Fixie ai

Name: Fixie ai
Brand: Ultravox.ai

AI Native

Ultravox AI is a speech-native, real-time voice agent platform that provides audio-in/audio-out APIs and SDKs for building speech-to-speech conversational agents and combines a managed realtime service with publicly available model weights and code.

Est 2022|Valuation $51.5M|Raised $16.5M|6 Ref Customers

LinkedIn 6K|Web traffic 20K|FTEs 15|HQ Seattle, Washington, US

MarketsInteractive Voice Response (IVR)Agent Builder Platforms

Published Oct 15, 2026

01Common complaints

- When procurement requires documented enterprise compliance artifacts - If an organization’s procurement process mandates SOC 2 Type II or ISO 27001 evidence prior to vendor approval and the vendor cannot provide those artifacts, consideration should be given to providers with documented compliance. - When the primary requirement is post-call analytics rather than real-time S2S - Conversation intelligence platforms that focus on recording, transcription, and QA analytics may be a better fit when analytics on recorded calls is the primary objective. - When a vertically integrated contact-center stack is preferred - Organizations seeking a single vendor that bundles telephony, number provisioning, recording compliance, and agent desktop tooling may reduce integration overhead with telephony-centric vendors. - When emotional expressiveness is the top requirement - Use cases requiring highly expressive or affective voice (therapy-style interactions, companionship, brand voice acting) should evaluate specialist expressive-voice providers and models focused on prosody and affect. - Publicly stated limitations and roadmap items - The vendor has indicated that some speech-generation capabilities are forthcoming, which may mean certain production voice-generation features were under active development at the time of the materials reviewed. - Scale limitations - Plan-based limits are documented in public materials: Free/PAYGO concurrency cap is described as 5 calls; paid tiers are described as having no hard cap but remain subject to infrastructure limits under extreme load. - API and implementation constraints - Rate limits and call-creation constraints can affect bursty workloads; integration patterns should include backoff and retry strategies to reduce 429 errors. - Security and compliance gaps - Public documentation did not include SOC 2/ISO/GDPR/HIPAA/PCI statements in the materials reviewed; this can be a procurement blocker for regulated buyers.

02Why they buy

- Speech-native processing (differentiator) - The vendor positions the product as operating without a separate ASR stage; the vendor states this approach can preserve paralinguistic cues and reduce latency compared with transcribe-first pipelines. - Combination of open-weight artifacts and a managed platform - The availability of code repositories and model weights alongside a hosted managed service may appeal to organizations that value transparency and optional self-hosting, balanced against the operational requirements of self-hosting. - Published entry-level pricing and a free entry-level tier - The vendor publishes per-minute pricing and documents a no-cost entry-level tier to facilitate initial experimentation. - Areas where competitors may have advantages - More mature public documentation of enterprise compliance and audit artifacts - Larger presence on mainstream review platforms, providing a broader set of buyer-reported experiences - More mature, end-to-end telephony tooling depending on competitor offerings

03Field reports

Important note on review availability

- Major review platforms - No product-specific review pages for Ultravox/Fixie were identified on major review platforms in the sources reviewed. - Vendor-provided testimonials - The vendor homepage includes brief quote snippets and logo-based endorsements; these are vendor-selected and lack independent methodology or sampling context. - Community and social signals (anecdotal) - Community discussion threads include comparative statements such as a Reddit user describing Ultravox as a credible native speech-to-speech alternative in specific deployments; such statements are anecdotal and not representative sample evidence. - Public discussion threads about model releases are mainly re-shares of vendor blog content rather than in-depth hands-on reviews. - Open-source / GitHub signals - Models and code are available on public repositories and model-hosting sites; issue and pull request activity on those repositories can provide indirect signals of adoption and friction but were not exhaustively captured in this review. - Reported reasons for selection (inferred) - Organizations have adopted the platform to pursue speech-native speech-to-speech workflows as an alternative to cascaded ASR-based pipelines, to enable BYOT telephony flexibility, and to leverage open-weight models for transparency or potential self-hosting. - Possible reasons for switching to competitors (inferred) - Procurement requirements for documented enterprise certifications (SOC 2/ISO) can favor more established vendors with publicly documented compliance artifacts. - Use cases with a primary emphasis on emotional prosody or highly expressive voice may prefer solutions specialized in expressive speech. - Organizations seeking a fully integrated telephony bundle (numbers, carrier, recording compliance tooling) may prefer vertically integrated telephony/contact-center vendors.

Public customer case studies with metrics

No publicly named customer case studies with quantified outcomes were identified in the reviewed sources.

Vendor-published performance claims (not independent ROI)

The vendor has published benchmark results for model performance on audio tasks; these are vendor-reported model performance claims and do not constitute independent business ROI. Organizations should request the vendor's full evaluation methodology and reproducibility details before relying on these claims.

Independent evaluation signals

An academic benchmarking preprint included Ultravox among the configurations evaluated. Reviewers should consult the preprint for full results and methodology to assess reproducibility and context.

04Best fit

Ideal Customer Profile

- Company sizes - Likely fit: - Seed to mid-market software companies building voice agents as a product feature - Enterprises with engineering capacity and a defined voice automation roadmap - Practical sizing guidance (not vendor-stated): engineering teams ranging from small teams to large engineering organizations depending on product scope and integration complexity. - Industries and verticals where adoption is plausible (hypothesized) - Customer support and CX automation - Appointment scheduling and service dispatch - Marketplaces with two-sided support workflows - Fintech and insurance intake calls (subject to compliance validation) - Roles and personas - Decision makers: - Head of AI/ML Engineering - VP Engineering / CTO - Head of Customer Operations / Contact Center Technology - End users: - Voice platform engineers - Product engineers responsible for call-flow logic - QA teams responsible for conversational testing - Pain points the product addresses - Latency and interruption/turn-taking issues observed with cascaded ASR pipelines - Need for a voice-agent platform that supports tool-calling, RAG, and telephony integrations - Qualification criteria (good fit) - Real-time speech-to-speech experiences are required - Organization can build its own orchestration layer using APIs and SDKs - Interest in open-weight models for transparency or potential self-hosting - Deal disqualifiers (poor fit) - Procurement requires publicly documented SOC 2/ISO compliance artifacts prior to procurement - Requirement for a no-code, low-engineering turnkey solution - Need for a bundled telephony solution including numbers, carrier, or integrated compliance tooling out-of-the-box

05The pitch

What the product does (core value proposition in objective terms)

- Ultravox AI (formerly Fixie.ai) provides a speech-native, real-time voice agent platform (Ultravox Realtime) with APIs and SDKs to build speech-to-speech conversational agents without a traditional ASR → LLM → TTS pipeline. - The vendor states this approach aims to reduce latency and preserve paralinguistic cues (tone, cadence, pitch) relative to text-transcription-based approaches. - The offering includes: - A managed service (Ultravox Realtime) for production voice agents - Open-weight / open-source components (model weights available on community model hubs; code available in public repositories)

Target market and market positioning

- Primary targets include developer teams and companies building voice AI agents (contact centers, phone automation, embedded voice assistants) that require low latency and telephony flexibility (Bring-Your-Own-Telephony). - Market positioning emphasizes: - Speech-native inference (audio-in / audio-out) - Real-time performance - REST APIs and SDKs tailored for developer integration - Published per-minute pricing for pay-as-you-go usage is provided by the vendor

Get a comprehensive analysis of Fixie ai including market position, competitive landscape, adoption trends, and peer benchmarks.

Download full report →