Product Workflows Customers Pricing Security Docs Sign in Book a demo

Litigo product · Depositions

Deposa

Deposition intelligence: transcripts and video into partner-grade digests. Key admissions, witness-contradiction maps across the full matter, impeachment outlines, and time-coded clips — every finding tied to the exact line and timestamp.

Book a demo → See pricing

Speech AI Multimodal AI Agentic analysis

Primary hardware · H100 / L4 / DGX

The problem

A single deposition can run 400+ pages. Associates spend days digesting transcripts, hunting contradictions, and building impeachment outlines — and key admissions and inconsistencies across witnesses get missed under time pressure.

What Deposa produces.

Drop the transcript, audio, or video. Get structured, cited output before lunch.

Digest

Admissions, auto-tagged.

An auto-digest of key admissions with issue tagging across the testimony.

Contradictions

Across every witness.

Cross-witness contradiction mapping over the entire matter, not one transcript at a time.

Impeachment

Outlines, generated.

Impeachment outlines built from the inconsistencies Deposa finds.

Video

Time-coded clips.

Speaker-diarized audio/video transcription with clips tied to each quote.

Traceability

Line-cite on every claim.

Every extracted claim carries a line cite and timestamp back to source.

Built on the NVIDIA stack.

Layer	Technology	Why it's here
Hardware	H100 / L4 · DGX	ASR and summarization inference over long audio/video; fine-tuning the legal speech and summarization models.
Speech & video	Riva · PyNvVideoCodec · DALI	GPU speech-to-text with diarization and legal vocabulary; GPU video decode for testimony clips.
Serving	CUDA · cuDNN · TensorRT · Triton	Low-latency digest serving, near real time.
Platform	NIM · NVIDIA AI Enterprise	Deployable speech + digest microservices, cloud or on-prem.

Transcribing and diarizing hours of legal audio/video, then embedding and cross-referencing thousands of statements, is GPU-bound. Riva + TensorRT deliver accurate, near-real-time digests; CPU transcription is too slow and less accurate on legal jargon.

Why it holds up.

Legal-tuned speech

Legal-domain speech models plus cross-witness contradiction mapping over an entire matter go well beyond generic transcription tools.

A growing annotated corpus

A compounding corpus of annotated depositions improves accuracy per firm; the contradiction-mapping logic is hard to copy.

Priced by the hour

Per-deposition-hour and per-digest pricing inside Litigo, with a trial-prep tier for firms.

Cited, not summarized away

Every finding links back to the exact line and timestamp — usable in prep and at trial.

Part of the ecosystem

Deposa feeds admissions and contradictions to Verdex for evidence strength, to Citel for quote verification, and to the Litigo Writer agent — sharing the model and evidence layer.

Digest your next depo.

See Deposa on a real transcript, in 30 minutes.

Book a demo →