s.4nt.org — Suttas, Side by Side · Help

A reading guide for the parallel sutta viewer at s.4nt.org.

What you're looking at

Each sutta page displays the original Pāli alongside one or more English translations in side-by-side columns,
aligned segment by segment.
The columns scroll together, so the same passage is always at the same height across all translations.

The columns

Switching translations

Each column has a dropdown at the top showing the current translator's name.
Click it to swap that column for a different translation — for example,
switch the right column from Sujato to Thanissaro, or back to Pāli.

Reordering columns

Columns can be dragged left or right.
Grab the small grip icon (⠿) in a column header and drag it to the position you want.
The reading layout adjusts instantly.

Adding and removing columns

• Click the + button at the far right of the column bar to add a new column.
• Click the × on any column header to remove it.

A typical starting layout is Pāli + one English translation.
You can expand to three (or more) columns to compare multiple translators directly.

Navigation

Table of contents

The ☰ button (top left) opens and closes the table of contents sidebar.
It shows every section of the sutta — you can click any entry to jump there instantly.

Within the sidebar: - The − and + buttons at the top collapse or expand all sections at once. - Individual sections can be folded with the small triangle button beside each heading. - Resize the sidebar: drag the vertical divider on its right edge (just left of the ☰ button) to make the table of contents wider or narrower. The width is remembered across pages. - Scrolling the list: hover over the contents and use your mouse wheel, or drag the scrollbar on its right edge.

Copying

Two quite different things can be copied off a page, and the words for them collide, so keep them apart:

• Page contents — the text: every line, in every column you have open, as an HTML table.
• Table of Contents (of links) — the outline: headings only, each one a hyperlink back here. No text.

Both live in one place:
⋮ More → 📋 Copy.
But the everyday way to copy text isn't in a menu at all.

Just select and copy

Drag across whatever you want and press Ctrl-C (⌘-C on a Mac). No menu appears and nothing needs clicking —
what lands on your clipboard is already a proper HTML table:
one column per translation, in the order you have them on screen,
with the section headings as full-width rows between the passages.

Paste it into LibreOffice Writer, Word, Google Docs,
WordPress or Blogger and it arrives laid out, not as a wall of run-together text.
Paste it into a spreadsheet — Calc, Excel, Sheets — and each translation lands in its own column.

You don't have to see a passage for it to copy. The table is built from the text itself,
not from what happens to be drawn on the screen.
A big page only renders the part you're looking at, so a selection that sweeps past sections you never scrolled to —
or past sections you folded shut — still brings every one of them along.
On DN 16, selecting the whole sutta copies all 1,623 lines even though only a few hundred were ever on screen.

Two rules govern how much you get:

• Inside a single cell, copying behaves completely normally: highlight half a sentence, copy, and you get just those words.
• Across two or more cells, you get whole rows — every column of every line you touched, even the ones you only clipped the edge of.

📋 beside any heading

Hover over a section heading in the main pane and a small 📋 appears at its right.
One click copies that whole section — its heading, every line, every column — without any dragging.

⋮ More → 📋 Copy → Page contents

The entire page as one HTML table.
Same output as select-and-copy, without the selecting.

⋮ More → 📋 Copy → Table of Contents (of links)

The page's L24 outline (the numbered sections in the table of contents) as an indented, linked table of contents:
the sutta title at the top linking to the page,
then each heading nested under its parent, every entry a link back to that section here.
Paste into a plain-text editor instead and you get a clean tab-indented outline.

This one only appears on outline texts;
a plain segment-numbered page has no curated sections to export.

What the HTML table includes

The same 📋 Copy menu — reachable from ⋮ More, or from ⚙ Settings → Copy — carries four switches.
They apply to every route above, select-and-copy included:

Right-to-left columns (Arabic) and the italic bridge columns keep those properties either way —
only the background tints are optional.

Sutta ref

A passage can be addressed in two different notations.
They look almost identical — the separator is the only difference — but they mean different things.

L24 ref — mn125.4.3

An all-dots address (. throughout).
This is lucid24's own reference system:
a curated, human-facing outline of a text — chapters, sections,
and sub-sections that an editor has chosen and numbered for study.
On a text that has an L24 system (for example the Visuddhimagga,
where 8.4.16 is the sixteen-step ānāpānassati discussion),
these are the entries you see in the table of contents, and each one is a clickable anchor — … /index.html#8.4.16 lands on it.

SC ref — mn125:4.3

A colon before the last number (nik<n>:<paragraph>.<segment>).
This is SuttaCentral's segment id — the canonical,
machine-assigned coordinate of one exact line of source text, the same id used across SuttaCentral and Bilara.
It always points at a single segment, whatever the editorial outline happens to be.

The difference in one line: the L24 dot-ref is an editorial address (where a passage sits in a curated reading outline);
the SC colon-ref is a source address (which exact segment of the canonical text it is).
One text can carry both — the same line is, say, L24 8.4.16.3 and SC vism8:220.1.
The dot-ref is stable to how a reader navigates;
the colon-ref is stable to the source.

Vinaya — root, commentary, subcommentary

The Vinaya's three layers share one numbering, so a passage keeps its address as you move between them:
take the root's address and add -a for the commentary, -t for the subcommentary.

That works because the commentary has no numbering of its own.
Every block of the Samantapāsādikā opens with the mūla paragraph number it expounds — 459.,
or a range like 44-9. — and that is exactly the paragraph number SuttaCentral publishes on the root segments.
So the number at the head of a commentary paragraph is the root paragraph it is about,
and the Sāratthadīpanī anchors to the same series:
all three layers agree on one number.

Two consequences worth knowing:

• The commentary is selective. It anchors about 1,400 blocks against the Vinaya's 3,340 paragraphs, so a root paragraph with no block of its own is covered by the preceding one — in the Parivāra one stretch of 153 paragraphs sits under a single anchor. Six root pages have no commentary at all: the two Pātimokkhas (whose commentary is a separate work, the Kaṅkhāvitaraṇī), Parivāra 2 and 5, and the nuns' sekhiya and adhikaraṇasamatha.
• The Samantapāsādikā's long opening — the Bāhiranidāna, on the three councils and the mission to Laṅkā — expounds no rule, so it has no root page to sit beside. It gets its own: pli-tv-vi-a-pre.

The ‖ EBT button

On grouped-nikāya pages (AN, SN, …)
a sutta header may carry a small ‖ EBT button.
Click it for two kinds of cross-reference, shown in a popup (the numbers never clutter the header itself):

• EBT parallels — early-Buddhist-text parallels to this sutta in other collections: full parallels and resembling passages, each a clickable link.
• Pāli-edition sutta number (where it diverges) — where a printed Pāli edition numbers this sutta differently from SuttaCentral, that edition's own reference is listed (e.g. Siam 20.127–128, the Royal Thai Tipiṭaka). Only editions whose numbering actually diverges appear; at present that is the Siam (Royal Thai) edition.

A sutta shows the button if it has either.
It's the same affordance you'll find on the collection index pages.

Quick-jump box

The search box in the top bar (open it with the 🦘 button or by pressing /) takes a ref or a shortcut word, then Enter.
Your recent jumps are saved and shown as a dropdown while the box is empty.
It understands everything below.

Site shortcuts

Bare words that jump to a landing page (a leading slash is optional, so home and /home are the same):

Jump to a text

The address form depends on how a collection is organised.
In every case you can leave the numeric part off a book-level ref to land on that page's top.

(A ref only lands somewhere if that text has been published;
otherwise the box flags it.)

Land on a specific place

Add a coordinate to any text ref to land inside it:

• An SC ref (colon) — the source segment: mn10:2.5, an3.65:2.4.
• An L24 ref (dots) — a curated outline section, on texts that have one: mn125.2.3, vism8.4.16.

Staying on the current page

You don't have to retype the work's own name.
On the page you're reading, a bare L24 ref jumps to that section — type 8.4.16, not vism8.4.16 —
and an SC ref jumps to that segment — type :2.5 (or,
on a multi-chapter work like the Visuddhimagga, keep the chapter:
8:216.1), not the full vism8:216.1.
The full, name-qualified forms above still work everywhere (and are what you need to jump to a different text).

Linking straight to a ref — ?q=

Any page on the site also takes a ref in the URL, written ?q=<ref>,
and jumps to it as the page loads — the link does the typing for you.
It accepts exactly what the box accepts,
which means a citation can be linked without knowing which page the passage lives on:
the site works that out from the ref itself.

The ref can be written in any of three conventions, and a link may use whichever one the passage is normally cited in:

• SuttaCentral segment ids — the colon form, mn125:4.3, addressing one exact line of source text. This is SuttaCentral's own numbering, the same id their site puts after the # in a URL, and it is described in full under SC ref.
• lucid24.org outline addresses — all dots, no colon: vism8.4.16, mn125.2.3. These are lucid24's curated study outline rather than a source coordinate — see L24 ref. A text only answers to one if its outline has been published here.
• DPR abbreviations — the -a (commentary) and -t (subcommentary) names used by the Digital Pali Reader: mn-a125, an-a3.65, dhp-a163, snp-a5.3. These matter most for the commentarial layers, which SuttaCentral largely does not carry, so a DPR-style abbreviation is often the only ref a passage has.

One thing to know about the commentarial refs:
their numbering here has in places been re-synced onto SuttaCentral's,
so that a commentary ref lines up with the root passage it comments on. snp5.3 and snp-a5.3 are the same verse,
and dhp-a163 is the story on Dhammapada verse 163 —
even where the source edition numbers that story differently (in the Pāli it is the 133rd).
So you can usually take a root ref, add -a, and get its commentary.

Two conveniences worth knowing:

• Your browser will treat it as a search engine. Register https://s.4nt.org/?q=%s as a custom search keyword and you can type a ref into the address bar to open the passage.
• The ?q= part removes itself from the address bar once it has done its work, so the URL you copy afterwards is the plain anchor to where you actually are.

In-text navigation

Click any section heading in the main text to scroll it into view.
Each row also has a subtle highlight on hover so you can track which verse the translator notes refer to.

🔍 Searching

The 🔍 button in the top bar opens the search box — or just press s anywhere outside a text field.
It has two scopes, chosen with the tabs at the top:

• This page — highlights every match in the sutta you are reading. Like your browser's own Find, but Pāli-aware, and it looks in the translation columns rather than at the whole document.
• Whole site — searches every page at once and lists the passages it finds. This one needs a program running behind the files, so it appears only in the local preview (serve.bat). On s.4nt.org the tab is simply not there, and 🔍 is the page-local find.

Matching options — ⚙

The ⚙ button opens the matching options.
They apply to both scopes.

Two things about Whole word in Pāli.
Compound separators are removed before anything is matched,
so deva·putta counts as the single word devaputta and whole-word deva does not find it —
which is usually what you want from a compound.
And because the text is matched as characters rather than parsed into words, inflection still hides things:
whole-word sati will not find satiyā.

This page

Type, and every match on the page is highlighted at once;
the counter beside the box shows 1 / 17.

• Enter / Shift-Enter, the ‹ › buttons, or ↑ ↓ — step through the matches.
• Esc closes the box and clears the highlights.
• Collapsed sections are skipped, the way your browser's own Find skips them.

Whole site

Switching to Whole site opens the search as its own window over the page.
The first search in a column takes a few seconds while that column is prepared —
the box says what it is doing, and it is cached afterwards.

The window has four parts:

• Search in (the left rail) — which parts of the corpus to search. Unticking a division takes it out of the search, so every count changes. Your choice is remembered between sessions, and a text added to the site later is searched by default.
• Also (the row under the search box) — a second word that filters the results. See below.
• Show only (the chips above the results) — which of the divisions that matched to list. This narrows only the list: the counts still describe the whole search, so you can read the Visuddhimagga's ten hits while seeing that the ṭīkās hold 903. Click a chip to see only that division, click it again to go back.
• The results, grouped by the page they are on.

Each hit renders as a small parallel table of its own:
the column you searched on the left,
the Pāli beside it when you searched a translation, and a gloss column on the right.
Two settings under ⚙ control that — Show alongside (which column to gloss with;
Bodhi-esque by default, the one English column that covers the whole corpus) and Lines around each hit (0–
10 segments of context either side).

A result row is not a link. The text is ordinary selectable prose, so a click-a-word dictionary extension works on it and a stray click cannot throw away a search that took a minute. The ref chip is what navigates: click it to open the passage in a new tab, or press and hold (or right-click) for new tab / new window / this window. Enter presses the selected row's chip.

📋 copies what is on screen — either as a reference list or as a table,
using the same table conventions as the viewer's own "Copy as table".

Also — filtering the results by a second word

The Also row narrows a result set by a second string:
keep only the passages that also match it, or push out the ones that do.

Type a word and press Enter.
It becomes a chip you can act on, rather than a line of syntax you have to retype:

• The sign — + requires the word, − excludes it. Click it to flip.
• The scope — how far from the match the word is allowed to be. Pick it from the little menu on the chip; on within ± a number box appears beside it, so you can widen or tighten that one filter without retyping it.
• ✕ removes that filter. Backspace in an empty box takes the last one back.

Typing -word adds an exclude directly.
Add as many filters as you like;
a passage has to satisfy all of them.

The scope is the part worth understanding. There are four, from tightest to loosest:

• on the matched line — the word has to be in the same segment as the match. A segment is often half a clause, so this is very tight.
• within ± N lines — the word has to be within N segments either side, and never past the edge of the page. This is the adjustable middle; the default is 5.
• in the same sutta / section — the word has to be somewhere in the same unit: the sutta, chapter or story the ref names before its colon. ja521:46.2 is Jātaka 521, vism-t12:48.2 is chapter 12 of the ṭīkā.
• anywhere on the page — the loosest, and the one to be careful with (see below).

Searching for jhāna and filtering on sāriput shows the ladder plainly — 26 matches on the line,
219 within ±5, 623 within ±25, 4,000 in the same sutta, 9,156 anywhere on the page, out of 13,850 unfiltered.

Why "anywhere on the page" is the blunt one. A page is a unit of publishing here, not of text. kn/ja/index.html is one page carrying all 547 Jātakas, and scmy/vin/pli-tv-pvr8 is one page carrying a single segment — so this scope asks a wildly different question depending on where you are. in the same sutta / section is almost always the one you want instead: on a single-sutta page like an MN discourse the two are identical, and on a grouped page only the section scope means what you meant.

One limitation worth knowing:
± N lines is counted in the column the match was found in. Columns don't have the same number of lines (an empty cell is never indexed),
so there's no cross-column "five lines away" to count.
If you're searching the Pāli and want to filter on an English word,
use in the same sutta / section, which looks across every column you have selected.

The headline says what the filters took away, not just what is left:
"735 matches across 169 pages, filtered down from 13,850".
A required word is highlighted in the results in its own colour (green,
beside the query's yellow) so you can see where it landed.
If the filters leave nothing, the box says so and offers to clear them.

Which settings the filters share

A filter term is matched with the same options as the search word. Fold diacritics,
Fuzzy Pāli and Whole word all apply to it exactly as they apply to the query,
and it is looked for in the same columns.

Regex is the one exception. A filter term is always a plain string,
so . * ( in one are just those characters —
you can search with a pattern and filter with a literal in the same query.
(And with Regex on, Whole word is off for everything, filters included.)

Leaving and coming back

A whole-site search costs real seconds to compute, so it is kept:

• Reopening 🔍 on the same query repaints the results you left without asking the server again, with a note saying how long ago it was and a Search again button.
• Coming Back to the page you searched from reopens the window on those results. Arriving by clicking a hit does not — that is a request to read the sutta.
• Clicking outside the window does not close it. ✕ or Esc do.

Footnotes

Many translations include translator notes.
They appear as small superscript markers (¹ ² …)
inline in the text.
Footnotes open in a panel in the lower part of the left sidebar,
sharing that space with the table of contents above it.
When no note is showing, the panel collapses to a thin "Footnotes ▾" strip at the bottom of the sidebar.

Opening a footnote

There are three ways:

• Click a marker in the text to open that specific note.
• Click the "Footnotes ▾" strip and choose which translator's footnotes to show from the small menu.
• In a column's translator dropdown, pick "see their N footnotes" (shown only when that translator has notes on the current sutta).

Working with the panel

• 🦘 jumps the main text to the verse the note belongs to (the segment reference is shown in the button's tooltip).
• ◀ ▶ step to the previous / next note (shown only in single-note mode).
• all switches to "show-all" mode — the default — listing every note for that translator as one scrollable commentary; the step arrows hide, since every note is already visible.
• 📋 copies the note text to the clipboard.
• ✕ closes the panel (it collapses back to the strip).

Resizing the footnote panel

Drag the horizontal divider between the table of contents and the footnotes up or down to give each as much room as you like —
the split is remembered.

On narrow screens (phones), footnotes instead slide up from the bottom of the screen as a sheet.

⚙ Display settings

Click the ⚙ Settings button (top right) to open the settings panel.

The sutta index

The index page lists all suttas in the collection with:

• The sutta number, title, and vagga (chapter group).
• Translation badges showing which translations are available for each sutta.
• A 🔒 icon on suttas that have been proofread and locked.

Click any row to open that sutta's parallel viewer.

Reading offline

Every sutta page is fully self-contained — all text and styling is embedded in the HTML file.
If you save a page to your computer (File → Save Page As… in your browser), it will open and work completely offline, with no internet connection required.

Translations available

Each translation is reproduced from its original source site, under that source's own licence.
Where a source publishes authoritative or updated terms, those govern — please consult the linked site.

Additional translations are added as alignments are completed.

PTS — unmapped suttas

The PTS (Woodward, Hare, and Rhys Davids) translations are aligned to SuttaCentral's sutta numbering.
But the digitized PTS source carries some translator text that has no SuttaCentral sutta id to display it beside — for example,
where the PTS edition's own numbering runs past SuttaCentral's,
or where it spells out a repetition series (peyyāla) that SuttaCentral collapses into a single range.
So that text never appears inline on the AN/SN pages.

Rather than discard it, we collect every such passage — verbatim — on one reference page:
PTS — unmapped suttas.
It is grouped by nikāya and book, and pure repetition series are condensed to one representative sutta with a count,
so the page is complete in meaning without thousands of near-duplicate copies.

Proofreading & verbatim validation

The segment-by-segment alignment to SuttaCentral's Pāli numbering is done by AI, and a verbatim check —
automated against the original digital source translation — confirms that no wording was added, dropped, or changed.
(The verbatim check ignores whitespace added for display,
and normalizes some punctuation characters such as quotation marks and apostrophes.)

Русский (Dhamma.gift)

A curated Russian collection, distributed at Dhamma.gift/ru/.
The many individual translators are listed under this single title on the source site.
The text is already keyed to SuttaCentral's Pāli segment numbering at the source,
so it is shown verbatim beside the Pāli and the English editions.

Terms: free distribution only, no commercial use — see the site for the individual translators and terms.

See the source's translation methodology for how these translations were prepared.

Sujato_corrected

This is identical to the unaltered Bhikkhu Sujato SC edition, with the following terms restored to Pāḷi:

• samādhi → keep as samādhi (not "immersion")
• jhāna → keep as jhāna (not "absorption")
• nibbāna → keep as Nibbāna (not "extinguishment" or "quenched")

And these English translations changed to align with genuine EBT:

• mettā → friendliness (not "love")
• ariyasāvaka → noble one's disciple (not "noble disciple")
• pīti → mental joy (not "rapture") — in jhāna context, to contrast with physical sukha
• sukha → physical pleasure (not "bliss") — in jhāna context; sukha vedanā in general encompasses both physical and mental
• sampajāno → lucidly-discerning (not "aware" or "situational awareness") — this is an inflection of the paññā (discernment/wisdom) faculty, not mere awareness
• vitakka-vicāra → directed verbal thoughts and pondering them (not "placing the mind and keeping it connected")

This is by no means a comprehensive set of corrections —
just some of the most egregious violations of core EBT principles.

Sujato's footnotes are unaltered and contain views incompatible with EBT.
Some are useful;
readers are encouraged to exercise discernment and critical thinking.

Pāli Editions

The site offers Pāli itself in more than one form:

• SC — SuttaCentral's segmented Mahāsaṅgīti Tipiṭaka edition. This is the default Pāli column and the alignment spine every translation on the site is matched to, segment by segment. Per SuttaCentral's own publication record, the text is drawn from The Buddhist Era 2500 Great International Council Pāḷi Tipiṭaka (40-volume Roman-script edition, B.E. 2548 / 2005), prepared by the Mahāsaṅgīti Tipiṭaka Buddhavasse 2500 for the M.L. Maniratana Bunnag Dhamma Society Fund. It is based on the Vipassanā Research Institute's Chaṭṭha Saṅgāyana CD-ROM edition (B.E. 2537–2542, tipitaka.org), fully proofread and recited four times between 2000–2002 and 2006–2007 to verify every Pāḷi sound and correct printing errors from the original 40-volume edition, with the 1958 Myanmar Chaṭṭha Saṅgāyana edition as primary reference and 18 additional national-script editions consulted for cross-checking. SuttaCentral extracted this Pāli text from the Mahāsaṅgīti edition's source XML, making only minor editorial changes to numbering and punctuation — no changes to the text itself. (Source · SC edition page · publication scpub64.)

• Siam (Thai script) — the Royal Thai (Siam / Syāmaraṭṭha) recension of the Pāli canon, in its native Thai script, digitized from the printed 2015 Royal Thai Tipiṭaka edition by the D-Tipiṭaka project. It is a separate manuscript lineage from the Mahāsaṅgīti edition used elsewhere on the site, so occasional wording differences between the two columns reflect the editions themselves, not translation error.

• Siam (IAST) — the same Royal Thai text, transliterated word-for-word into IAST (roman-script Pāli) for readers who don't read Thai script.
• Myanmar (မြန်မာ) — the Chaṭṭha Saṅgāyana Tipiṭaka (CST), the Sixth-Council Burmese recension, in its native Myanmar script, taken from the VipassanaTech tipitaka-xml digitization of the Vipassanā Research Institute (tipitaka.org) edition. This is the same lineage the SC Mahāsaṅgīti text is based on (see SC above), so it typically agrees with the SC column word-for-word; where the printed CST elides a repeated passage with …pe… (peyyāla), that segment may show a shorter or blank match.
• Myanmar (IAST) — the same CST text, transliterated word-for-word into IAST (roman-script Pāli) for readers who don't read Myanmar script.

The Siam and Myanmar columns are each matched to the SC Pāli segment-by-segment by an automatic,
deterministic text-matching process —
not by AI and not by a human editor — so the displayed text is always taken verbatim from the source;
a handful of passages (mostly verse/gāthā sections,
where word order differs more between editions) may show a shorter or blank match at a given segment rather than an invented one.

AI Translations

Translations generated by an artificial intelligence model,
rather than adapted or ported from an existing published edition.
Two kinds sit here:
those rendered directly from the source language (Bodhi-esque, and the two DeepSeek columns),
and bridge columns that translate the Bodhi-esque English into another modern language.
The five bridge columns — Arabic, Chinese, Spanish, French and Hindi —
are generated from the same input by the same pipeline, so each covers exactly what Bodhi-esque covers:
5,764 units across all five nikāyas, segment for segment.

Bodhi-esque

A fresh English translation rendered directly from the Pāli in the style of Bhikkhu Bodhi —
his characteristic register and technical vocabulary (e.g. concentration for samādhi,
taints for āsava, with jhāna, Nibbāna, and Dhamma kept in Pāli).

• Generated, not aligned: produced segment-by-segment from SuttaCentral's segmented Pāli, not derived from any existing English edition.
• Copyright-clean: Bhikkhu Bodhi's own copyrighted translations are not used as input. Only a glossary of his preferred terms and a description of his style guide the output, so the result is an independent translation, not a reproduction.
• Model: generated with DeepSeek Flash (V4), June 2026.

As with any machine translation, it is a study aid, not an authoritative edition —
read it alongside the Pāli and the human translations.

Arabic-esque

A Modern Standard Arabic (اللغة العربية الفصحى) translation, bridged from the Bodhi-esque English column —
the model is given the Bodhi-esque English segment-by-segment and renders each into Arabic under the same segment keys.

• Bridged, not direct-from-Pāli: the source is the Bodhi-esque English, so the Arabic inherits its segment alignment 1:1 (each Pāli segment → one Arabic line). Any error already in the English carries through, and it is one further translation step removed from the Pāli.
• Model: generated with DeepSeek V4 Flash, July 2026.

As with any machine translation — and doubly so for a translation-of-a-translation —
it is a study aid, not an authoritative edition.
The column is displayed right-to-left in an Arabic (Naskh) typeface, sized up from the Latin columns for legibility.

Chinese-esque

A Traditional Chinese (繁體中文) translation, bridged from the Bodhi-esque English column —
the model is given the Bodhi-esque English segment-by-segment and renders each into Traditional Chinese under the same segment keys,
using established Chinese Buddhist terminology (e.g. 世尊 for the Blessed One,
涅槃 for Nibbāna, 緣起 for dependent origination).

• Bridged, not direct-from-Pāli: the source is the Bodhi-esque English, so the Chinese inherits its segment alignment 1:1 (each Pāli segment → one Chinese line). Any error already in the English carries through, and it is one further translation step removed from the Pāli.
• Traditional, not Simplified: the text is rendered in Traditional characters (繁體), shown in a Traditional-Chinese typeface.
• Model: generated with DeepSeek V4 Flash, July 2026.

As with any machine translation — and doubly so for a translation-of-a-translation —
it is a study aid, not an authoritative edition.

Spanish-esque

A Modern Standard Spanish (español estándar moderno) translation, bridged from the Bodhi-esque English column —
the model is given the Bodhi-esque English segment-by-segment and renders each into Spanish under the same segment keys,
using the established Spanish Buddhist vocabulary (el Bienaventurado for the Blessed One,
sufrimiento for dukkha,
atención plena for sati), while bhikkhu, Dhamma, kamma, Nibbāna and jhāna stay in Pāli.

• Bridged, not direct-from-Pāli: the source is the Bodhi-esque English, so the Spanish inherits its segment alignment 1:1 (each Pāli segment → one Spanish line). Any error already in the English carries through, and it is one further translation step removed from the Pāli.
• Model: generated with DeepSeek V4 Flash, July 2026.

As with any machine translation — and doubly so for a translation-of-a-translation —
it is a study aid, not an authoritative edition.

French-esque

A Modern Standard French (français standard moderne) translation, bridged from the Bodhi-esque English column —
the model is given the Bodhi-esque English segment-by-segment and renders each into French under the same segment keys,
using the established French Buddhist vocabulary (le Bienheureux for the Blessed One, souffrance for dukkha,
pleine conscience for sati), while bhikkhu, Dhamma, kamma, Nibbāna and jhāna stay in Pāli.

• Bridged, not direct-from-Pāli: the source is the Bodhi-esque English, so the French inherits its segment alignment 1:1 (each Pāli segment → one French line). Any error already in the English carries through, and it is one further translation step removed from the Pāli.
• Model: generated with DeepSeek V4 Flash, July 2026.

As with any machine translation — and doubly so for a translation-of-a-translation —
it is a study aid, not an authoritative edition.
French quotation marks and spacing conventions are followed, so quoted speech reads « like this ».

Hindi-esque

A Modern Standard Hindi (मानक हिन्दी) translation in Devanagari script,
bridged from the Bodhi-esque English column —
the model is given the Bodhi-esque English segment-by-segment and renders each into Hindi under the same segment keys,
using the tatsama vocabulary Hindi Buddhist literature actually uses (भगवान् for the Blessed One,
दुःख for dukkha, स्मृति for sati, प्रज्ञा for paññā).

• Bridged, not direct-from-Pāli: the source is the Bodhi-esque English, so the Hindi inherits its segment alignment 1:1 (each Pāli segment → one Hindi line). Any error already in the English carries through, and it is one further translation step removed from the Pāli.
• Pāli loanwords kept Pāli, not Sanskritised: धम्म (Dhamma), निब्बान (Nibbāna), कम्म (kamma) and झान (jhāna) are written in Devanagari in their Pāli form — झान rather than ध्यान, which in ordinary Hindi means only "meditation". Proper names likewise keep their Pāli spelling (सारिपुत्त, not शारिपुत्र).
• Model: generated with DeepSeek V4 Flash, July 2026.

As with any machine translation — and doubly so for a translation-of-a-translation —
it is a study aid, not an authoritative edition.
The column is shown in a Devanagari typeface, sized up from the Latin columns for legibility.

DeepSeek — Line by Line

Generated with DeepSeek V4 Flash (June 2026),
translating from the source language into plain English one line at a time.
This guarantees full coverage:
every line of the source has a corresponding line of translation, with no gaps or skipped material.

DeepSeek — V4 Flash (Paragraph Chunking)

Also generated with DeepSeek V4 Flash (June 2026), but unlike the line-by-line version above,
larger chunks — typically full paragraphs — are given to the model at once.
The intention is that additional context produces a more natural, better-informed translation.
The tradeoff is weaker alignment:
the output doesn't map to the source line-for-line as precisely as the line-by-line version does.

Standard Disclaimers Apply

As with any AI-generated content, these translations should not be taken as authoritative on their own.
Always cross-check against reliable human translations and, where possible, the source language itself.

Why Allow AI Translations At All?

Given that AI translation carries a real risk of error, mistranslation, or subtle distortion,
it may seem odd for an EBT-focused site to include it rather than relying solely on established scholarly translations —
so it's worth explaining why AI-generated translations are included here.
Several reasons:

1. Full coverage — Many collections and individual suttas have no existing English translation at all, or their only existing translation is old, inconsistent, or poor enough that a modern AI model gets the core meaning across more accurately.
2. Speed relative to scholarship — Quality human translation is scarce and slow; a single translator may spend years on a single collection. AI translation offers an interim, clearly-labeled option while proper scholarly work catches up.
3. Independent cross-check — Even where a good human translation already exists, an independently generated AI version offers a second data point, useful for spotting translator bias, inconsistency, or error by comparison — not as a replacement.
4. Consistency across a large corpus — A single model applying consistent methodology and terminology across a huge collection can surface patterns that would be difficult for multiple human translators, working independently across different eras and styles, to maintain.
5. Transparency over omission — Clearly labeling material as AI-generated, along with the standard disclaimers, discloses the risk openly rather than hiding it — it's a labeled tool a reader can weigh accordingly, not a replacement presented as authoritative.

In short:
AI translations don't replace the human scholarly record — they fill its gaps,
sit alongside it as a cross-check, and are labeled honestly as what they are.

Frank (lucid24) translator

A first-draft column:
frank's own vocabulary applied across the corpus, with human proofreading still in progress.

As of August 2026,
fewer than 100 suttas are completely translated by frank himself. Those are the ones he has taken through by hand and stands behind as his own work.

The rest of the corpus is a first draft produced by AI, under two hard requirements:
follow frank's glossary of preferred renderings for the technical vocabulary —
where it departs from conventional English deliberately rather than accidentally (undistractible-lucidity for samādhi,
Dharma-remembrance for sati,
bodily-pleasure for sukha, pain-&-suffering for dukkha,
asinine-inclination for āsava) — and give full coverage of every Pāli segment,
so every segment of the source has a line of translation, with no gaps and no skipped material.

Why the first draft was rebuilt

The original draft translated the Pāli from scratch, line by line,
and shared each translated line canon-wide through a single translation memory.
That draft was abandoned:
correcting its mistakes and distortions by hand cost far more time than the draft itself had saved.
So the first draft was regenerated under a changed instruction.
It keeps both requirements above — full Pāli-segment coverage, and frank's glossary —
but now takes the Sujato_corrected edition as the authority for grammar,
syntax, and referents, and for which sense to pick when a Pāli-to-English dictionary entry offers several.
The model is no longer construing the Pāli on its own;
its job is to relexicalize an already-sound English rendering into frank's vocabulary,
and to restore what that rendering paraphrased away.

• Glossary-enforced, not merely requested: each technical lemma's rendering is checked in the output, not just asked for in the prompt, so one Pāli word reads the same way throughout the canon instead of drifting between synonyms from sutta to sutta.
• Coverage: 4,436 units — all of DN, MN, SN and AN, plus 14 books of the Khuddaka Nikāya — with every Pāli segment carrying a line of translation.
• Model: AI-assisted, generated with DeepSeek V4 Flash, August 2026.

It is a first draft and says so.
Read it against the Pāli beside it and against the human translations —
and expect its technical vocabulary to look unfamiliar by design, not by error.

EBT/LBT Classification Legend

Explanation

This system marks four categories along two axes:
chronological origin (early vs. late) and compliance with core EBT (benign vs. malignant).

• EBT (Early Buddhist Teachings) — material understood to derive from the earliest layer of the tradition, closer to the historical source.
• LBT (Late Buddhist Teachings) — material developed in later periods, building on or diverging from the earlier layer. The Abhidhamma/Abhidharma literature is the largest early body of it, and the schools diverge sharply in how far they depart from the sutta layer — see Differences in Abhidharma Schools.
• ✅ Benign — content that is doctrinally consistent with core EBT, whether it's genuinely early material or a later elaboration that stays faithful to it.
• ⛔ Malignant — content that departs from or contradicts core EBT, whether misattributed as early when it isn't, or developed later in ways that undermine the earlier teaching.

The Four Categories

• ✅ EBT — genuine early material, compliant with core EBT (compliance here is closer to definitional, since this is the core)
• ⛔ EBT — material falsely claiming early/authoritative status, and/or not actually compliant with core EBT despite the label
• ✅ LBT — later development that remains compliant with core EBT; a legitimate extension, honestly presented as later
• ⛔ LBT — later material that departs from or contradicts core EBT, regardless of how it's presented

This gives a quick visual shorthand for sorting sources or claims by how faithfully they track the earliest teachings,
without needing to write out the full judgment each time.

Note on Terminology

"Benign" and "malignant" here are not value judgments on any particular sect, tradition, or its practitioners.
The terms are framed strictly in relation to genuine core EBT compliance,
and describe the effect that compliance or non-compliance would have on the outcome of practice for someone who intends to follow genuine EBT.
A teaching is "benign" if following it supports that intended outcome,
and "malignant" if following it undermines or diverts from that outcome —
regardless of the sincerity, tradition, or good faith of those who hold it.
This framing reflects the charter and goal of the Early Buddhism Meditation Preservation Society (EBMPS):
preserving and clarifying genuine EBT for practitioners seeking it.

Why Include Malignant Material at All?

Given that the explicit purpose of this site is to preserve and clarify genuine Early Buddhist Teachings,
it may seem counterintuitive — even contradictory —
to devote space to material that the classification system itself labels malignant,
so it's worth explaining directly why that material is included here rather than simply excluded.
Several reasons:

1. Partial genuine content — Many ⛔-classified sources still contain some percentage of authentic, useful EBT material embedded within them. Excluding the whole text would mean losing that genuine content along with the corrupted portions.

2. Tracing corruption — Studying malignant material lets us trace how, when, and where deviation from core EBT entered the tradition — valuable for understanding the historical process of doctrinal drift, not just its end products.

3. Discernment training — Seeing a malignant text alongside the genuine EBT it diverges from is often the fastest way to learn the difference. Discernment is built through contrast, not by reading only "correct" material in isolation.

4. Meeting practitioners where they are — Many people already treat certain ⛔-classified texts as authoritative, because that's what their tradition handed them. Engaging directly with those specific texts — rather than ignoring them — is the only way to actually reach practitioners working from them.

5. Credibility of the classification system — A system that only ever produces ✅ ratings would look like a rubber stamp. Publicly classifying and explaining malignant material demonstrates that the criteria are being applied consistently, not selectively deployed to flatter the tradition.

6. Harm reduction over blanket avoidance — Simply telling someone "don't read that text" rarely works once they've already encountered it elsewhere. Precisely flagging which parts are corrupted — rather than a blanket condemnation — lets someone already engaging with the material salvage what's genuine and discard what isn't.

7. Inoculation against future distortions — Studying known patterns of corruption gives practitioners a template for recognizing new or unclassified material that follows similar patterns, rather than requiring every text to be pre-vetted for them.

8. Scholarly completeness — If EBMPS is to function as a serious reference rather than a curated devotional reading list, it needs to represent the actual landscape of Buddhist literature — not a filtered subset. Omitting malignant material would leave the archive incomplete and less useful for genuine comparative study.

In short:
classifying malignant material isn't an endorsement of it —
it's part of how genuine EBT gets preserved, taught, and defended.

Canonical Precedent: Suttas That Contrast Wrong and Right

Several suttas already use a contrastive method — pairing wrong and right views,
factors, or practices — to illuminate what is correct.
This gives canonical precedent for the classification approach used here.

Directly paired wrong/right formulations

• MN 117 (Mahācattārīsaka Sutta) — explicitly pairs wrong view/wrong path factors against right view/right path factors, including the mundane-vs-transcendent right view distinction.
• SN 45.8 (Vibhaṅga Sutta) — analytical definitions of the eightfold path factors; often read alongside MN 117, which builds on this kind of analysis.
• AN 10.103 / AN 10.104 (Micchatta Sutta, "Wrongness") paired with AN 10.105 / AN 10.106 (Sammatta Sutta, "Rightness") — the ten wrong factors (wrong view through wrong liberation) set directly against the ten right factors.
• MN 9 (Sammādiṭṭhi Sutta) — Sāriputta's exposition of right view, built through repeated wrong-understanding/right-understanding contrasts across many topics (the unwholesome/wholesome, nutriment, dependent origination, and more).

Corruption / counterfeiting of the teaching itself

• SN 16.13 (Saddhammapaṭirūpaka Sutta) — the "counterfeit Dhamma" text; Kassapa asks the Buddha why the true Dhamma doesn't last, and the Buddha explains how a counterfeit version arises and displaces it. Likely the single most directly relevant sutta to this project's premise.
• AN 4.180 (Mahāpadesa Sutta), also found in DN 16 — the "Four Great References," criteria for testing whether something attributed to the Buddha should be accepted as genuine by checking it against the Sutta and Vinaya. This is essentially the canonical precedent for doing an EBT/LBT classification at all.
• MN 22 (Alagaddupama Sutta) — the snake-simile sutta on wrongly grasped vs. rightly grasped Dhamma; warns about misusing or misapplying the teaching even when the source material is genuine.

Wrong practice/view systems contrasted with the genuine path

• DN 2 (Sāmaññaphala Sutta) — surveys six contemporary teachers' wrong doctrines before presenting the Buddha's gradual training, using the contrast to illuminate what's distinct about right practice.
• DN 1 (Brahmajāla Sutta) — catalogs 62 wrong views as a way of mapping the terrain around right view.
• MN 76 (Sandaka Sutta) — contrasts four kinds of flawed holy life with the genuine one.
• MN 8 (Sallekha Sutta) — a long list of paired wholesome/unwholesome states used to define "effacement."

Theravāda Aṭṭhakathā & Ṭīkā Chronology

(ChatGPT AI-generated summary, 2026-07.)

A rough timeline of when the Theravāda Aṭṭhakathā (commentaries) and Ṭīkā (subcommentaries) were composed.

This timeline is Theravāda-only.
The Abhidhamma/Abhidharma of the other schools ran on its own track — for who was writing what, and where the schools agree and diverge, see Differences in Abhidharma Schools.

Overview Timeline

Early Sinhala Aṭṭhakathā (Lost)

Approximate period:
c. 250 BCE–400 CE

After the establishment of Buddhism in Sri Lanka, explanatory traditions developed around the Mahāvihāra.
Important early commentarial traditions included:

• Mahā-Aṭṭhakathā
• Mahāpaccarī
• Kurundī
• Andhakaṭṭhakathā

These were composed in Sinhala (Sīhala) and are no longer extant as independent works. Their existence is known mainly because later commentators, especially Buddhaghosa, state that they used these earlier sources.

Buddhaghosa & the Classical Pali Aṭṭhakathā

Buddhaghosa

Approximate date:
c. 400–450 CE

Buddhaghosa reorganized and translated earlier Sinhala commentarial material into literary Pali.
Major works include:

• Visuddhimagga
• Samantapāsādikā (Vinaya commentary)
• Sumaṅgalavilāsinī (Dīgha Nikāya commentary)
• Papañcasūdanī (Majjhima Nikāya commentary)
• Sāratthappakāsinī (Saṃyutta Nikāya commentary)
• Manorathapūraṇī (Aṅguttara Nikāya commentary)

Dhammapāla & Later Aṭṭhakathā

Dhammapāla

Approximate date:
late 5th–early 6th century CE

Associated works include commentaries on:

• Udāna
• Itivuttaka
• Theragāthā
• Therīgāthā
• Cariyāpiṭaka

By approximately the 6th century CE, the major Pali commentarial corpus was substantially complete.

Ṭīkā (Subcommentaries)

Ṭīkā explain the Aṭṭhakathā rather than directly explaining the Canon.

Early Ṭīkā

Approximate period:
7th–9th centuries CE

Some early subcommentarial works appear during this period.

Classical Ṭīkā Period

Approximate period:
10th–13th centuries CE

This was the major age of Theravāda scholastic expansion.
Important authors include:

• Ānanda
• Sāriputta
• Vācissara
• Other Sri Lankan scholastics

Many major Ṭīkā were produced during the medieval Sri Lankan kingdoms.

Simplified Timeline

• c. 500 BCEBuddha's teaching period
• c. 300 BCEBuddhism established in Sri Lanka
• c. 250 BCE – 400 CESinhala Aṭṭhakathā develop (now lost)
• c. 400 – 450 CEBuddhaghosa composes the major Pali Aṭṭhakathā
• c. 500 – 550 CEDhammapāla — further commentaries (esp. Khuddaka)
• c. 600 CEMajor Pali Aṭṭhakathā corpus substantially complete
• c. 700 – 900 CEEarly Ṭīkā (subcommentaries) begin
• c. 1000 – 1300 CEClassical Ṭīkā tradition (medieval Sri Lanka)
• c. 1400 – 1800 CELater Sri Lankan & Burmese scholastic works

Scholarly Sources

Oskar von Hinüber

A Handbook of Pāli Literature (1996) — standard reference for dating Pali texts; covers Aṭṭhakathā and Ṭīkā chronology.

K. R. Norman

Pali Literature (1983);
A Philological Approach to Buddhism (1997) — discusses historical development and linguistic dating.

A. K. Warder

Indian Buddhism — discusses Buddhist schools and Sri Lankan textual traditions.

Rupert Gethin

The Foundations of Buddhism — historical overview of Theravāda development.

G. P. Malalasekera

The Pāli Literature of Ceylon (1928) — classic study of Sri Lankan Pali literature.

Bhikkhu Ñāṇamoli

The Path of Purification (Visuddhimagga translation and introduction) — discusses Buddhaghosa and the commentarial tradition.

Summary

The Theravāda exegetical tradition developed over approximately 1,500 years:

• Lost Sinhala Aṭṭhakathā: c. 3rd century BCE–4th century CE
• Surviving Pali Aṭṭhakathā: mainly 5th–6th centuries CE
• Major Ṭīkā: mainly 10th–13th centuries CE
• Later scholastic Ṭīkā traditions: 14th–18th centuries CE

The broad scholarly picture:
the Canon was preserved first through oral tradition, then interpreted through early Sinhala commentaries, systematized by Buddhaghosa and later commentators, and further refined by medieval Ṭīkā writers.

Tips for comparative study

• Two translations + Pāli is a productive default — three or more is more fun. The Pāli column lets you check the original word when two translators differ.
• Reorder columns with drag and drop so the translations you're comparing sit next to each other. Footnotes share the left sidebar, so put a translator's column toward the left, near its notes, when reading text and notes together.
• Segment numbers on (SC segment refs) makes it easy to note a verse number and find the same passage on suttacentral.net for commentaries or other translations.
• Footnote "all" mode works well for studying a short sutta in depth — you get the notes as a continuous commentary alongside the text.
• If you are comparing two English translations and want Pāli close at hand but less prominent, put Pāli in the middle and shrink it by making the browser window narrower — the outer columns stay readable while Pāli serves as an anchor reference.

About

s.4nt.org presents the discourses of the Pāli Canon — the earliest record of the Buddha's teaching —
with the original Pāli and one or more English translations side by side,
aligned segment by segment, so the same passage sits at the same height across every column.

The site is a project of the Early Buddhism Meditation Preservation Society (EBMPS), at 4nt.org.

The pages are produced by a custom tool that re-aligns existing translations to SuttaCentral's segmented Pāli text one segment at a time —
"un-babbling" them back into a single, comparable structure.
Each page is fully self-contained:
saved to disk, it works offline with no server and no tracking.

See What you're looking at above for how to use the viewer.

Legal

The texts. Each translation is reproduced from its original source under that source's own licence —
see Translations available above for the source website and the specific terms of each.
Copyright in every translation remains with its respective author, translator, or publisher;
the texts are presented here for study and free distribution only, never for sale or commercial use.

The site. The site software and the original site material (layout, segment-alignment data,
and this guide) are produced by the Early Buddhism Meditation Preservation Society and offered for non-commercial use.

Contact. For corrections, licensing questions, or to report an error,
write to the Early Buddhism Meditation Preservation Society via 4nt.org.

This is a good-faith summary, not legal advice. Where it conflicts with the licence a source publishes for its own text, that licence governs.