Murf AI

Generate studio-quality AI voice overs with Murf AI: agents turn text into natural speech across a catalog of voices, re-voice existing recordings, translate narration scripts, and dub audio or video into new languages, with output stored durably in FlowRunner.

6 actions API key available

Murf AI website ↗ Platform Documentation ↗ Capability data verified 2026-07-25

A finished product video lands in shared storage marked ready for localization

Create Dubbing Job submits the media URL with the quarter's target locales

Get Dubbing Job Status polls until the job reports COMPLETED, copying each dubbed track and its SRT subtitle file into durable FlowRunner storage

Per-locale preview links post to the localization channel with credits consumed and remaining

A native-speaking reviewer for each locale approves the dub or flags mispronounced product terms

Approved tracks upload to the video platform; flagged locales queue for a corrected rerun

The release tracker updates with which locales shipped and which are in revision

What This Integration Enables

Murf AI turns voice production from a scheduling problem into a pipeline step. The teams that get the most from it are not replacing a voice actor on one video; they are the ones with a localization matrix, twelve videos times eight locales, that was never going to get recorded by humans at all. FlowRunner agents run that matrix: translate the script, render it in a multi-native voice, dub the finished video, and store every output durably so the links do not die with Murf's temporary URLs. - Generate Speech returns per-word timings alongside the audio, the raw material for caption and lip-sync workflows - Translate Text is tuned for narration, so translated scripts sound spoken rather than literal - Change Voice re-voices existing recordings while keeping wording and timing, which makes brand-voice consistency retroactive - Dubbing runs asynchronously with polling handled inside the flow, and completed tracks include matching SRT subtitles - Every customer-facing output pauses for [human-in-the-loop](/concepts/human-in-the-loop) review, in the reviewer's own language

Without FlowRunner

Voice work is a booking Narration waits on studio time, talent availability, and a per-language invoice

Localization ships blind Dubbed content goes out in languages nobody on the team can check, or it does not go out at all

Old recordings age out A product rename makes forty demo videos wrong, and re-recording them all is nobody's budget

With FlowRunner

Voice is a pipeline step Scripts become studio-quality narration in the flow that produces the video, in the voice the brand always uses

Every locale gets a named reviewer Dubs and translations pause for a native speaker's approval before anything reaches customers

Re-voicing is a batch job Change Voice re-renders existing recordings with current wording and a consistent voice, timing intact

Use Case Scenarios

E-learning narration straight from the script sheet

Course scripts live row by row in [Google Sheets](/integrations/google-sheets). The agent pulls each row, calls Generate Speech with the course's designated voice and a Conversational style, and writes the stored audio URL back to the row. For the Spanish and German editions, Translate Text runs first and a multi-native voice renders each locale. The instructional designer reviews the batch through preview links in [Slack](/integrations/slack) before the module builder assembles anything. A course revision becomes an afternoon of review instead of a week of recording.

One product video, eight markets

When the quarterly launch video is finalized, the agent submits it with Create Dubbing Job for all target locales and polls Get Dubbing Job Status until every track completes, storing dubbed media and SRT files durably. Each locale's preview goes to its regional marketer for approval. Approved tracks publish to [YouTube](/integrations/youtube) as localized versions with subtitles attached, and the campaign tracker in [Notion](/integrations/notion) updates per market. The launch stops being English-first with translations to follow eventually.

A back catalog re-voiced into one brand voice

Years of demo videos and podcast intros were recorded by whoever was available. The agent walks the archive, runs Change Voice on each recording with the brand's chosen voice while retaining original timing, and files the re-rendered audio next to the source. Marketing reviews the queue in batches, and approved versions replace the originals in the help center. The catalog sounds like one company without a single re-recording session.

Human-in-Loop Highlight

Create Dubbing Job is where this connector spends money on output the team often cannot verify: each job draws credits per locale, and the dubbed tracks ship in languages the people who built the flow may not speak. An agent that dubs and publishes in one motion is an agent that can mispronounce your product name in Japanese, at scale, with nobody noticing until a customer does. So the dubbing flow never publishes its own output. It completes the job, stores the tracks, and asks the named reviewer for each locale: "French dub complete for the onboarding video, 4 minutes 12 seconds, credits remaining on the account attached. Approve for publish?" A rerun costs credits; a shipped mispronunciation costs credibility in a market you are trying to enter. The reviewer decides which price gets paid.

Agent processes routinely

Detects exception requiring judgment

Clear match Continues automatically

Ambiguous Routes to human via preferred channel

Human decides

Agent resumes with decision

Agent Capabilities

6 actions

Speech Synthesis

Generate Speech Converts text into studio-quality speech and stores the audio as a FlowRunner file. Supports voice styles, multi-native locales, pitch and rate adjustment, target duration, and output in MP3, WAV, FLAC, PCM, OGG, A-Law or U-Law at sample rates up to 48000 Hz. Returns per-word timings for caption and lip-sync work plus the account's remaining character quota.
Change Voice Re-voices an existing recording with a Murf voice while keeping the original wording and timing. Takes a publicly reachable audio URL, optionally retains the source prosody and accent, accepts a manual transcription for noisy sources, and saves the converted audio to FlowRunner file storage.

Voices

List Voices Retrieves the full voice catalog available to the account: voice identifier, display name, gender, primary locale, accent, supported speaking styles, and the additional locales each multi-native voice speaks. The lookup that drives voice selection logic in flows.

Translation

Translate Text Translates a batch of text snippets into a target locale with an engine tuned for voice-over narration. Returns each source paired with its translation plus character counts and credits consumed. The first half of every localized narration pipeline.

Dubbing

Create Dubbing Job Submits an audio or video file for AI dubbing into one or more target locales. Asynchronous: returns a job identifier immediately, accepts a queue priority, an optional source locale, and an optional webhook for completion. The credit spend this page's human gate governs.
Get Dubbing Job Status Checks a dubbing job's progress: overall status, credits consumed and remaining, and per-locale download URLs for dubbed media and matching SRT subtitles. Can copy completed tracks into FlowRunner file storage so links survive Murf's temporary URL expiry.

Frequently Asked Questions

What can FlowRunner do with Murf AI?

FlowRunner agents can run Generate Speech, Change Voice, and List Voices in Murf AI, plus 3 more actions.

Does connecting Murf AI to FlowRunner require OAuth?

No. Murf AI connects to FlowRunner with an API key, no OAuth flow required.

Can Murf AI trigger a FlowRunner workflow automatically?

Murf AI doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.

Start building with Murf AI

$100 in credits. No card required. Connect in minutes.

Start Building Free Talk to us