Get started

Use cases

Six things people
actually build on AVFlow

Every one of these is a working demo, not a diagram. Each page gives you the full Job JSON, the reasoning behind its shape, and a link to the code that submits it. Where AVFlow has no direct path, the page says so and shows the real workaround.

Browse the demo repo Quick start
  1. 01

    Meeting recording & notes

    Record a call in speaker layout to your own storage, and get a transcript you can summarise.

    Conferencing products that ship a "record this call" button and email a summary with action items afterwards. One Job produces the composited video, the mixed audio, and the transcript together.

    Read it →

    Sources

    • livekit

    Nodes

    • video_mixer (speaker)
    • audio_mixer
    • asr

    Sinks

    • segment → S3 (HLS + WebVTT)
    <base>_subs.vtt Priya: we ship the migration Thursday
    1280×720 · 25 fps · HLS + WebVTT → S3
  2. 02

    Vertical co-host switching

    A 1080×1920 guest stream whose layout you retarget mid-broadcast without dropping the output.

    Live shopping and creator apps where a host brings guests on and off screen. The audience sees the layout change; they never see the stream reconnect.

    Read it →

    Sources

    • livekit

    Nodes

    • video_mixer (custom, 9:16)
    • audio_mixer

    Sinks

    • rtmp_push
    1080×1920 · 30 fps · RTMP
  3. 03

    Captioned vertical voice room

    Broadcast an audio-only room as 9:16 video that reads well with the sound off.

    Social audio rooms that want a vertical livestream. The hard part is not the audio — it is giving an audio-only room something to look at, with captions the viewer can actually see.

    Read it →

    Sources

    • livekit (audio)
    • web_capture
    • video_generator

    Nodes

    • asr
    • audio_mixer

    Sinks

    • livekit (captions back)
    • rtmp_push
    web_capture of your overlay canvas · RTMP
  4. 04

    Live interpretation channel

    Publish a translated voice track back into the room for listeners to switch to.

    Cross-border town halls and creator streams where each viewer picks their own language. Run one Job per language and they coexist in the same room.

    Read it →

    Sources

    • livekit (audio)

    Nodes

    • audio_mixer
    • translate

    Sinks

    • livekit (translated track)
    audio only · no video is composed
  5. 05

    AI co-host in the room

    A speech-to-speech agent that listens, answers, and can be interrupted.

    An always-available co-host that keeps a show moving when the audience is quiet, or a live assistant that answers questions out loud. One source, one node, one sink.

    Read it →

    Sources

    • livekit (audio)

    Nodes

    • voice_agent

    Sinks

    • livekit (agent voice)
    audio only · speech in, speech out
  6. 06

    Per-participant moderation

    Review every participant separately, so a finding points at a person rather than a composite.

    Live social and creator platforms that have to answer "who did this, and when" — for trust and safety, for a takedown request, or for a regulator. This is the one pipeline here that deliberately never mixes.

    Read it →

    Sources

    • livekit

    Nodes

    • audio_resample (16 kHz mono)

    Sinks

    • image → S3
    • websocket
    no canvas · no mixer · streams stay separate

Run them yourself

One Next.js app, six demos, real Jobs. Bring an API key and a LiveKit project — the AI nodes need no vendor keys of your own.

github.com/avflowhq/demos All examples