zhenzhen-minimax-audio-lowprice

One node, four jobs: Comfly_minimax_audio_lowprice ("zhenzhen-minimax-audio-lowprice") is a MiniMax audio Swiss army knife. The model dropdown picks between minimax-music-2.6 for music generation, minimax-speech-2.8-hd and minimax-speech-2.8-turbo for text-to-speech, and minimax-voice-clone for cloning a voice from a reference clip. Whatever you choose, the result comes back as an AUDIO tensor you can preview or save. If you've been running separate TTS and music nodes, this collapses the whole audio corner of your graph into one box.

It lives in the T8mars/Comfyui-zhenzhen pack, the "Comfly" API-wrapper family that routes through the author's paid reseller service. This is a "lowprice" node: it uses the cheaper China-facing endpoint (api.seedance.nz) and takes its key through the shared zhenzhen-seedance2-low-price-api-settings node via the api_config input, rather than a per-node key field. All the usual submit-and-poll machinery lives behind it.

What to actually set

• model - required. Decide what you're doing first; the other fields only make sense in context.
• prompt - for TTS it's the words to speak; for music it's the description, with lyrics as a separate field and is_instrumental to drop the vocals.
• voice_id - the TTS voice (defaults to "Wise_Woman"), plus speed, volume, pitch for shaping.
• language_boost - helps non-English text; auto usually gets it right.
• custom_voice_id + reference_audio + clone_target_model - the voice-clone side. Feed it a sample through the reference_audio input and it synthesizes that voice at the target model's quality.
• output_format, sample_rate, bitrate, channel - the audio engineering stuff; the defaults (mp3, 32kHz, 128kbps, mono) are fine for most uses.
• need_noise_reduction / need_volume_normalization - free cleanups for the speech models.

There's also skip_error and a seed that's mostly ComfyUI cache control (fixed = cached result, other modes = new run; it isn't sent to the API).

Outputs

• audio - the result as an AUDIO tensor. Preview or save it and you're done.
• audio_url - hosted URL string if you want the file by hand.
• result_text - for cases where the API returns text (rare here), or diagnostic info.
• task_id - lets you look the job up in the site's async task list.
• response - the raw payload for debugging.

Installing it

Whole-pack install, no models, nothing on your GPU:

cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/Comfyui-zhenzhen
# restart ComfyUI

ComfyUI Manager → search "Comfyui-zhenzhen" also works. Then drop in a zhenzhen-seedance2-low-price-api-settings node, put your key in it, and wire its api_config here.

The honest bit

The pack's own changelog tracks these exact models - minimax-music-2.6, the speech-2.8 pair, and voice-clone - so the node tracks the real API surface, which is more than most wrapper nodes manage. But it's a reseller's endpoint: your text and reference audio go through a third-party server, and per-call costs meter up if you're generating a lot of TTS. For occasional voiceover or BGM it's genuinely convenient - one node, one key, four capabilities. If you're churning out narration daily, a local TTS (the open ecosystem has solid options) is still cheaper at scale.