Fix ValueError in FireRedTTS3Instruct.generate_tts

[Talpik]

Problem

Every call to fireredtts3.core.FireRedTTS3Instruct.generate_tts fails, on any device:

ValueError: not enough values to unpack (expected 3, got 2)

The text-frontend wrapper unpacks three values:

https://github.com/FireRedTeam/FireRedTTS3/blob/main/fireredtts3/core.py#L384

but the backend FireRedTTS3Instruct.generate_tts in fireredtts3/llm/fireredtts3_instruct.py returns only (gen_audio, gen_audio_sr) — which is also the two-value form documented in the README:

gen_audio, gen_audio_sr = instruct.generate_tts(...)

The sibling methods differ genuinely: generate_voice_design and generate_semantic_edit do return a third gen_text (the CoT plan / rewritten text), so only this one call site is wrong.

Repro

from fireredtts3.core import FireRedTTS3Instruct
instruct = FireRedTTS3Instruct("pretrained_models")
instruct.generate_tts(prompt_text=..., prompt_audio=..., prompt_audio_sr=..., text=...)

Fix

Unpack two values. One-line change, no behaviour change elsewhere.

Testing

ICL zero-shot cloning through FireRedTTS3Instruct.generate_tts now completes and produces correct audio; verified together with generate_voice_design, generate_semantic_edit and generate_acoustic_edit, which were already fine.

🤖 Generated with Claude Code