Pilots · launched in October 2026

Eight small AI tools we build and run ourselves

Each tool does one job: you upload a file or paste text, see the price, pay per order and download the result. They run on Claude Haiku 5.5, file text is deleted after 24 hours, and every result needs a human check — these are new pilots without customer reviews yet.

The tools

SRT Rescue
$4 per file
Locale Pack
$5 per order
Recall Pack
$4 per order
Spec Check
$7 per order
Screenshot Review
$9 per order
Catalog Copy
$7 per order
Brief to Diagram
$7 per order
CSV Brief
$5 per order

SRT Rescue

$4 per file

Translates SRT and VTT subtitles between Russian and English. Cue numbers and timestamps are copied from your file, not regenerated.

Tested on 20 October 2026: 28 of 28 test orders completed; in every one, cue IDs and timestamps matched the source byte for byte.

Where it falls short: Russian slang is sometimes translated with the opposite meaning. The built-in check verifies structure, not meaning.

Locale Pack

$5 per order

Translates up to 40 short UI strings into German, Spanish, French or Russian and returns JSON with the same keys; placeholders are checked.

Tested on 20 October 2026: 11 of 11 test orders completed; median time 3.1 seconds.

Where it falls short: Formal and informal address can differ between orders (Sie in one, du in another).

Recall Pack

$4 per order

Turns 1–20 of your own facts into Anki cloze cards (TSV, HTML, JSON). Every answer is a substring of your fact.

Tested on 20 October 2026: 10 of 12 test orders completed; median time 5.4 seconds.

Where it falls short: Occasionally produces a weak card; in one order the model distorted a quote.

Spec Check

$7 per order

Turns up to 8 interface requirements into a draft test matrix with quotes from your source and open questions (CSV, HTML, JSON). Nothing is executed.

Tested on 20 October 2026: 9 of 12 test orders completed; median time 8.4 seconds.

Where it falls short: Coverage of the requirements is not always stable from order to order.

Screenshot Review

$9 per order

Reviews 1–3 interface screenshots and returns up to 5 observations with marked areas (HTML and JSON).

Tested on 20 October 2026: 11 of 11 paid test orders completed after a fix (0 of 8 before it); on screens with planted defects it found both main defects on one and 3 of 5 on another.

Where it falls short: The marked areas do not always sit on the right element: the written observations are more reliable than the boxes.

Catalog Copy

$7 per order

Turns 1–50 CSV rows of product specs into English descriptions assembled from your facts, plus an import CSV and a JSON of the source facts.

Tested on 20 October 2026: 11 of 11 test orders completed after a fix (1 of 8 before it).

Where it falls short: This is not free copywriting: text is assembled from your facts by a template, field labels are always English, and input formats are strict.

Brief to Diagram

$7 per order

Draws a draft process diagram from a text brief; each block quotes the sentence of the brief it came from (SVG, HTML, JSON).

Tested on 20 October 2026: 16 of 19 test orders produced a diagram after two fixes (0 of 19 before them); 2 of the 3 refusals were correct (an explicit loop, a poem instead of a process).

Where it falls short: Cyclic processes are refused. A free pre-check now shows this before payment.

CSV Brief

$5 per order

Turns a fixed-schema weekly operations CSV into a short report with aggregates you can verify (HTML and CSV). No financial analysis.

Tested on 20 October 2026: 15 of 15 test orders completed after two fixes (0 of 15 before them); an independent recount of the aggregates found no discrepancies.

Where it falls short: Russian output reads choppy, and in one order the report came out in English although Russian was selected.

How we tested

On 20 October 2026 we ran about 250 test orders across the eight tools through the live pipeline. Payment was simulated with a signed webhook — no real card payments were made in the test. The model cost of the whole test campaign was under $1. Live smoke tests of two tools had passed, yet almost every paid order went to a refund; the write-up explains why.

Questions

How do I pay?

On the tool's own page: by card through Stripe. The price is shown before you pay; you pay per order.

What happens to my files?

File text is deleted after 24 hours and is not written to logs.

Can I use the output as is?

No. The tools check what can be checked mechanically — for example timestamps in SRT Rescue or placeholders in Locale Pack — not meaning. Treat the result as a draft and review it before you publish or ship it.

How much does each tool cost?

SRT Rescue $4 per file; Locale Pack $5 per order; Recall Pack $4 per order; Spec Check $7 per order; Screenshot Review $9 per order; Catalog Copy $7 per order; Brief to Diagram $7 per order; CSV Brief $5 per order. The price is shown on the tool's own page before you pay.

Which model runs the tools, and are there customer reviews?

The tools run on Claude Haiku 5.5. They are new pilots launched in October 2026 and have no customer reviews yet, so none are published here.

Who runs these tools?

Amplituda, a process automation studio in Riga, Latvia. The same team builds automation for companies on n8n and Claude.

Need this inside your own process?

The tools are small on purpose. When the job is part of a bigger workflow — requests, reports, catalogues, CRM — we build it as automation on n8n and Claude for your team.