I have been experimenting with local two-way audio on a physical Tapo C200 (HW 1.0, FW 1.3.17) and I think there is a reasonably small, self-contained media protocol layer that could fit in python-kasa.

The goal would not be to add media-player, TTS, transcoding or WebRTC support to python-kasa, but only to expose the low-level local Tapo media primitives required for microphone/talkback.

What I verified

On the C200, the media protocol is separate from the regular management API:

management: HTTPS :443
media:      HTTP  :8800 /stream

Talkback works locally using:

HTTP Digest authentication
→ open `talk` session (`mode: aec`)
→ receive a session id
→ G.711 A-law / PCMA, 8 kHz mono
→ minimal MPEG-TS
→ multipart stream with X-Session-Id
→ camera speaker

I validated this with a standalone Python POC against the physical camera:

• audible generated tones played successfully through the speaker;
• 8/8 loopback checks captured the expected tone through the camera microphone;
• a talk session remained stable for 30 seconds;
• reopening sessions repeatedly works;
• talkback works while preview is active;
• talkback also works while the Tapo app live view is open;
• no FFmpeg or other media framework is required for the speaker path.

The speaker path can be implemented using standard Python networking plus a small MPEG-TS muxer.

Relationship with existing python-kasa support

python-kasa already exposes or queries useful SmartCam information such as:

• getAudioConfig;
• audio_capability.device_speaker;
• audio_capability.device_microphone;
• getMediaEncrypt;
• vhttpd;
• the existing SmartCam credentials / SHA-256 credential hashing logic.

The existing SslAesTransport itself is not reusable because the media protocol is different, but some authentication/capability primitives can be shared.

Reference implementation

go2rtc already implements the same Tapo media protocol:

https://github.com/AlexxIT/go2rtc/tree/master/pkg/tapo

I used it as a protocol reference, but the proposed implementation would not depend on or embed go2rtc.

The relevant protocol-specific parts are small compared with the rest of go2rtc's media stack.

Proposed scope

I suggest starting with speaker talkback only.

Conceptually:

async with await camera.open_talk() as talk:
    await talk.write_pcma(data)

Where data is already PCMA / G.711 A-law at the codec/rate advertised by the camera.

Internally, python-kasa would handle:

• media connection on port 8800;
• Digest authentication;
• talk session lifecycle;
• multipart framing;
• minimal MPEG-TS encapsulation;
• real-time pacing;
• reconnect / timeout errors.

It would deliberately not handle:

• MP3/AAC decoding;
• generic resampling;
• TTS;
• WebRTC;
• Home Assistant / Music Assistant concepts;
• video streaming.

Possible internal structure

Something along these lines:

SmartCamDevice
├── management transport (:443)
└── media client (:8800)
    └── TalkSession
        └── write_pcma(...)

The exact public API should of course follow the project's existing conventions.

Microphone / preview

I also validated microphone capture:

preview
→ encrypted multipart video/mp2t
→ AES-CBC decrypt
→ MPEG-TS demux
→ PCMA 8 kHz

That works on the same C200, but I would prefer to keep it out of the first PR because it adds considerably more complexity and firmware/codec variations.

A second PR could add a low-level preview/audio iterator once the talkback transport is accepted.

Open questions

Before turning the POC into a proper PR, I would like feedback on:

1. Is a low-level SmartCam media transport within python-kasa's intended scope?
2. Would a speaker-only first PR be preferable?
3. Should the public API accept only encoded PCMA frames initially, or would a small PCM16 → G.711 helper also be appropriate?
4. Is there a preferred place for this transport/session layer in the current SmartCam architecture?

If this direction is acceptable, I can prepare a focused first PR with unit tests for authentication, session setup, multipart framing and deterministic MPEG-TS output, plus validation against the physical C200.