Most teams with raster well logs should use AI digitizing with human QC. It is the best way to digitize well logs in 2026 when you have more than a handful to convert: cost per log is lowest, and every curve is still checked by a person before it becomes LAS. If the wells are public and already digitized, buying the LAS from a data vendor or log library is faster still. Below are all five options, the criteria we ranked them on, and which one fits your situation.
Key Takeaways
- AI digitizing with human QC ranks first for most teams. It handles hundreds or thousands of logs at a per-log price, and a person approves every curve against the source scan. NthDS is $35 per log for logs you already have.
- Don't digitize what you can buy. For public wells, a data vendor or log library may already have the LAS. Buying it by the well is cheaper and faster than converting the raster yourself.
- Desktop software costs staff time. Tracing a log by hand takes 4 to 6 hours. That works for a few logs a month and doesn't scale to a data room.
- Scanning is not digitizing. A TIFF is a picture of the log. You can look at it, but you can't compute with it until the curves are converted to LAS.
What well log digitizing actually is (and what it is not)
Well log digitizing turns a raster image of a log into numbers. The input is a scan of the paper log, usually a TIFF or PDF, sometimes a JPEG from a phone. The output is a LAS file: curve values by depth, with a header for the well name, API number, operator, and curve units. Petrel, Kingdom, Techlog, and every other interpretation package load LAS directly.
Digitizing is not scanning. Scanning paper to TIFF is a necessary first step for paper logs, but a TIFF is still a picture. You can view it and correlate by eye, but you can't run a cutoff, build a crossplot, or map a property from it.
Digitizing is also not generic OCR. OCR reads printed characters. It reads the header and the depth numbers, but not the curves, which are lines, not text. Curve tracing needs models built for log images.
The same distinction applies to mud logs and other well file tables. Mud logs come back with mud weights tied to depth, and their hydrocarbon indicators and annotations can be extracted in full, but that is a separate job from wireline curve digitizing and is often priced separately.
How Well Log Digitizing Works, Step by Step
Whether a person or a model does the tracing, digitizing a log takes the same six steps.
- Scan and register depth. Paper and film logs are scanned to TIFF or PDF. Depth registration ties pixel rows to the printed depth labels and sets the vertical scale for each section, since one print often carries a correlation section and a 5-inch detail section at different scales.
- Capture the header. Well name, API or UWI number, operator, KB or DF elevation, run date, logging company, and each curve's name, units and scale go into the LAS header. A wrong API number or KB throws off every tie downstream, so the header is checked before the curves.
- Trace each curve. Find the tracks and the grid, then follow each curve: SP, gamma ray, resistivity, density, neutron, sonic. The hard parts are curves that wrap past the edge of a track onto a backup scale, curves that cross each other, and faded or smudged sections. Traced by hand in desktop software such as NeuraLog, a log takes 4 to 6 hours. Nspect does the tracing in minutes.
- Convert to engineering units. Pixel positions become values using each track's scale, linear or logarithmic (resistivity usually runs on a log scale), with a null value wherever the trace is missing.
- QC against the source scan. Overlay every digitized curve on the raster and check values against expected ranges for depth shifts, spikes, missed wraps and lost sections. With NthDS, a person approves each curve before export.
- Export LAS. Write LAS 2.0 with the header and curve values by depth, or CSV. Where a well was logged in several runs, splice them into one curve per measurement before loading it into your interpretation package.
How We Ranked These Options
We compared five ways to get LAS from raster logs. Each was scored on the same five criteria, all measurable:
- Cost per log at a 1,000-log job. All-in cost, including staff hours for methods you run yourself.
- Turnaround for 1,000 logs. Calendar time from upload to finished LAS.
- QC before delivery. Whether each curve is compared to the source image and approved by a person, and whether the review is recorded.
- Output. LAS 2.0 with header metadata (API, well name, operator, curve mnemonics and units), ready to load without rework.
- Data control. Where your scans go during the work: your network, a vendor cloud, or a third party.
Disclosure: NthDS sells options 1 and 2. We kept the criteria objective so you can apply them to any vendor you are evaluating, including us.
The Best Ways to Digitize Well Logs in 2026
These are the five approaches buyers actually choose between. They are ranked by how well each scores across the five criteria for a typical job of hundreds to thousands of logs. Your volume, budget, and security rules may change the order, so the use-case table further down maps each one to a buyer.
1. AI digitizing with human QC, best for hundreds to thousands of logs on a deadline
Models find the tracks and depth scale, trace each curve, and convert it to engineering units. Then the digitized curves are overlaid on the source scan for review. Breaks and wraps get fixed, and a person approves each curve before export. The AI does the tracing; your geologist keeps the judgment calls.
With NthDS, scans go in as PDF, TIFF, JPEG, or PNG, one file or 100+ at a time. Rotated, faded, and photocopied pages are accepted. Output is LAS 2.0 with header metadata, or CSV.
- Price: $35 per log for logs you already have (mud logs may cost more). Self-serve accounts start with $35 of free credit, enough for the first log, no card required.
- Turnaround: Nspect traces a log in minutes; batches, including QC, are quoted by volume and scan quality. On one project, NthDS delivered about 2,500 gamma ray curves in LAS in a single week.
- QC: curve overlay on the source image, range checks on values, and human approval before export.
- Data control: digitizing runs in the NthDS cloud.
Start self-serve with a free account, or request a demo for a large batch.
Best suited for: operators, A&D teams, and consultants with more than a few dozen logs and a date to hit. It is also the right choice when you need every curve checked but can't spend staff hours tracing.
2. Buy LAS from a data vendor or log library, best for public wells you don't hold logs for
Many public wells have already been digitized and sit with a data vendor or in a log library. If you need offset wells, a regional study, or logs from wells you never operated, buying the LAS beats digitizing it. There is no raster to find and nothing to trace.
Subscription data vendors sell digitized curves inside annual contracts, which costs more if you only need some wells. Their coverage also has gaps, because regional log libraries hold logs the national vendors don't.
On NDEX, you search 2M indexed wells and buy available LAS one well at a time, with no subscription. LAS availability and delivery vary by well. If a well you need has only a raster, digitize it with option 1.
- Price: per well on NDEX. See NDEX pricing. Subscription data vendors usually require an annual contract.
- Turnaround: LAS is made on demand from the well's raster logs; delivery varies by well.
- QC: check the curve set and depth range in the listing before you buy, then review as you would any LAS.
- Data control: nothing leaves your network, because you're downloading, not uploading.
Best suited for: geologists, landmen, and analysts who need public offset wells or a quick regional set. If the wells you need aren't digitized yet, use option 1 on the rasters.
3. Desktop digitizing software, best for a few logs a month with a trained tech
Desktop software puts the raster on screen, and a technician calibrates the grid and depth scale, then traces each curve, with point-and-click or semi-automatic tools. The result can be very good, because a trained person is looking at every inch of the log.
The cost is time. Tracing a full log by hand takes 4 to 6 hours. At 1,000 logs, that is 4,000 to 6,000 staff hours before QC, on top of license and seat costs. Quality also depends on who does the tracing and how tired they are by the 50th log.
- Price: software license plus staff time, usually the largest part.
- Turnaround: limited by headcount. One person does a log or two a day.
- QC: whatever your team builds. Often the same person who traced the curve checks it.
- Data control: stays on your workstation.
Best suited for: teams with a small, steady flow of logs and a technician already trained on the software. It also works for a single critical well where you want your own person to trace every inch.
4. Manual digitizing service, best for small one-off jobs with no deadline
You ship scans to a service, and their technicians trace the curves by hand and send back LAS. You don't need software or staff, and the vendor handles the labor.
Turnaround is the weak point. Every curve is traced by a person, so large jobs queue behind the vendor's available labor. Pricing is usually quoted per log or per curve, so ask for both before you compare.
- Price: quoted per log or per curve.
- Turnaround: set by the vendor's labor. Ask for a committed date on your volume.
- QC: varies by vendor. Ask for a sample with overlays before you commit.
- Data control: your scans go to the vendor, sometimes offshore. Ask where the work happens.
Best suited for: a small batch with no deadline when you would rather pay per log than learn a tool. For anything with a date attached, compare turnaround against option 1.
5. Scan only and keep the rasters, best when you only need to look at the logs
Scanning paper logs to TIFF or PDF preserves them and makes them shareable. For correlation by eye, a quick look at a show, or an archive you rarely open, images may be enough.
The limit is that you can't compute anything from a picture. No petrophysics, no cutoffs, no crossplots, no mapping. Most teams that start here end up digitizing the logs that matter later, so scan at a resolution good enough to digitize from.
- Price: the lowest of the five.
- Turnaround: fast for scanning alone.
- QC: image quality only.
- Data control: depends on who scans.
Best suited for: archives you are preserving rather than analyzing. Pick out the wells you'll actually interpret and digitize those.
Side-by-Side Comparison
The table scores each option on the five criteria above: cost per log at a 1,000-log job, turnaround, QC before delivery, output, and data control.
| Approach | Cost per log | Turnaround, 1,000 logs | QC before delivery | Output | Where your scans go |
|---|---|---|---|---|---|
| 1. AI digitizing + human QC | $35 (NthDS) | Quoted per batch | Overlay, range checks, human approval | LAS 2.0 + header, CSV | Vendor cloud |
| 2. Buy LAS from a data vendor or log library | Per well (NDEX) or annual subscription | Varies by well | Your normal LAS review | LAS | Nowhere (download) |
| 3. Desktop software | License + 4–6 staff hours | 4,000–6,000 staff hours | Whatever your team builds | LAS | Your workstation |
| 4. Manual service | Quoted per log or curve | Limited by vendor labor | Varies by vendor | LAS | Vendor, sometimes offshore |
| 5. Scan only | Lowest | Fast | Image quality only | TIFF / PDF image | Whoever scans |
Which Approach Fits Your Use Case
| Who you are | Typical job | Best approach | Why |
|---|---|---|---|
| Independent geologist or consultant | 5–50 logs for a prospect | AI digitizing, self-serve | First log free ($35 credit); no software license or install |
| A&D or data room team | 1,000+ wells, fixed close date | AI digitizing with batch QC | Per-log price; turnaround quoted per batch |
| Landman or analyst | Public offset wells | Buy LAS by the well from a data vendor or log library | No raster to find; pay only for the wells you need |
| Team with a trained tech | A few logs a month | Desktop software in-house | Existing seat and skills; volume too low to outsource |
| Archive owner | Boxes of paper logs | Scan, then digitize the wells you'll use | Preserve everything; spend on the logs that matter |
What Results Look Like
These are real NthDS projects. The clients are not named.
- A top exploration company, behind schedule. The incumbent vendor was on pace for 13 to 15 months against a 9-month deadline. NthDS classified about 100,000 documents in a week, extracted about 2,500 gamma ray curves to LAS in a week, finished 30,000 pages of tables across 3,500 wells in 45 days, and came in about 30% under projected spend. Read the case study.
- A large U.S. independent with 130,000+ SEG-Y volumes. Crawler built one inventory of the company's seismic data across its operating divisions and keeps it current against vendor transmittals. Read the case study.
- A large operator bidding in a federal offshore lease sale. Crawler read decades of public BOEM and BSEE mud logs, pulled shows and zones of interest by depth, and ranked the auction blocks by hydrocarbon indicators. Read the case study.
FAQ
What are the steps to digitize a well log?
Six steps: scan the log and register depth, capture the header, trace each curve, convert pixel positions to engineering units, QC every curve against the source scan, and export LAS 2.0 or CSV. Tracing is the slow step by hand, 4 to 6 hours a log; Nspect does it in minutes, and a person approves each curve before export.
How much does it cost to digitize a well log?
It depends on the method. NthDS AI digitizing is $35 per log for logs you already have, and new accounts get $35 of free credit, so the first log is free. Desktop software costs a license plus several hours of staff time per log, and manual services quote per log or per curve.
How long does it take to digitize a well log?
Tracing a log by hand in desktop software takes 4 to 6 hours. Nspect does the tracing in minutes; we quote full batches, including QC, by volume and scan quality. On one project, NthDS delivered about 2,500 gamma ray curves in LAS in a single week.
What is the difference between a raster log and a LAS file?
A raster log is a picture of the paper log, usually a TIFF or PDF. A LAS file holds the curve values by depth as numbers, so Petrel, Kingdom, or Techlog can load, plot, and compute with them.
What file formats can be digitized?
Scanned logs in PDF, TIFF, JPEG, or PNG, including multi-page, rotated, and faded scans. Output is LAS 2.0 with header metadata, or CSV.
How accurate is AI well log digitizing?
Accuracy depends on scan quality, so the QC step matters more than the headline number. NthDS overlays every digitized curve on the source scan, checks values against expected ranges, and a person approves the curve before export.
Can mud logs be digitized too?
Yes. We deliver digitized mud logs with mud weights tied to depth, and we offer full extraction of the hydrocarbon indicators and annotations on each log. Mud logs can cost more than wireline logs because there is more on each page to read.
Where can I get LAS for wells I don't have logs for?
From a data vendor or log library. Subscription data vendors sell LAS inside annual contracts. NDEX lets you search 2M indexed wells and buy available LAS one well at a time, with no subscription. LAS availability and delivery vary by well.
Start With One of Your Own Logs
Upload a raster log and check every curve against the original scan before you commit to a batch. Digitizing logs you already have is $35 per log.
- A few logs: create a free account and upload them yourself. New accounts get $35 of free credit, so the first log is free, and no card is required.
- Hundreds or thousands of logs on a deadline: book a 15-minute demo. We'll quote the batch and run a sample first.