Use identical ElevenLabs voice and settings for every chunk; reduce expressive variance. Chunking: render 20–45s chunks split at sentence/paragraph boundaries; add SSML breaths or 200–300 ms between paragraphs. Stitching: apply 100–200 ms overlap with 10–20 ms fade ramps to crossfade joins. Post chain (44.1kHz/24‑bit): high‑pass 80 Hz → de‑click/denoise → normalize to -16 LUFS → gentle comp (2:1, 10/100 ms) → de‑esser → limiter -0.1 dB → dither if reducing bit depth. Adjust overlap to mask prosody jumps.