Where these 13 reports land
Build it0
Scope it3
Keep it10
Category report
Voice AI
13 products ranked by how much of their useful core an AI coding agent can reproduce: 0 strong builds, 3 scoped builds, and 10 weak replacements.
How the score breaks down here
- Interface
- 25
- Core workflow
- 18
- Data access
- 16
- Operations
- 10
- Trust & safety
- 16
16/100 average buildability
Why people keep paying here
- proprietary models12 of 13 reports
Model quality and inference operations are part of the product.
- content rights11 of 13 reports
Licensed content and distribution rights are not reproducible with an LLM.
- scale infra11 of 13 reports
Reliability at the vendor's scale is an operations problem, not a prompt.
The typical achievable core in this category: turn supplied scripts into speech or avatars through documented model APIs.
Scope it
Thrumble
A single inbound receptionist agent is a credible focused build with LiveKit Agents or Pipecat plus a SIP trunk. Thrumble as a product is larger: agent config UI, knowledge bases, tools, call transfer, outbound campaigns, phone numbers, transcripts, and production telephony. You can DIY the core conversation loop; you still own carriers, media ports, failed calls, and ops.
Scope it
ekto
The pipeline is no longer exotic: capture mic audio, segment it with a voice activity detector, transcribe with Whisper, translate, speak it back with a local TTS voice. An agent can wire that into a working local app in a focused implementation and it will genuinely translate a conversation. What it will not do out of the gate is stay graceful for an hour: chunk boundaries clip words, speaker turns bleed together, latency creeps as the buffer grows, and the sentence by sentence pacing that makes these apps usable in real conversation is a tuning problem, not a coding problem. You also get no phone app, which is where voice translation actually happens. Fine for a desk setup and travel prep, unconvincing when you are holding it out to a stranger in a market.
Scope it
NaturalReader
The core loop is buildable, but a dependable replacement becomes a substantially larger project. For NaturalReader, read user-provided text and documents aloud with a small set of licensed or local voices. The hard boundary is voice catalog, ocr, document formats, mobile apps, and commercial licensing options, plus models, compute, rights, and safety operations.
Keep it
Murf
A consolation build is possible, but the paid product's decisive value sits outside a solo rebuild. For Murf, turn user-authored scripts into labeled voiceover projects using licensed local voices. The hard boundary is proprietary voices, studio editor, stock media, collaboration, and commercial rights, plus models, compute, rights, and safety operations.
Keep it
HeyGen
A consolation build is possible, but the paid product's decisive value sits outside a solo rebuild. For HeyGen, assemble a labeled avatar-style video from user-owned media without cloning real people. The hard boundary is proprietary avatars, lip sync, translation, rendering, templates, and rights operations, plus models, compute, rights, and safety operations.
Keep it
Speechify
A consolation build is possible, but the paid product's decisive value sits outside a solo rebuild. For Speechify, read user-supplied documents aloud with a local TTS engine. The hard boundary is high-quality voice catalog, ocr, mobile capture, cross-device sync, and publisher partnerships, plus models, compute, rights, and safety operations.
Keep it
Synthesia
A consolation build is possible, but the paid product's decisive value sits outside a solo rebuild. For Synthesia, combine user-owned visuals with clearly labeled synthetic narration and captions. The hard boundary is proprietary avatars, rendering fleet, templates, localization, rights, and enterprise controls, plus models, compute, rights, and safety operations.
Keep it
Captions
A consolation build is possible, but the paid product's decisive value sits outside a solo rebuild. For Captions, caption and reframe user-owned short videos with local models. The hard boundary is proprietary editing models, mobile capture, avatars, cloud rendering, and creator workflow, plus models, compute, rights, and safety operations.
Keep it
PlayHT
A consolation build is possible, but the paid product's decisive value sits outside a solo rebuild. For PlayHT, run a local licensed TTS model and organize voiceover projects. The hard boundary is proprietary voices, real-time api, cloning workflow, compute, and commercial rights, plus models, compute, rights, and safety operations.
Keep it
Colossyan
A consolation build is possible, but the paid product's decisive value sits outside a solo rebuild. For Colossyan, create labeled training videos from user-owned slides and licensed local narration. The hard boundary is proprietary avatars, enterprise learning workflow, translation, and rendering, plus models, compute, rights, and safety operations.
Keep it
D-ID
A consolation build is possible, but the paid product's decisive value sits outside a solo rebuild. For D-ID, animate a user-created illustration with local speech and a visible synthetic label. The hard boundary is proprietary talking-head models, cloud rendering, api, moderation, and rights workflow, plus models, compute, rights, and safety operations.
Keep it
LOVO
A consolation build is possible, but the paid product's decisive value sits outside a solo rebuild. For LOVO, produce labeled voiceovers from licensed local voices and user-authored scripts. The hard boundary is large proprietary voice catalog, editor, cloning, rights, and hosted rendering, plus models, compute, rights, and safety operations.
Keep it
WellSaid Labs
A consolation build is possible, but the paid product's decisive value sits outside a solo rebuild. For WellSaid Labs, produce clearly labeled voiceovers from user-authored scripts using licensed local voices. The hard boundary is premium proprietary voices, enterprise rights, pronunciation tools, workflow, and support, plus models, compute, rights, and safety operations.