> For clean Markdown of any page, append `.md` to the page URL.
> For a complete documentation index, see https://docs.sarvam.ai/llms.txt.
> For full documentation content in one file, see https://docs.sarvam.ai/llms-full.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.sarvam.ai/_mcp/server.

# Text to Speech Best Practices

## Writing Text for Synthesis

* Punctuate deliberately - commas and periods control pacing, not just grammar. Under-punctuated text tends to sound rushed.
* Spell out abbreviations and numbers that could be read multiple ways (e.g. dates, currency) if you notice mispronunciation.
* Match the language dropdown to the actual language of your text - a mismatch is the most common cause of mispronounced words.

## Choosing Speed and Tone

Speed doesn't just change playback pace - it changes perceived tone. Slower reads as deliberate and authoritative; faster reads as energetic and casual.

* Use slower speeds for narration, instructional content, and anything where authority or clarity matters.
* Use faster speeds for ads, promos, and social content where energy matters more than precision.

## Keeping a Consistent Voice

* Pick a persona once per project and reuse it across every asset - switching personas mid-project is the most common reason a series of clips ends up sounding disjointed.
* If you'll reuse a persona across Text to Speech and Dubbing, select it in the [Voice Library](/creative-voice-library-browse) once rather than re-picking it in every tool.
* Generate a short test line before committing to a long script - catching a mispronunciation or awkward pacing choice on one sentence is far cheaper than after synthesizing several paragraphs.