RESOURCES
Changelog
Newest first; breaking changes ship as a new version and are announced here first.
2026-10-19
4K output and true 5.1 surround mixes
Finished videos render up to 4K, matching the source; the mix ships as true 5.1 surround or stereo, chosen per job, in 24-bit PCM WAV, and Studio deliveries can take ProRes masters. Additive; no request changes.
2026-09-29
Live source languages: 32
Live sessions (experimental) are planned to take 32 source languages (the list): 18 of the 30 output languages plus 14 source-only ones; live output stays the 30. On GET /v1/languages, live_source now marks the 32, and nn (Norwegian Nynorsk) appears as a live-only source row (output: false, tier: null).
2026-09-16
99 source languages, 30 output languages
source_lang now accepts any of 99 source languages (the list), each transcribed and translated; the 30 output languages are unchanged and target_langs still takes only those. GET /v1/languages gains an additive output boolean (tier is null on source-only rows); a source-only code in target_langs returns 422 unsupported_language.
2026-09-16
Live sessions marked experimental
The live surface (sessions, WHIP ingest, WHEP playback, RTMP and SRT in & out, the live_source flag on GET /v1/languages) is experimental and not generally available; the reference now describes the planned surface (Live streaming), set up with the team under a Studio contract. The live latency figures published on 2026-08-23 and 2026-09-03 are withdrawn.
2026-09-04
The language set is 30
The supported set is 30 languages (Languages): new, as source and target on uploads, are Cantonese, Norwegian, Swedish, Danish, Catalan, Croatian, Serbian, Bulgarian, Slovak, Lithuanian, Latvian, Belarusian, and Kazakh; Hindi, Turkish, Polish, Filipino, Tamil, Telugu, Bengali, and Czech are no longer offered as source or target (a request naming one returns 422 unsupported_language). Live sessions (experimental since 2026-09-16) were planned to accept 11 of the 30 as source (restated in the 2026-09-29 entry above), and GET /v1/languages reflects the set.
2026-09-03
Live latency restated end to end
Live latency was restated as an end-to-end figure plus per-stage figures for translation, voice, lip-sync, and transport (all withdrawn 2026-09-16); the earlier audio-only figure was withdrawn at the time. Live sessions are experimental since 2026-09-16.
2026-08-26
API access is enterprise-only
API access is enterprise-only: keys are issued under Studio contracts (private integrations, volume pricing, a named point of contact) -- book a call. Live sessions (experimental since 2026-09-16) are designed to deliver each speaker's translated voice on the original video.
2026-08-23
Live latency published; live_source on languages
Live latency was documented (figures withdrawn 2026-09-16). Live sessions (experimental since 2026-09-16) accept a subset of the languages as source (restated in the 2026-09-04 entry above); GET /v1/languages returns an additive live_source flag per language, Planned live sources lists the set, and uploads keep every language as source.
2026-08-23
Audio-only dubs: dub_mode
POST /v1/dubs accepts dub_mode: "audio": the dub delivers the audio tracks only, metered at the contract's audio-only rate, and audio sources (podcasts, voice tracks) use this lane. "full" (video + lip-sync) stays the default (Create a dub).
2026-08-16
SRT in & out on live sessions
Live sessions (experimental since 2026-09-16) speak SRT (Secure Reliable Transport) on both legs: ingest.protocol: "srt" returns an srt_url + passphrase pair (AES on the link), and outputs with protocol: "srt"get each language's dubbed feed pushed to the srt_url you provide. Protocols still mix freely per session (SRT in & out).
2026-08-16
Audio tracks on every output
Every finished language now delivers its audio as standalone files: format=audio returns the full mix as it sits in the mp4; format=voice returns the dubbed voice track alone, no scene bed, for your own mixing. Same signed-URL redirect as the video, covered by the same dub.output_ready webhook (Outputs & subtitles).
2026-08-13
The Familiar API reference is public
The full /v1 reference is live on this site. Documented surface:
- Dubbing: create, read, list, delete · the transcript & review flow · outputs & subtitles · metadata translation.
- Creators & control: creator profiles · Do Not Translate · context, each at the creator | stream | job scopes.
- Live streaming (experimental since 2026-09-16): sessions with WHIP ingest, WHEP playback, and RTMP in & out.
- Platform: webhooks · the languages · usage.
The API is enterprise-only: private access under a Studio contract, with volume pricing, raised limits, ProRes masters on request, and a named point of contact. Book a call and your key is issued with the contract.