Audio to Irodori Ref Config

This is the niche-of-a-niche node in the ALICE Lab Audio Tools pack, and if you're here you probably already know what Irodori TTS is. Quick version for everyone else: Irodori is an open Japanese text-to-speech model aimed at visual-novel and game voice work, and it reaches ComfyUI through the separate comfy-Irodori-TTS pack. That pack's sampler takes a special IRODORI_REF_CONFIG object describing the reference audio used for voice cloning. The awkward part: getting an arbitrary audio clip you've generated or processed in ComfyUI into that format has always been a detour through file loaders.

Audio to Irodori Ref Config closes that loop. It takes any standard ComfyUI AUDIO value and converts it to the exact IRODORI_REF_CONFIG that IrodoriTTS Sampler's ref_config input expects. You wire it, you don't write a config file.

How it works

The mechanism is small and honest, straight from the source. The node takes the first audio batch, averages all channels to mono (matching what Irodori-TTS itself does before codec encoding), clamps to 16-bit PCM, and writes a WAV under ComfyUI/temp/alice_lab_audio_tools/irodori_ref/. The filename is a SHA-256 content hash, so rerunning identical audio reuses the same file instead of cluttering your temp dir. It then returns a config dict with the ref_wav path, the normalize flags, and your max_ref_seconds - and it does all of this without importing comfy-Irodori-TTS. The IRODORI_REF_CONFIG output type is just a string that matches what that pack declares, so the two interoperate without a hard dependency. The README confirms the pack loads normally even when comfy-Irodori-TTS isn't installed - you just have nowhere to plug the output.

The inputs that matter

• audio - the ComfyUI AUDIO you want as reference (say, a slice you cut with Media Range (Input)).
• normalize_ref_audio - boolean, default off. On, it normalizes the reference to −16 dBFS and sets the "ensure max" flag, the same options IrodoriTTS's own Reference Audio loader exposes.
• max_ref_seconds - float, 1–120, default 30. Cap on how much reference the sampler will use.

Single output: irodori_ref_config of type IRODORI_REF_CONFIG.

What to actually expect

If you're doing Japanese TTS with a custom or processed reference voice, this is genuinely handy - you can run a clip through an editor or the Audio Mixer, trim it, and hand it straight to the sampler without touching disk. If you're not doing Irodori work, skip this node entirely; its one output is only meaningful to that one sampler.

Install

# ComfyUI Manager → search "ComfyUI ALICE Lab Audio Tools" → Install → restart
# or:
cd ComfyUI/custom_nodes
git clone https://github.com/alice-lab-dev/ComfyUI-ALICE-Lab-Audio-Tools
# restart, then Add Node → ALICE_Lab → Audio

Same pack-wide requirements as the rest of ALICE Lab Audio Tools: ffmpeg + ffprobe on the process PATH (macOS Homebrew paths auto-checked) and a current ComfyUI with the AUDIO type. No pip deps, no model downloads. And remember: the receiving IrodoriTTS sampler comes from comfy-Irodori-TTS, which you install separately. One honest caveat from the source: only the first audio batch is used and it's averaged to mono - so feed it the exact mono reference you intend, not a multi-take batch.