Otter Alternative for Confidential Client Calls: The Audio Path Decides

10/21/2026

By Kyle Nelson, Founder, Fazit

ogTitle: "Otter Alternative for Confidential Client Calls: The Audio Path Decides"

The search for an Otter alternative usually starts with the bot. You want a notetaker that doesn't appear in the participant list, doesn't announce itself, and doesn't make your client ask who else is on the call. Otter now offers bot-free capture, so that box is ticked. The harder question for confidential client calls is what happens to the audio after the bot is gone.

Fazit is a native macOS menubar app that turns 1:1 client calls into finished Markdown notes in an Obsidian vault. No bot joins the call, transcription and note generation run on the Mac, and the call audio is never written to disk. That last part is the difference between a tool that looks private and a tool that is private.

What "bot-free" actually means in 2026

Bot-free describes what your meeting looks like, not where your audio goes. Granola, Krisp, Circleback, and now Fireflies and Fathom all capture meetings without a visible bot. They removed the bot. They did not remove the server. The audio still streams to vendor clouds and third-party speech services, as Bot-Free Is Not the Same as Private explains.

Otter's own bot-free option follows the same pattern. The OtterPilot bot stays out of the participant list, and the audio still travels to Otter's infrastructure for transcription. For a consultant under an NDA, a lawyer with privilege obligations, or a therapist with confidentiality duties, that's the disclosure that matters. The NDA you signed already decided whether you can send client audio to a third party, as An AI Notetaker for Consultants lays out.

The audio path is the only comparison that survives a security review

Fathom, Otter, Fireflies, tl;dv, and Granola all write a clean summary now. That race is over. The comparison that still means something is architectural: where in the pipeline a durable artifact of your conversation is created, and who holds it afterward. Five of six create one on someone else's infrastructure, according to How to Compare AI Notetakers in 2026.

Fazit is the pipeline where the artifact never exists. A Core Audio process tap reads the call app's output and the microphone directly. The audio exists only in a fixed-size RAM ring buffer with no write API, destroyed on every exit path including errors. Every note records audio_retained: false in its frontmatter. There is no audio file to leak, no recording to subpoena, and no vendor database holding a copy of the conversation.

What the Otter litigation changed

Otter.ai faces a consolidated federal class action, In re Otter.AI Privacy Litigation, in the Northern District of California. The complaint alleges that OtterPilot records meetings without everyone's consent and feeds the audio into model training. No court has ruled yet, but the complaint is a checklist of what any cloud notetaker does with your calls, as Is Otter.ai Safe? details.

The legal risk has a price. Under California Penal Code § 637.2, the Otter wiretap class action puts the number at $5,000 per recording, plus a transcript that becomes discoverable evidence in any lawsuit. Discovery and wiretap statutes both point at the same artifact: the recording. The one you never made can't be subpoenaed, as The AI Notetaker Legal Risk Nobody Priced In explains.

Consent is still your job

Fazit's architecture reduces the vendor's exposure and the audio-file exposure. It does not remove your duty to have consent where the law requires it. California law prohibits intentionally recording a confidential communication without the consent of all parties, and Washington law requires consent from all parties to legally record in-person or telephone conversations, according to the Justia 50-state survey. Germany's § 201 StGB makes recording non-public speech without all-party permission a criminal offense, according to German counsel cited in Fazit's research.

The New York City Bar advises attorneys to obtain client consent before using AI to record a client call, and to consider confidentiality and privilege issues before recording, transcribing, or summarizing client conversations with AI. That advice applies to every professional with confidentiality obligations, not just lawyers. This is not legal advice, and consent and professional-conduct duties apply to the person running the call.

What Fazit does differently

Fazit captures two audio streams separately: your microphone and the call app's own audio. That gives you perfect You/Them attribution on a 1:1 call with no diarization guesswork. Transcription runs on-device using Parakeet via CoreML, and notes are generated by a local model on localhost. Neither the audio nor the transcript leaves the machine.

The output is a plain Markdown file written into a folder of your Obsidian vault, or Apple Notes. There is no Fazit database holding a copy. The note contains a summary, key moments, action items with owners, and a paste-ready follow-up email. When the call ends, you click Stop in the menubar and the note appears in your vault.

The comparison table

Tool	Bot joins?	Audio leaves device?	Audio file created?	Note location
Otter.ai	Optional	Yes	Yes	Vendor cloud
Granola	No	Yes	Yes, deleted after transcription	Vendor cloud
Fireflies	Optional	Yes	Yes	Vendor cloud
Fathom	Optional	Yes	Yes	Vendor cloud
Fazit	No	No	No	Your Obsidian vault

Four of the five cloud tools keep a recording. Granola deletes audio after transcription, and the audio still reaches Granola's systems and third-party services before deletion, as Granola Record Audio documents. Fazit is the row where the copy was never eligible to exist.

The hardware made this possible

Speech recognition moved to the cloud a decade ago because laptops could not run the models. Apple Silicon's Neural Engine and NVIDIA's Parakeet TDT v3 ended that. A server-class speech model now transcribes your calls on a MacBook, faster than the conversation happens, as Cloud Transcription Was a Hardware Workaround explains.

That hardware shift is the precondition for audio that never becomes a file. When transcription runs at roughly 110 times real time on an M4 Pro, there's no reason to store the audio. The sound exists in RAM, becomes text, and is destroyed. The note is the only artifact.

What you give up

Fazit is an early-stage product. It makes no maturity, scale, or certification claims, and holds no SOC 2 or HIPAA certification. It requires macOS 14.4 or later on Apple Silicon. There is no Windows, Linux, iOS, or Android version. Multi-party speaker labels are deliberately cut from the MVP because local diarization tops out around 13 to 18 percent error, and speaker recognition is the exact feature Fireflies is being sued over under Illinois BIPA.

The trade is real. You give up the enterprise search, the CRM sync, and the always-on auto-detection. You get a note that is a file you can grep, in a vault you own, with no recording behind it. For confidential 1:1 client calls, that's the trade most professionals should make.