Serverless video rendering vs AWS Lambda
Compare PennyCompute output pricing with AWS Lambda compute, request and storage billing, without inventing a universal video-minute rate.
Open cloud renderingTwo different bills
PennyCompute is a prepaid job API: you upload sources or a Remotion bundle, submit a service ID, and pay when the job succeeds. AWS Lambda is general-purpose compute: you (or a renderer such as Remotion Lambda) run functions, and AWS bills the time and memory those functions use, plus a charge per request. A video job on Lambda usually also needs object storage, logs and egress, which appear on other AWS lines. The useful comparison is the operating model and the published rate units, not a single made-up “Lambda minute of video” price.
PennyCompute’s live output rates
Fetched on 22 November 2026: https://pennycompute.com/ffmpeg-api-pricing lists FFmpeg at $0.0075 per minute of HD output, $0.00375 for SD, $0.015 for 4K and $0.0015 for audio only, billed by the second, with a 10-second minimum per output. The FFmpeg rate follows the output’s shorter side: SD below 720 lines, HD up to 1080 and 4K above. https://pennycompute.com/pricing lists Remotion at $0.0037 per render plus $0.0048 per minute of output up to 1080p, or $0.017 per minute above 1080p, billed by the second. Only a successful job posts a charge. Failed, cancelled and expired jobs, and time in the queue, cost nothing. A new account starts with $5.
What AWS Lambda’s public price list covers
The AWS Lambda pricing page (https://aws.amazon.com/lambda/pricing/, accessed 22 November 2026) describes on-demand charges for requests and for compute in GB-seconds, with a monthly free tier that the page states as one million requests and 400,000 GB-seconds. Worked examples on that page use $0.20 per one million requests and $0.0000166667 per GB-second for a 1024 MB x86 function. A compute example on the same page uses $0.0000133334 per GB-second. Those are compute and request figures for a function you configure. They are not a price per minute of finished video, and they do not include S3, CloudWatch, data transfer or any application license.
Remotion on Lambda, as Remotion publishes it
If the workload is a Remotion composition, Remotion’s own Lambda cost examples are the public figures to read first: https://www.remotion.dev/docs/lambda/cost-example (accessed 22 November 2026). Under the configuration Remotion lists (2048 MB RAM, 10240 MB disk, us-east-1, Remotion 4.0.381), a one-minute video in the same S3 bucket is $0.017 warm or $0.021 cold, and a ten-minute remote HD video is $0.103 warm or $0.108 cold, Lambda compute only. On PennyCompute, the same output lengths at the live up-to-1080p Remotion rate are $0.0085 for one minute and $0.0517 for ten minutes. These are not the same renders: Remotion’s examples are measured compositions on AWS; PennyCompute’s figures are arithmetic from the rate card. Remotion also notes S3, logs, bandwidth and, for teams of four or more, a company license on top of Lambda compute.
FFmpeg-style transcoding on AWS
A self-built FFmpeg Lambda has no single public “HD output minute” list price, because the bill follows function memory, duration, storage and transfer. AWS also sells AWS Elemental MediaConvert as a managed transcoding service, priced on demand per normalized minute of output, with Basic and Professional tiers and feature multipliers, documented at https://aws.amazon.com/mediaconvert/pricing/ (accessed 22 November 2026). That page does not reduce to one HD rate beside $0.0075 without reproducing its multiplier table, so this page does not invent one. If you already use MediaConvert, compare your actual invoice to PennyCompute’s output-minute card rather than to a shorthand rate.
What you operate in each model
On Lambda you provision functions, set memory and ephemeral storage, watch concurrency limits and timeouts, ship the renderer, and wire storage and IAM. Remotion Lambda automates much of that for Remotion projects, still inside your AWS account. On PennyCompute you keep an API key and a prepaid balance, submit ffmpeg.* or remotion.render.v1 jobs, and retrieve the file when the job succeeds, or get a webhook. You do not configure Lambda memory or an AWS region for that job. Capacity for a given service is visible at GET /v1/services/{id}/capacity; a refused or failed job is not charged.
When a job API is enough, and when Lambda is the better fit
A job API is usually enough when you want FFmpeg operations or a Remotion composition rendered as a finished file, priced by output, without running AWS for that step. Lambda (or Remotion Lambda) is a better fit when the render must live in your AWS account, when you need Remotion’s distributed Lambda renderer, or when the work is not a single output file. Confirm live PennyCompute rates on /pricing before you budget, and measure a real composition before you compare a production Lambda bill with a rate card.
Frequently asked questions
What does PennyCompute charge for serverless video rendering?
On 22 November 2026, https://pennycompute.com/ffmpeg-api-pricing lists FFmpeg at $0.0075 per minute of HD output, $0.00375 for SD, $0.015 for 4K and $0.0015 for audio only, billed by the second with a 10-second minimum per output. https://pennycompute.com/pricing lists Remotion at $0.0037 per render plus $0.0048 per minute of output up to 1080p, or $0.017 per minute above 1080p, billed by the second. Failed, cancelled and expired jobs cost nothing.
How does AWS Lambda bill a render?
AWS Lambda on-demand billing, documented at https://aws.amazon.com/lambda/pricing/ (accessed 22 November 2026), charges for requests and for compute in GB-seconds. Worked examples on that page use $0.20 per one million requests and $0.0000166667 per GB-second for an x86 function, and $0.0000133334 per GB-second appears in a compute example. A video render also typically incurs storage, logs and data transfer, which are billed by other AWS services.
Can I compare a PennyCompute minute to a Lambda GB-second directly?
Not as a single number that applies to every composition. PennyCompute prices a completed output by its length and frame size. Lambda prices the memory and duration of the functions that did the work, which vary with the composition, concurrency and cold starts. Remotion publishes its own Lambda compute examples for named renders; those are the fairest public figures to put beside PennyCompute’s Remotion rate card, with the exclusions Remotion lists.