must be verified by load testing before launch; these estimates are not a capacity or uninterrupted-service commitment. Application and database compute specifications do not scale in direct proportion to user count.
GB (Gigabyte) is a decimal capacity unit; GiB (Gibibyte) is a binary capacity unit. 1 GiB is approximately 1.074 GB. The specifications below follow each platform's stated units.
5.2 Render
| Service / fee | Configuration and billing | Monthly fee: 100 users | Monthly fee: 500 users | Monthly fee: 1,000 users |
|---|---|---|---|---|
| Team workspace | Pro, 2 administrators; $25 per workspace/month | $25.00 | $25.00 | $25.00 |
| Application service | 1 CPU core, 2 GB memory, 1 instance; $25/month | $25.00 | $25.00 | $25.00 |
| Database compute | 0.5 CPU core, 1 GB memory, 1 instance; $19/month | $19.00 | $19.00 | $19.00 |
| Database storage | 10 GB provisioned; 10 × $0.30/GB/month | $3.00 | $3.00 | $3.00 |
| Scheduled jobs | 1 Cron Job, 3,600 minutes reserved; $0.00016/minute, minimum $1 per job/month | $1.00 | $1.00 | $1.00 |
| Outbound transfer | Allowances of 5/25/50 GB; Pro includes 25 GB, then $0.15 per GB | $0.00 | $0.00 | $3.75 |
| Runtime logs | Allowances of 1/5/10 GB, built-in 14-day retention, no separate per-GB fee | $0.00 | $0.00 | $0.00 |
| Database backups | 7-day point-in-time recovery, included with the paid database plan | $0.00 | $0.00 | $0.00 |
| Monthly total | $73.00 | $73.00 | $76.75 |
CPU (Central Processing Unit) core count and memory capacity are the basis for comparing compute specifications in this table.
Render charges for provisioned storage. All three scenarios use 10 GB to cover estimated first-year data and reserved capacity. Pro has a fixed monthly workspace fee, not a separate charge per administrator. Calculated scheduled-job compute costs are $0.576, so the $1 monthly minimum applies. The fixed configuration totals $73; the 1,000-user scenario adds $3.75 for 25 GB of excess transfer. Official pricing, compute specifications, workspace pricing, scheduled-job pricing, outbound transfer pricing
5.3 Microsoft Azure
| Service / fee | Configuration and billing | Monthly fee: 100 users | Monthly fee: 500 users | Monthly fee: 1,000 users |
|---|---|---|---|---|
| Team management | 2 administrators, standard role permissions, no additional charge | $0.00 | $0.00 | $0.00 |
| Application service | Linux App Service Basic B1, 1 core, 1.75 GB memory; $0.018/hour × 730 | $13.14 | $13.14 | $13.14 |
| Database compute | PostgreSQL General Purpose D2ds_v5, 2 virtual cores, 8 GiB memory; $0.244/hour × 730 | $178.12 | $178.12 | $178.12 |
| Database storage | 32 GiB provisioned; 32 × $0.138/GiB/month | $4.42 | $4.42 | $4.42 |
| Scheduled jobs | WebJobs, 3,600 minutes reserved, sharing App Service resources | $0.00 | $0.00 | $0.00 |
| Outbound transfer | Allowances of 5/25/50 GB, using the first 100 GB free allowance | $0.00 | $0.00 | $0.00 |
| Log ingestion | Allowances of 1/5/10 GB of Analytics Logs; first 5 GB free per billing account, then $2.99 per GB | $0.00 | $0.00 | $14.95 |
| Database backups | Non-geo-redundant backups, with capacity within the provisioned 32 GiB | $0.00 | $0.00 | $0.00 |
| Monthly total | $195.68 | $195.68 | $210.63 |
The database uses the General Purpose tier rather than Burstable, which allows temporary performance bursts and is not recommended by the provider for production. The minimum database storage configuration is 32 GiB. Charges depend on provisioned capacity and do not decrease with lower actual data volume. Database specifications and use cases
Compute and storage use Singapore's public pay-as-you-go prices. The exact storage fee is $4.416; table amounts are rounded to two decimal places. Official unit rates: App Service B1, PostgreSQL D2ds_v5, database storage.
WebJobs uses existing App Service resources without provisioning another instance. App Service plan charges
The exact monthly cost of the fixed configuration is $195.676. The 1,000-user scenario adds $14.95 for 5 GB of log ingestion. The monitoring free allowance is shared by the billing account; all three scenarios assume the full 5 GB remains available. If other services have used it, backend monthly fees for 100/500/1,000 users are approximately $198.67/$210.63/$225.58. Outbound transfer also assumes the free allowance remains available. Azure Monitor pricing, Singapore log ingestion rate, outbound transfer pricing
5.4 Cost differences and additional charges
| Monthly active users | Render platform/month | Render average/user/month | Azure platform/month | Azure average/user/month |
|---|---|---|---|---|
| 100 users | $73.00 | $0.730 | $195.68 | $1.957 |
| 500 users | $73.00 | $0.146 | $195.68 | $0.391 |
| 1,000 users | $76.75 | $0.077 | $210.63 | $0.211 |
Average platform cost per user is the unrounded backend monthly fee divided by monthly active users, shown to three decimal places. It excludes translation and model fees and is not the subscription price charged to users. With a fixed configuration, average cost decreases as user count rises. If load testing or actual operation indicates an upgrade is needed, the estimate must be revised.
| Fixed configuration comparison | Render | Microsoft Azure |
|---|---|---|
| Database compute and storage/month | $22.00 | $182.54 |
| Application memory | 2 GB | 1.75 GB |
| Database memory | 1 GB | 8 GiB |
| Provisioned database storage | 10 GB | 32 GiB |
Render costs approximately $122.68 less per month for 100 and 500 users, and $133.88 less for 1,000 users. The main difference comes from the starting database specifications: the Azure configuration provides more database compute and storage. These figures therefore compare the configurations available for this project, not unit prices for identical hardware specifications.
The main reasons costs may increase are:
- More capacity or load: Upgrading application compute, database compute, or provisioned storage.
- Excess traffic: Render Pro charges $0.15 per GB above its included allowance; Azure's next tier after free outbound transfer in Asia is $0.12 per GB.
- Longer retention and redundancy: Longer log or backup retention, geo-redundant backups, or additional application or database redundancy.
- Additional environments and support: A separate test environment, paid alert rules, purchased support, or other network services.
With only one administrator, Render could use a Hobby workspace. The same compute, storage, and 50 GB of transfer would cost approximately $54.75/month. Its included transfer allowance is 5 GB, with charges for another 45 GB. Log retention drops to 7 days, and database point-in-time recovery to 3 days. This estimate budgets for Pro with two company administrators. Render plan conditions, transfer charges, Azure bandwidth pricing
All monthly fees above exclude Soniox, OpenAI, domain registration, development and maintenance labor, payment processing, and taxes. Both options are managed services, so company maintenance of Linux hosts does not need to be listed separately.
6. Assessment
Under the fixed configurations and usage assumptions in this report, backend monthly fees for 100/500/1,000 monthly active users are US$73.00/US$73.00/US$76.75 for Render and US$195.68/US$195.68/US$210.63 for Microsoft Azure. Average platform costs per user are in Section 5.4. All exclude translation and model fees.
For this service scope and public pricing, Render Pro, a paid Web Service, Render Postgres, and Cron Job can be evaluated first, with applications, databases, and scheduled jobs managed in one workspace. Microsoft Azure can use the company's existing resource and permission management structure. Dedicated Azure operational support, contract discounts, or centralized procurement requirements should be included in the final choice where applicable. Both configurations require peak connection and data-write load validation before launch; this report is not a capacity guarantee.
Ongoing cost management should review platform bills, model usage, and maintenance hours together. Platform fee differences can be quantified directly. Management effort and maintenance hours need to be assessed from actual operating records; no unverified labor savings are included.