ElevenLabs gives you one of the most generous free tiers in AI audio: 10,000 text-to-speech credits every month, no credit card required. That is roughly ten minutes of finished narration — enough for several short videos or a couple of podcast intros. This guide walks through the exact workflow for creating AI voiceovers on the free plan without wasting a single character.
What the free plan actually includes in 2026
Sign up at elevenlabs.io and you land on the Free tier with these limits:
- 10,000 TTS credits per month, refreshed monthly and non-transferable — unused credits do not roll over.
- About 10 minutes of audio per month, since one credit roughly equals one character on the Multilingual v2 model (around 1,000 characters per spoken minute, or about 1,500 words of script).
- Access to the public voice library — hundreds of pre-made voices with filters for gender, age, accent, and use case.
- Voice Design (around 350 credits per generation) for building a custom voice from a text description.
- Sound Effects at about 50 credits each, plus limited Music generations and Studio (up to 3 long-form projects).
- Speech-to-Text (Scribe) with 92-language support and a separate Voice Isolator pool of 60,000 credits.
Two restrictions matter. First, the free plan does not grant commercial-use rights — if you monetise the content, upgrade to Starter (about $6/month) which unlocks commercial use. Second, voice cloning unlocks on paid tiers; the free plan is for pre-made and designed voices, not cloning your own.
Step 1: Sign up and set your script budget
Create your free account with an email — no card details are asked for. Before generating anything, budget your 10,000 characters. A practical rule: 1,000 characters yields roughly one minute of speech. A 60-second Short needs 900–1,100 characters. Plan your month backwards from the scripts you need, and always keep a 10–15% credit buffer for regenerating lines that come out wrong.
Free plan tip: choose the cheaper models where quality allows. Multilingual v2 costs 1 credit per character; the Flash and Turbo v2.5 models cost 0.5–1 credit per character depending on settings, so the same script can cost nearly half as much.
Step 2: Pick the right voice for your content
Open Text to Speech in the dashboard and browse the voice library. Use the filters instead of auditioning voices randomly — they save credits you would otherwise burn on test generations. For narration, voices tagged "narration" or "audiobook" with deep, steady delivery perform best; for Shorts and Reels, energetic conversational voices tagged "entertainment" or "social media" hold attention better.
Listen to each voice's samples before selecting it. When you find a contender, generate a short test paragraph (your actual opening lines, not lorem ipsum) before committing to the full script. A voice that sounds great on marketing copy can sound wrong on your material, and discovering that after a 3,000-character generation is expensive on a free quota.
Step 3: Choose the model deliberately
ElevenLabs currently offers three main model families, and the choice affects both credit cost and output character:
- Multilingual v2 — the most consistent and reliable option, 29 languages, no SSML support. Best default for polished voiceovers.
- Flash / Turbo v2.5 — faster, good quality, and the only models that accept SSML-style pause and pronunciation tags like
<break time="1.5s" />(max 3 seconds per break) and<phoneme>. Cheaper per character too. - Eleven v3 — the most expressive model with emotion tags for tone shifts, but less predictable; expect retakes, which cost credits on a free plan.
For a first voiceover on the free plan, Multilingual v2 is the safe pick. Switch to Flash or Turbo when you need pause control or want to stretch your monthly credits.
Step 4: Tune the voice settings
Below the text box you will find four sliders that matter more than the model choice for final quality:
- Stability — high values (0.75–0.85) give consistent, professional delivery; lower values (0.3–0.5) add variation and energy but risk odd emphasis.
- Similarity — how strictly the output sticks to the voice sample; keep it high (0.85–0.9) for narration.
- Style — exaggeration of the voice's delivery style; keep near 0 for professional work, raise to 0.5+ for energetic creator content.
- Speed — playback speed; 1.0 is natural, 0.95 can add gravitas to documentary narration.
Tried starting points: natural/professional voiceovers work well at stability 0.8, similarity 0.9, style 0.1, speed 1.0. Conversational creator voices sit around stability 0.55, similarity 0.85, style 0.35, speed 1.0. Test on one paragraph first — regenerating a full script after the fact is the fastest way to burn your monthly quota.
Step 5: Format your script for speech
Write your script the way you want it to sound, not the way you would write prose. Use short sentences and paragraph breaks — a blank line between paragraphs produces a natural 0.3–0.5 second pause on the v2 models. Spell out numbers and abbreviations the way they should be pronounced ("2026" reads fine; "Q3" may not). Avoid ellipsis as a pause device — it can be vocalised as an actual sound on some models.
Keep each generation under your remaining budget and split long scripts into sections. Generating in chunks lets you re-record one bad paragraph instead of the entire script, and it gives you natural edit points for video.
Step 6: Generate, review, and download
Paste your script, hit generate, and listen to the full output before downloading — never bulk-download and review later. Common issues on the free tier: mispronounced names, rushed pacing, or emphasis on the wrong word. Fix these by rewording the sentence slightly rather than regenerating the same text, which tends to produce the same result.
Download as MP3 (free tier exports at 128 kbps; 192 kbps is gated to paid plans). If the voiceover goes into a video, import it into your editor and add background music at about −20 dB below the narration so it supports rather than fights the voice.
Making the free quota last: a monthly workflow
Treat 10,000 characters as a production budget and it goes far:
1. Batch scripts monthly — write all scripts at the start of the month, then record in one session so you learn exactly how much each format costs.
2. Test cheap, commit once — audition voices and settings on a 200-character sample paragraph; generate the full script only after the sample sounds right.
3. Use Turbo for drafts — generate timing drafts on the cheaper Turbo model, then regenerate final audio on Multilingual v2 only for the version that ships.
4. Post-process pauses — instead of regenerating for pacing, split the MP3 and insert silence with a free audio editor; it costs zero credits.
5. Reuse what works — save your voice + settings combination as a favorite so every project starts from a known-good configuration instead of re-testing.
With this discipline, the free plan comfortably covers one voiceover per week at Short length, or two to three full-length video narrations per month — genuinely useful output for zero dollars.