COMPARISON · NLP MODERATION APIs

A verdict you can replay,
on your own terms.

Tisane is a linguistics-based NLP API that analyzes abusive content across many languages. Filter8 is the deterministic alternative for the high-volume text hot path — a replayable, explainable verdict plus cleaned text, sub-millisecond, flat-cost, and nothing retained after the response.

Try the playground → Read the docs
100 free calls · nothing retained after the response · not sent to a third-party model.
Deterministic & replayable
An NLP model returns labels and a confidence. Filter8's verdict is a pure function of (bundle, input) — the same answer every time, with a trace you can replay in CI or an audit.
Flat cost, private
Per-call/usage pricing scales with volume and keeps text on a third party. Filter8 is flat at scale — and keeps nothing after the response.
Evasion + PII + ratings
26 evasion classes folded before matching, contact/PII classifiers, and rating-aware decisions from E to M — in one sub-millisecond call.

Tisane vs Filter8 — on text

CAPABILITY Tisane Filter8
ApproachHosted NLP analysisDeterministic engine
Cost modelPer call / usage — linearFlat — marginal cost ≈ compute
LatencyNetwork round-tripSub-millisecond, in-process
OutputAbuse labels + analysisVerdict + clean_text
Explainable / replayableAnalysis fields✓ reproducible explain trace
Data residencyText sent to TisaneNothing retained after the response
Evasion (leet · homoglyph · spacing · zero-width)Model-dependent26 classes, deterministic
Contact / PII (incl. spelled-out)VariesBuilt-in classifiers
Rating-aware (E–M · minors)—Per-rating, per-category ceilings
Multilingual linguistic analysis✓ core strength— English-first, deterministic
Tisane is a product of Tisane Labs. Comparison as of July 2026 based on public information and Filter8's positioning; capabilities and pricing change — verify current details with the vendor before relying on specifics.
WHERE TISANE WINS

Honest about the trade

Tisane's strength is multilingual linguistic analysis — many languages, nuanced abuse types, and natural-language understanding a deterministic English-first engine doesn't match. If broad multilingual nuance is your problem, that's a real fit. Filter8 is deliberately narrow: the fastest, cheapest, most private path for high-volume text — and it escalates the nuanced or multilingual tail to a model when you want that depth.

Tisane comparison questions

Is Filter8 a Tisane alternative?
For the deterministic, high-volume text path, yes — replayable, sub-millisecond, flat-cost. Tisane's strength is multilingual linguistic analysis; route that nuanced tail to a model async.
Can I self-host?
Not yet — Filter8 is a managed, hosted API today. It keeps nothing after the response and logs offsets, never the text.
Is deterministic accurate enough?
For known patterns and evasion on the hot path, yes and far faster. Escalate genuinely ambiguous cases to a model, off the hot path.
Does Filter8 do many languages?
It's English-first and deterministic today. For broad multilingual nuance, pair a model for the review tail — deterministic core, model on the edge.
COMPARE MORE
vs Perspective — the retiring API → vs bad-words — the npm wordlist → vs CleanSpeak — the enterprise suite →

Deterministic moderation, on your terms.

Replayable and flat-cost. Run your worst-case inputs and decide on evidence. 100 free calls, no card.

Get an API key Open the playground