VoiceMem is live · free · last verified 4 Nov 2026 · viability monitor
VoiceMem
Open-source dual-brain memory for real-time voice agents, streaming memory retrieval that adds roughly 300 tokens per turn
VoiceMem makes two bets most memory tools skip — separating facts from feelings, and starting retrieval mid-utterance — and it publishes numbers that make them worth testing. If your voice agent is token-bound or latency-bound on memory, benchmark this before renewing a managed contract. Just budget for a v0.0.2 research codebase you own end to end.
Verdict generatedVerified 9d ago · liveness 67/100 · cite: rightaichoice.com/tools/voicemem
- Developers building real-time voice agents that need low-latency, emotionally aware persistent memory
- Researchers studying voice AI memory architectures, with a technical report, eval scripts, and ChatMem-400K
- Teams self-hosting an open-source memory backend who want every layer swappable
- Prototypes and MVPs where per-turn token cost is a hard constraint
- Teams needing production SLAs, vendor support, or a managed service — this is a v0.0.2 research project
- Non-technical users who want a hosted API and no self-hosting
- Projects requiring ready-made multi-platform client apps
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip VoiceMem if you need a hosted memory API with an SLA and vendor support, or if nobody on your team can debug a Python streaming pipeline and manage local model downloads.
The software is Apache-2.0 and free, but the default model bundle is a separate Hugging Face download (zhifeixie/VoiceMem_Default_Models_Env) that you host and store yourself.
VoiceMem is $0 under Apache-2.0 — the whole dual-brain memory system, the ASR/emotion/embedding modules, VoiceMem Utils, and the ChatMem-400K dataset. That puts it below Mem0 and Zep, which charge for a managed memory backend, and its published ~300 tokens per single-turn query (versus Mem0's 6,956) keeps your model spend lower too. The trade is that you pay in engineering hours: no managed tier, no support contract, v0.0.2 code you self-host.
In short
VoiceMem — Open-source dual-brain memory for real-time voice agents, streaming memory retrieval that adds roughly 300 tokens per turn. Best for Developers building real-time voice agents that need low-latency, emotionally aware persistent memory, Researchers studying voice AI memory architectures, with a technical report, eval scripts, and ChatMem-400K, Teams self-hosting an open-source memory backend who want every layer swappable. Free to use.
What is VoiceMem?
VoiceMem — Open-source dual-brain memory for real-time voice agents, streaming memory retrieval that adds roughly 300 tokens per turn. Best for Developers building real-time voice agents that need low-latency, emotionally aware persistent memory, Researchers studying voice AI memory architectures, with a technical report, eval scripts, and ChatMem-400K, Teams self-hosting an open-source memory backend who want every layer swappable. Free to use.
VoiceMem is an Apache-2.0 memory layer for real-time voice agents, built by the xzf-thu team and sitting at about 2.9k GitHub stars. The core idea is architectural: instead of dumping every exchange into one vector store, memory splits into a left brain that organizes facts through schemas and entities, and a right brain that tracks persona, emotion, and relationships via standalone and cross-entity nodes. An agent can recall what you said next to who you are and how you sounded saying it. The pipeline streams from start to finish. Audio segmentation, ASR, memory extraction, and graph writes run while the speaker is still talking, and speculative prefetching fires retrieval mid-utterance so results are ready before the turn ends. Queries route first, then rank, and only Top-K memories reach the model context — the project puts a single-turn query at roughly 300 tokens. Everything is decoupled, including the underlying memory engine. Bundled components cover ASR, speaker verification, scene detection, emotion recognition, and local embeddings. Around the core sit VoiceMem Utils for out-of-the-box use, model families that read VoiceMem memories directly (including a fine-tuned Qwen reply model), the ChatMem-400K dataset for fine-tuning, and an interactive web demo with two-stage barge-in and a PCM-based output timeline. v0.0.2 opened a pluggable speech-synthesis layer and fixed the event-date chain. Accuracy is published rather than implied: 91.2% on LoCoMo with only Top-5 memories and 69.44% on PersonaMem. This targets developers and researchers who need persistent, emotionally aware voice memory and will run the backend themselves — where Mem0 and Zep would hand you a managed service instead.
Key features
- Dual-brain memory: left brain stores facts via schemas and entities, right brain stores persona, emotion, relationships
- Fully streaming pipeline: audio segmentation, ASR, memory extraction, and graph writes while the user speaks
- Speculative prefetching inside a voice turn (0–300 ms) so retrieval starts before the speaker finishes
- Top-K memory routing and ranking to cap context per query at roughly 300 tokens
- Published 134 ms response time vs Mem0's 1,440 ms
- 91.2% on LoCoMo with Top-5 memories (Mem0: 61.68%) and 69.44% on PersonaMem
- Multi-modal memory from real audio: voice, speaker identity, sound events, multi-party conversation, music
- Built-in ASR, speaker verification, scene detection, emotion recognition, and local embedding modules
- Swappable memory engine and pluggable speech-synthesis layer (v0.0.2)
- SessionBuffer isolates per-session context; temporary turns purged at session end
- Two-stage barge-in: VAD pauses and preserves the audio queue, clears on stop or stable ASR text
- PCM-sample output timeline with AudioWorklet render progress for interrupt handling
- VoiceMem model families (fine-tuned Qwen reply model) that read VoiceMem memories directly
- Offline memory engine mode plus a continuous streaming/VAD interface
- Interactive web demo with streaming partial transcription
Behind the verdict
The pitch here is narrow and honest: memory for real-time voice, not a general-purpose assistant toolkit. If you're building a voice agent that has to remember a caller's allergies, tone, and history without blowing your context budget, that's precisely the problem VoiceMem was shaped around. The dual-brain split is the part worth your time. Facts get schema and entity structure for accurate lookup; persona and emotion live on their own nodes with cross-links. In practice that means the agent can answer a dietary question and also notice the caller sounds frustrated — two retrievals that managed memory services tend to blur into one fuzzy match. Where it earns its keep is latency and tokens together. Speculative prefetching starts the query before the speaker finishes, and Top-K routing caps what reaches the model. We'd reach for this when a per-turn token ceiling or response time is a hard constraint, or when a multi-party, emotionally loaded conversation is the whole point of the product. The published LoCoMo and PersonaMem figures are the right things to replicate on your own data before committing. Pass if you need SLAs, vendor support, or a hosted API. This is a v0.0.2 repository — you clone it, pip install, download models, warm them up, and debug at the source level. Relative to Mem0 or Zep, VoiceMem trades managed convenience for control over every stage, including swapping the memory engine and, since v0.0.2, the speech-synthesis layer. If your team has no one who wants to own that stack, the trade isn't worth it. Watch the operational edges: local models load lazily, so warmup before the first audio chunk is not optional; ingestion is slower than search because it extracts facts and builds graphs, which is fine for async writes but surprises people
Models under the hood
as of 2026-11-08
Limitations
- As of September 2026, VoiceMem is in early development (v0.0.2, released 09/01/2026) and may have incomplete documentation or unstable APIs.
- The project's performance claims (134 ms response, ~300 tokens per single-turn query, 91.2% LoCoMo at Top-5, 69.44% PersonaMem) come from the project's own README and technical report and may need independent verification.
- Being open-source infrastructure released under Apache-2.0, it requires technical expertise to install, configure, and integrate into existing voice agent pipelines; there is no managed cloud service or commercial support.
as of 2026-10-11
Researching VoiceMem? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Is VoiceMem still active?
VoiceMem is live · free · last verified 4 Nov 2026 · viability monitor.
Viability score
How well maintained and how widely used is VoiceMem? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: November 2026
How we score →What's new in VoiceMem
Checked 4 days agoAcross the latest 5 updates: 4 launches and 1 changelog entry.
v0.0.2 — 修复事件日期链路，移除右脑冗余记忆类别，开放可插拔语音合成层
Fixed the event-date chain, removed a redundant right-brain memory category, and opened a pluggable speech-synthesis layer so the TTS stage can be replaced.
v0.0.1 — 发布初代 VoiceMem 和 Technical Report
First VoiceMem release, published alongside the project's technical report documenting the dual-brain streaming architecture.
开源 VoiceMem 模型系列，可直接读取并理解 VoiceMem 提供的记忆
VoiceMem model families open-sourced; they can directly read and understand memories produced by the VoiceMem pipeline, including a fine-tuned Qwen reply model.
发布 VoiceMem Utils，开箱即用
VoiceMem Utils package released for out-of-the-box usage, giving a plug-and-play wrapper around the core memory system.
开源 ChatMem-400K 数据集
ChatMem-400K dataset open-sourced for fine-tuning VoiceMem models on conversational data.
What do users say about VoiceMem?
What people actually say about VoiceMem — is it worth it?
We scanned public community sources for VoiceMem on Sep 21, 2026 and could not establish that the discussion we found is about this tool rather than something else sharing its name. Only 2 of the posts we fetched could be positively tied to VoiceMem. Rather than publish a sentiment score built on the wrong subject, we publish nothing here and re-run the scan.
How we verified this
Every listed fact about VoiceMem — pricing, features, links, status — is re-read from the vendor's own pages by our verification pipeline and each change is logged with a date; the last pass ran on 4 Nov 2026. Analysis is drafted with AI models and checked against our editorial standards before publishing; see how we verify.
Verification history
We have re-verified VoiceMem 12 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — changed pricing page
- — changed who it suits, pricing tiers, features, summary, our verdict, description
- — changed pricing page, features, summary, description, pricing tiers, our verdict
- — changed pricing page
- — changed features, our verdict, pricing tiers, summary, description
- — changed summary, pricing tiers, features, our verdict, description
Showing the 6 most recent of 12 verification passes.
Free to cite with attribution — this page re-verifies continuously.
How much does VoiceMem cost in 2026?
VoiceMem: Free to use. Pricing last verified 4 Nov 2026.
Plans compared
Every published VoiceMem tier, its price as the vendor publishes it, who it fits, and what it adds over the tier below. The 12-month cost projection follows.
| Plan | Price | Billing | Who it's for |
|---|---|---|---|
| Open Source | Free | — | — |
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Where the pricing makes sense
The company stage and team size where VoiceMem's pricing actually pencils out — and where peers do it cheaper.
VoiceMem is $0 under Apache-2.0 — the whole dual-brain memory system, the ASR/emotion/embedding modules, VoiceMem Utils, and the ChatMem-400K dataset. That puts it below Mem0 and Zep, which charge for a managed memory backend, and its published ~300 tokens per single-turn query (versus Mem0's 6,956) keeps your model spend lower too. The trade is that you pay in engineering hours: no managed tier, no support contract, v0.0.2 code you self-host.
Has VoiceMem's pricing changed?
No plan price for VoiceMem has moved since we began logging in Aug 2026; the pricing page is re-read on every verification pass.
Who should pick VoiceMem?
Pick VoiceMem if you are Developers building real-time voice agents that need low-latency, emotionally aware persistent memory, Researchers studying voice AI memory architectures, with a technical report, eval scripts, and ChatMem-400K, Teams self-hosting an open-source memory backend who want every layer swappable, Prototypes and MVPs where per-turn token cost is a hard constraint.
Look elsewhere if you are Teams needing production SLAs, vendor support, or a managed service — this is a v0.0.2 research project, Non-technical users who want a hosted API and no self-hosting, Projects requiring ready-made multi-platform client apps, Deadlines with no room for model downloads, local warmup, and source-level debugging.
Skip it if Skip VoiceMem if you need a hosted memory API with an SLA and vendor support, or if nobody on your team can debug a Python streaming pipeline and manage local model downloads.
GeneratedReal-world workflow fit
Concrete scenarios for the personas VoiceMem actually fits — and what changes day-one when you adopt it.
Clone the repo, pip install voicemem, pull the default models with hf download, then call warmup() and ingest a WAV file to store a fact like a nut allergy before running a streaming query against it.
Outcome: A working dual-brain memory store that answers 'what are my dietary restrictions?' from the left brain and reports emotional context from the right brain, with no per-seat cost.
Run the evaluation folder against LoCoMo and PersonaMem, compare Top-5 retrieval accuracy and per-turn token counts against the published 91.2% and ~300-token figures, then swap the underlying memory engine to isolate the dual-brain contribution.
Outcome: Independently reproduced numbers instead of vendor README numbers, plus a decoupled component graph for ablation studies.
Wire the streaming API in as a VAD-style consumer, let speculative prefetch (0–300 ms) start retrieval during the question, and use Top-K routing to cap the memory tokens injected per turn.
Outcome: Retrieval that resolves before the speaker finishes and a memory-context footprint measured in hundreds of tokens per turn rather than thousands.
Use cases
- Build a voice assistant that remembers user preferences and past conversations over long periods.
- Create an empathetic voice agent that tracks emotional states and responds appropriately.
- Implement a low-latency memory layer for real-time voice interaction with speculative prefetching.
- Fine-tune VoiceMem language models on your own conversational data using the ChatMem-400K dataset.
- Replace the built-in memory engine with your own while keeping the dual-brain architecture intact.
- Integrate VoiceMem into open-source voice agent frameworks for personalized interactions.
- Run VoiceMem as an offline batch memory engine over recorded audio, ingesting WAV files and searching them by text query.
Setup time & first value
How long it actually takes to get something useful out of VoiceMem — broken out by persona, not the marketing-page minute.
A developer comfortable with Python and Hugging Face can reach a first successful ingest-and-search in one to two hours, most of it spent on pip install voicemem, the model download, and warmup(). Getting the streaming path wired into an existing voice agent with barge-in and AudioWorklet handling is a multi-day integration, not an afternoon.
Switching to or from VoiceMem
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From Mem0: move from the managed API to self-hosted storage by re-ingesting source audio through VoiceMem, since it extracts facts, tags, and graphs locally rather than accepting an uploaded memory record.
- →From Zep: replace the managed memory backend with the Apache-2.0 system, then re-extract persona and emotional context that Zep's graph did not capture.
- →From a flat transcript store: point VoiceMem's ingest at your recorded audio to build the left-brain schema/entity graph and right-brain persona nodes the transcript never had.
- →From a custom in-house memory layer: keep your retrieval logic by swapping it in as the underlying memory engine and retaining the dual-brain structure around it.
- ↗To Mem0: if you need a managed backend with support rather than v0.0.2 source, export your factual store and re-key it into Mem0's hosted API.
- ↗To Zep: for a managed graph memory service with a commercial SLA, port the entity relationships and drop the streaming prefetch that only matters in real-time voice.
- ↗To a plain vector database: if the emotional and persona layer is overkill, keep only the left-brain embeddings and query them directly.
- ↗To a commercial voice-agent platform: if you would rather buy turnkey voice memory than run ASR, embedding, and graph writes locally, the whole pipeline is what you would be replacing.
Resources & guides
Tutorials & Learning
YouTube returned 6 videos for “VoiceMem”, and we withheld 6: 6 could not be judged, because “VoiceMem” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about VoiceMem.
Official links
What are the best VoiceMem alternatives?
The closest alternatives to VoiceMem we list are Mem0, Cognee and Powermem. Ranked by direct product-type match, not category overlap.
| Alternative | Pricing | Starts at | Last verified |
|---|---|---|---|
| Mem0 | Freemium | Pricing not published | — |
| Cognee | Freemium | Pricing not published | — |
| Powermem | Free | Free | — |
Alternatives to VoiceMem
View all
Tools that pair well with VoiceMem
Common stack mates teams adopt alongside VoiceMem, with the specific reason each pairing earns its keep.
Mem0
Mem0 is an AI memory layer that gives agents persistent context across sessions, users and apps
Cognee
Cognee is open-source agent memory that turns scattered company data into a cited knowledge graph agents can recall.
Powermem
Open-source, self-hosted memory layer that gives AI agents persistent recall via hybrid retrieval.
VoiceMem vs its closest rival
VoiceMem's closest head-to-head is with Mem0.
Our VoiceMem vs Mem0 verdict written Sep 21, 2026 quotes $249, which no longer matches the verified price table, so it is being re-reviewed and is not shown.
Head-to-head comparisons
Voicemem vs Mem0
Verdict being re-reviewed against current pricing and data.
Voicemem vs Vectorize
These aren't really substitutes — they overlap only at the abstract level of 'agent memory.' Pick Vectorize if you're building text or MCP-based agents (Claude Code, Cursor, Google ADK) and want memory that turns failures into reusable judgment, with a free MIT self-host path and a managed cloud option when you'd rather not run infrastructure. Pick VoiceMem only if you're building a real-time voice agent that needs sub-150ms streaming memory with persona and emotion — and you're comfortable with a v0.0.2 research project, self-hosting, and benchmark claims that come from the vendor's own report.
GeneratedVoicemem vs Granica Ai
These are not competitors — Granica sells an enterprise efficiency layer for petabyte-scale tabular data lakes and long-running agents, while VoiceMem is a free Apache-2.0 memory backend for real-time voice agents. Pick Granica if you are an enterprise data engineer trying to cut Iceberg/Delta Lake storage and processing spend (with a cloud perimeter and petabyte-scale data), or an AI team needing agent state persistence via Myelin. Pick VoiceMem if you are building a self-hosted voice agent prototype and need emotion-aware memory at low latency, and you can tolerate v0.0.2 research-grade code with no SLA.
GeneratedVoicemem vs Tobira
These aren't competitors — don't shortlist them against each other. Tobira is for people who want a public, censorship-resistant address and machine-readable profile for an AI agent (so other agents and crawlers can find and talk to it). VoiceMem is for builders wiring persistent, emotionally aware memory into a real-time voice agent. If you need agent discoverability, take Tobira. If you need voice memory with published latency and accuracy numbers, take VoiceMem. If you need both, use both.
GeneratedVoicemem vs Composio Mcp
These don't compete. Composio MCP is the plumbing that lets a coding agent (Claude, Cursor, Codex, ChatGPT) actually send the email, update the Salesforce record, or open the PR — priced freemium and climbing to $99/mo when your team outgrows the free tier. VoiceMem is an Apache-2.0 memory engine for people building real-time voice assistants who want their agent to remember not just facts but tone and relationships — free, self-hosted, and explicitly a v0.0.2 research project with no SLA. Buy Composio if your problem is app reach. Adopt VoiceMem if your problem is low-latency, emotionally aware recall in a voice loop. Nobody is choosing one over the other.
GeneratedVoicemem vs Arcade Ai
These aren't competitors — you would not swap one for the other, and a comparison page is the wrong place to decide. Arcade AI is a buy-an-enterprise-runtime decision for teams whose agents touch real user accounts and need per-action audit trails naming agent, user, and system. VoiceMem is a self-host, Apache-2.0 memory layer for people building a real-time voice agent who want persona and emotion recall at low latency, and who accept v0.0.2 maturity with no support. If you're asking "which of these do I pay for," the answer depends entirely on whether your problem is authorization or memory — and if it's both, you'd run them in different parts of your stack, not pick between them.
GeneratedFrequently asked questions
Categories
Best-of guides
For makers of VoiceMem
- Get a Verified Listing for VoiceMem ($39): a written decision within 1 working day.
- Free listing for another tool: Free reviews are decided in order, within 3 weeks of the review starting.
Used VoiceMem? Help shape our editorial sentiment research.