LAION · MOSS-8B voice-acting · prompting ablation

Does the DramaBox prompt format change MOSS-8B quality?

The question. A DramaBox prompt mixes what to say (dialogue in "quotes") and how to say it (stage directions). MOSS-TTS takes a separate text field (what) and instruction field (how). We always put the extracted dialogue in the text field, and ablate the instruction two ways:

· with_dialogue — instruction = the complete DramaBox prompt, dialogue still quoted inside.
· no_dialogue — instruction = stage directions only, the quoted dialogue removed (it already lives in the text field).

Crossed with a reference-audio arm (none @ temp 0.8 vs a random emotion snippet @ temp 1.0), over 25 prompts (15 character-consistent + 10 single-emotion), 16 takes each. Ranked by reward = invWER · (norm blend + norm genu).

0.947best-of-16 reward · winner: no_dialogue
-0.026with − no dialogue (best-16)
-0.128emotion-ref − no-ref (best-16)

1 · Which prompting style wins?

stylegroupsbest-16 rewardmean rewardinvWERWERblendgenu
with_dialogue500.9210.4710.6860.3553.0881.547
no_dialogue500.9470.4370.6240.4252.8261.554

Style within pathway type

pathwaystylebest-16 rewardmean rewardinvWER
ccwith_dialogue0.8520.4200.664
ccno_dialogue0.8540.3510.574
nonccwith_dialogue1.0250.5480.720
nonccno_dialogue1.0860.5650.700

2 · Reference audio: emotion snippet vs none

referencegroupsbest-16 rewardmean rewardinvWERgenu
no_ref500.9980.4910.6691.384
emo_ref500.8700.4170.6411.716

Adding a random emotion reference changes best-of-16 reward by -0.128 (26% of prompts better with the ref). Note the reference emotion is random w.r.t. the prompt, so this measures whether an unrelated emotional voice helps or fights the written direction. A full "does emotion follow the ref or the prompt?" attribution needs an emotion classifier on the outputs — flagged as follow-up.

3 · Sample-size scaling (best-of-k) by prompting style

0.350.530.710.891.07124681216with_dialogueno_dialogueno_refemo_refbest-of-k takes
kwith_dialogueno_dialogueno_refemo_ref
10.4700.4390.4930.417
20.6190.6010.6620.558
40.7410.7460.8000.688
60.8000.8170.8700.747
80.8360.8570.9050.788
120.8880.9110.9620.837
160.9210.9470.9980.870

4 · Prompt adherence — does the delivery match the written direction?

prompt-sim = VoiceCLAP-commercial cosine similarity between the direction text (the DramaBox prompt with the spoken dialogue removed) and the generated audio. It asks: does the voice actually perform "how to say it", separate from saying the right words? (These cosines sit around 0; the relative differences are what matter.)

mean prompt-simwith_dialogue / no_ref / ccno_dialogue / emo_ref / noncc
by prompting style-0.119-0.120
by reference-0.096-0.144
by pathway-0.096-0.155

Does selecting best-of-k by reward also raise prompt adherence?

-0.14-0.08-0.020.040.09124681216best-of-k by rewardbest-of-k by prompt-simbest-of-k takes
kprompt-sim of reward-pickprompt-sim of prompt-sim-pick (ceiling)
1-0.119-0.119
2-0.117-0.059
4-0.115-0.009
6-0.113+0.017
8-0.111+0.034
12-0.110+0.057
16-0.108+0.072

Key result: reward and prompt-sim are almost uncorrelated (r = -0.041). Selecting best-of-k by reward barely lifts prompt adherence (-0.119 → -0.108), while selecting by prompt-sim could reach +0.072. The emotion reference lowers adherence (-0.144 vs -0.096 no-ref) — a random-emotion voice pulls the delivery toward the reference and away from the written direction. If prompt adherence matters, add prompt-sim into the reward — best-of-N on reward alone won’t get it.

5 · Recommendations

Prompting: keep the dialogue in the text field either way; for the instruction, no_dialogue gave the higher best-of-16 reward (Δ 0.026). Stripping the dialogue from the instruction (directions-only) helped — the model isn't confused by seeing the words twice.

Reference audio: a random-emotion reference did not improve reward (Δ -0.128). Since it's emotion-random, prefer a reference whose emotion matches the prompt, or omit it and rely on the written direction.

Sample size: see the curve — most of the best-of-k gain is captured by k≈8; best-of-16 adds a little more. For bulk generation best-of-8 is a good quality/cost point.

Prompt adherence: the reward barely tracks how well the delivery matches the written direction (r≈0). If matching the performance direction matters, fold prompt-sim into the reward or select on it — best-of-N by reward alone does not improve it. And drop the random emotion reference: it measurably pulls the delivery away from the prompt.

1,597 scored takes · 25 prompts × 2 styles × 2 refs × 16 · no-ref temp 0.8 / rep 1.1 · emo-ref temp 1.0. Audio grids linked at top.