Model reference — Live feed
Live OCR reads text off a camera as it appears — vehicle plates, container and wagon numbers, labels on passing goods, equipment displays, badges and shipping documents held up to a lens — and acts on it immediately.
The engineering trick is to read as little as possible. Reading every frame in full is unaffordable and unnecessary, because the same text is visible for many frames. Instead a light detector finds text regions continuously, the region is tracked, and only the clearest frame of each occurrence is read. That makes live reading both cheaper and more accurate than reading every frame, and it is what allows several cameras to run on one card. Where the text is drawn from a known list — your own fleet, your own stock — checking the reading against that list corrects most residual errors.
Models in this group take a live camera or stream as input and must keep pace with it in real time. Each specification table gives three hardware tiers — Minimum, the smallest setup on which the model runs correctly; Medium, the usual production configuration; and High, a configuration sized for peak volume. The rate is how many camera streams or feeds one server of that tier can keep up with in real time, not a per-hour count: live work must fit inside the interval between frames, and a server that cannot keep pace drops frames rather than falling behind. Use these rates for initial sizing. Before production, benchmark your own data to validate accuracy, latency, throughput and cost. The sample input here is one 1080p stream at 25 frames per second.
Vendor: Baidu (PaddlePaddle)
What it does: reads text quickly enough to keep up with a live stream, tolerating angle and modest resolution. The usual live reader.
| Requirement | Minimum | Medium | High |
|---|---|---|---|
| GPU type | RTX 3060 | RTX 4090 | A100 80 GB |
| VRAM | 6 GB | 24 GB | 80 GB |
| vCPUs | 4 | 8 | 24 |
| RAM | 16 GB | 32 GB | 64 GB |
| Server | 1× RTX 3060 12 GB | 1× RTX 4090 24 GB | 1× A100 SXM 80 GB |
| Rate (camera streams handled at 25 frames per second) | ≈ 6 streams | ≈ 22 streams | ≈ 60 streams |
Vendor: Jaided AI
What it does: a light reader covering eighty languages, adequate where text is large and steady, such as a document held to a fixed camera.
| Requirement | Minimum | Medium | High |
|---|---|---|---|
| GPU type | RTX 3060 | RTX 4090 | A100 80 GB |
| VRAM | 6 GB | 24 GB | 80 GB |
| vCPUs | 4 | 8 | 24 |
| RAM | 16 GB | 32 GB | 64 GB |
| Server | 1× RTX 3060 12 GB | 1× RTX 4090 24 GB | 1× A100 SXM 80 GB |
| Rate (camera streams handled at 25 frames per second) | ≈ 8 streams | ≈ 28 streams | ≈ 80 streams |
Vendor: Ultralytics
What it does: finds where text is on each frame so only those regions are read. This is the component that makes live reading affordable, and it runs alongside other detection on the same stream.
| Requirement | Minimum | Medium | High |
|---|---|---|---|
| GPU type | RTX 3060 | RTX 4090 | A100 80 GB |
| VRAM | 6 GB | 24 GB | 80 GB |
| vCPUs | 4 | 8 | 24 |
| RAM | 16 GB | 32 GB | 64 GB |
| Server | 1× RTX 3060 12 GB | 1× RTX 4090 24 GB | 1× A100 SXM 80 GB |
| Rate (camera streams handled at 25 frames per second) | ≈ 12 streams | ≈ 40 streams | ≈ 110 streams |
Vendor: Microsoft
What it does: reads the difficult individual regions — stylised, handwritten, low contrast — that the fast reader returns with low confidence.
| Requirement | Minimum | Medium | High |
|---|---|---|---|
| GPU type | RTX 3060 | RTX 4090 | A100 80 GB |
| VRAM | 8 GB | 24 GB | 80 GB |
| vCPUs | 6 | 12 | 24 |
| RAM | 16 GB | 32 GB | 64 GB |
| Server | 1× RTX 3060 12 GB | 1× RTX 4090 24 GB | 1× A100 SXM 80 GB |
| Rate (camera streams handled at 25 frames per second) | ≈ 4 streams | ≈ 14 streams | ≈ 40 streams |
Vendor: StepFun
What it does: reads structured text such as equipment displays and printed tables held to the camera, returning layout as well as characters.
| Requirement | Minimum | Medium | High |
|---|---|---|---|
| GPU type | RTX 3060 | RTX 4090 | A100 80 GB |
| VRAM | 8 GB | 24 GB | 80 GB |
| vCPUs | 6 | 12 | 24 |
| RAM | 16 GB | 32 GB | 64 GB |
| Server | 1× RTX 3060 12 GB | 1× RTX 4090 24 GB | 1× A100 SXM 80 GB |
| Rate (camera streams handled at 25 frames per second) | ≈ 2 streams | ≈ 8 streams | ≈ 20 streams |
Vendor: Microsoft
What it does: locates and reads text without being fitted first, useful for standing up a live reading trial before committing to a tuned pipeline.
| Requirement | Minimum | Medium | High |
|---|---|---|---|
| GPU type | RTX 3060 | RTX 4090 | A100 80 GB |
| VRAM | 8 GB | 24 GB | 80 GB |
| vCPUs | 6 | 12 | 24 |
| RAM | 16 GB | 32 GB | 64 GB |
| Server | 1× RTX 3060 12 GB | 1× RTX 4090 24 GB | 1× A100 SXM 80 GB |
| Rate (camera streams handled at 25 frames per second) | ≈ 3 streams | ≈ 10 streams | ≈ 28 streams |
What is readable live depends on resolution, speed of motion, lighting and how long text stays in view. Our consultants review your camera positions and the text you need captured, then recommend the detection and reading models and the validation against your own data.
At the start of a project we may run a short proof of concept on a sample of your own data, measuring the accuracy and the throughput the model actually achieves on your material. That replaces the estimates on this page with real figures, so the cost and the schedule for the full engagement are known before it is committed.
Share a representative sample, expected volume, latency target and deployment location for Live OCR. We benchmark the shortlisted models, recommend the lowest-cost GPU configuration that meets the target, and scale it from pilot capacity to a dedicated production cluster.