Captions
Captions provides the Mirage API for generating videos and speech and adding styled captions to videos.
Verdict
Common use cases
- Transcribe and caption product demo videos
- Generate AI voiceover from script drafts
- Animate headshots into talking testimonials
- Export social clips with branded caption styles
- Monitor video job progress in standup threads
Integration
- Vendor
- Captions
- Category
- other
- Auth
- API_KEY
- Tools
- 8
- Composio slug
captions
Tools
- Add Captions
Start an asynchronous job that transcribes and renders a selected caption style onto an uploaded or existing completed video. Use CAPTIONS_GET_VIDEO_STATUS to monitor the returned job, then CAPTIONS_DOWNLOAD_VIDEO when it completes. This ca
- Download Video
Download the MP4 content of a completed video job and return an offloaded file reference. Use CAPTIONS_GET_VIDEO_STATUS first to confirm the job is COMPLETE.
- Generate Speech
Generate WAV speech synchronously from text using a known Mirage voice ID and return an offloaded audio file reference. This call consumes paid Mirage credits and cannot be cancelled or deleted once submitted.
- Generate Video
Start an asynchronous Mirage Video 1 job that animates a portrait image using speech audio. Use the returned job ID with CAPTIONS_GET_VIDEO_STATUS. This call consumes paid Mirage credits and cannot be cancelled or deleted once submitted.
- Get Caption Template
Retrieve one caption style by ID, including its name and preview video URL.
- Get Video Status
Retrieve the current state, progress, and error details for a video generation or captioning job.
- List Caption Templates
Return one page of available caption styles, including names and preview video URLs.
- List Videos
Return one page of video jobs owned by the connected Mirage account, including each job's status and progress.
Setup
Setup guide
- 11. In Switchy, open Settings and navigate to the Integrations page. 2. Find Captions in the MCP directory and click Connect. 3. You'll be prompted to paste a Captions API key; retrieve yours from the Captions dashboard under API Settings. 4. Paste the key and click Authorize. 5. Switchy confirms the connection and lists the eight tools now available. 6. Open any Project and type '@Captions list videos' to verify the integration works — you should see recent jobs from your account. 7. To test caption rendering, upload a short MP4 to the thread, then prompt '@Captions add captions to this video using template ID 12345' (substitute a real template ID from the list). 8. Poll job status with '@Captions get video status for job abc123' until it returns COMPLETE, then download the captioned file.
What teammates see: by default, memories from Captions are scoped to the Project (PROJECT visibility) - you can mark any memory PRIVATE or share it ORG-wide.
Works well with
Top models
Compatibility data appears once enough Projects have used this MCP together with a given model.
How Switchy teams use it
Starter prompts
List Caption Styles
@Captions list all caption templates and show me the preview URLs for each style.
Add Captions to Video
@Captions add captions to the video I just uploaded using the 'Bold Yellow' template. Start the job and give me the job ID.
Generate AI Speech
@Captions generate speech from this script using voice ID 'en-US-neural-female-1' and return the audio file.
Animate Portrait Video
@Captions generate a Mirage video using the portrait image I uploaded and the audio file from the previous message. Return the job ID.
Check Job and Download
@Captions get the status of job abc123. If it's complete, download the video and attach it here.
Example outputs
Illustrative - representative of the model's voice and quality, not literal recordings.
@Captions list my recent video jobs and show me which ones are ready to download
Found 7 video jobs in your Captions account. 4 are complete: 'product-demo-v2.mp4' (captioned, 2m 14s), 'testimonial-cut.mp4' (captioned, 1m 03s), 'mirage-avatar-intro.mp4' (AI-generated, 45s), and 'explainer-draft.mp4' (captioned, 3m 22s). 2 jobs are still processing (estimated 3-5 minutes remaining). 1 job failed due to an unsupported codec. All completed videos are ready to download via the Download Video tool.
This example demonstrates read-only job monitoring. The MCP returns job metadata (status, duration, type) but does not automatically download files. Users must explicitly call Download Video for each completed job they want to retrieve. Useful for batch workflow orchestration.
@Captions add captions to my uploaded video using the 'Bold Yellow' template, then download the result when it's done
Started captioning job #vid_8x3k9m2 using the 'Bold Yellow' template. The job is currently processing (12% complete). Transcription and caption rendering typically take 1-2 minutes per minute of video. I'll monitor the job status and download the MP4 once it reaches COMPLETE. Note: this action consumes Captions API credits and cannot be reversed once the job starts.
This example shows the two-step write workflow: start the async captioning job, then poll for completion before downloading. The MCP does not auto-retry failed jobs or refund credits. Users should confirm the correct template ID before triggering, as caption style cannot be changed mid-job.
@Captions generate a 30-second AI avatar video using voice ID 'emma-professional' reading this script, then summarize what the avatar will look like
Started Mirage Video job #vid_9p4n7q1 with voice 'emma-professional'. The job will animate a portrait using the provided 30-second script. Mirage Video 1 produces a talking-head style avatar with lip-sync and natural head movements. Processing takes 3-5 minutes. This job consumed 30 Mirage credits (non-refundable). Once complete, the output will be a 1920×1080 MP4. I'll notify you when the video is ready to download and review.
This example pairs the MCP's Generate Video action with AI reasoning about the output format. Mirage credits are deducted immediately when the job starts, regardless of success. Users must supply a valid voice ID (obtained separately) and should preview voice samples before committing to paid generation.
Use-case deep-dives
When Captions MCP beats manual video editing for client content
A 6-person creative agency ships 40+ social clips per week for three B2B clients. The team records long-form webinars and podcasts, then needs vertical cuts with burned-in captions for LinkedIn and Instagram. Captions MCP is the right call here: one AI can monitor job status across all eight tools, queue caption renders in batch, and hand off download links to the project manager without a human toggling between the Captions dashboard and Slack. The API key auth means the whole team shares one account. The threshold: if you need custom caption fonts or brand-specific animation timing beyond the template library, you'll still export to Premiere. But for teams shipping volume over pixel-perfect control, this MCP turns a 90-minute manual process into a 12-minute supervised queue.
Generate talking-head testimonials from text quotes with Mirage Video
A 3-person SaaS marketing team collects written customer quotes but lacks video budget for on-site shoots. They upload headshots and use the Generate Video tool to animate portraits speaking the testimonial script, then add captions with a branded template. Captions MCP works if you're producing 5-10 testimonials per quarter and the slightly synthetic Mirage aesthetic fits your brand (think modern SaaS, not luxury retail). The AI can batch-generate speech, queue video jobs, and monitor status without manual polling. The trade-off: Mirage credits are pay-per-use and non-refundable, so you can't experiment cheaply. If your brand requires documentary realism or you're producing one testimonial every six months, hire a videographer. For teams treating testimonials as a repeatable content format, this MCP makes the workflow scriptable.
When caption templates scale onboarding videos across regions
A 12-person ops team maintains 30 onboarding videos and needs Spanish and French versions for new regional hires. They record English narration, generate translated speech with the Generate Speech tool, then apply captions in the target language using the template library. Captions MCP is the right fit if your videos are under 10 minutes, your caption needs are stylistic (not legal compliance), and you're updating content monthly. The AI can list templates, preview styles, and queue caption jobs without a human clicking through the web UI. The boundary: if you need certified accurate captions for accessibility compliance or videos longer than 15 minutes, the async job model gets cumbersome and you'll want a dedicated localization vendor. For mid-frequency internal content where speed beats perfection, this MCP turns a multi-day turnaround into same-day delivery.
Frequently asked
What does the Captions MCP do in Switchy?
It lets your team generate AI speech, animate portrait images into talking-head videos, and burn captions onto existing footage — all from Switchy's chat interface. The MCP wraps Captions' Mirage API, so you can kick off video jobs, check their status, and download the final MP4 without leaving your workspace. Each operation consumes credits from your Captions account.
Do I need a paid Captions account to use this MCP?
Yes. You need an active Captions account with Mirage credits and an API key. The MCP authenticates using that key, and every speech or video generation call deducts credits from your balance. Free-tier limits or trial quotas apply exactly as they would if you called the API directly.
Can the Captions MCP edit existing videos or only generate new ones?
It can burn captions onto videos you've already uploaded to Captions, but it won't trim, splice, or re-encode arbitrary footage. For new content, it generates talking-head clips from a still image and audio file. If you need timeline editing or multi-track compositing, use a dedicated video tool and then add captions as a final step.
How does this compare to uploading videos to Captions manually?
The MCP automates the upload-job-download cycle inside Switchy, so you can script repetitive workflows or chain video generation with other MCPs. You lose the Captions web UI's preview player and manual style picker, but you gain the ability to trigger jobs from chat, monitor them programmatically, and integrate the output into broader automation.
Who on the team should connect the Captions MCP?
Whoever owns the Captions API key and understands your credit budget. Because every generate call charges your account immediately and can't be cancelled, you want one person responsible for monitoring usage. Other teammates can invoke the tools through Switchy, but the connector controls the purse strings.