Almost every free, browser-based text-to-speech tool caps how much text you can convert in one generation — commonly somewhere around 5,000 characters, or roughly 800-1,000 words. For a blog post or a short script, that's plenty. For a book chapter, a long report, or a full podcast episode, it isn't, and you'll need a workflow for producing longer audio in sections.
Why the limit exists
Generating speech from text is computationally expensive, and it scales with length — a longer input takes proportionally longer to synthesize and uses more server resources per request. A per-generation cap keeps individual requests fast and keeps the service usable for free, without a single long request tying up resources other users are waiting on. It's a practical tradeoff, not an arbitrary restriction.
A workflow that actually works
1. Split at natural breaks, not arbitrary character counts
Don't just cut your text every 5,000 characters regardless of what's happening in the sentence. Split at paragraph or section boundaries — the end of a scene, the end of a chapter section, a natural pause point in a script. This matters because a section that starts or ends mid-sentence often gets read with slightly off pacing at the cut point, since the model has no context about what came immediately before or after within that chunk.
2. Keep voice and settings identical across every section
This is the single most important step for a consistent-sounding final file. Write down (or screenshot) the exact voice, pitch, rate, and volume settings you use for the first section, and use precisely the same values for every subsequent section. Even a small, unintentional settings change between sections is often audible when they're stitched together.
3. Generate each section separately
Work through your split sections one at a time, downloading each as its own MP3 file with a name
that preserves the order (chapter1-part1.mp3, chapter1-part2.mp3, and so
on).
4. Combine in an audio editor
Free tools like Audacity handle this well: import each MP3 as a separate track in sequence, and export the combined result as one file. Add a very short silence (a quarter to half a second) between sections if the join sounds abrupt — this alone often fixes most of the "stitched together" feeling.
Where this is worth the effort — and where it isn't
It's worth it for content you're actually publishing: an audiobook chapter, a long-form podcast episode, a full report someone will listen to end-to-end. It's usually not worth the effort for a quick draft check — if you just want to hear how a long document sounds roughly, generating the first section alone is often enough signal without doing the full stitching workflow.
