Comparison

Subtitles King vs Deepgram

Deepgram is for streaming voice. Subtitles King is for finished video.

TL;DR

At a glance.

FeatureSubtitles KingDeepgram
Free tier✓ 100 MB videos$200 credit
Cost (paid)Pro waitlist~$0.43/hr (Nova)
MCP server
Self-host✓ free✓ enterprise only
Real-time streaming✓ best in class
Burned video output
Open source
Latency to first textMinutesMilliseconds

When Deepgram wins

  • Best-in-class real-time streaming transcription latency.
  • Custom model training and fine-tuning on your data.
  • Strong contact center / live caption use cases.
  • Enterprise on-prem deployment available (paid).

When Subtitles King wins

  • Free open-source self-host — no enterprise contract required.
  • Returns the finished, subtitle-burned video, not just transcript text.
  • MCP-native for AI agent integration.
  • Simpler API surface for the common "subtitle this video" use case.

Code

Integration, side by side.

Subtitles King
# Subtitles King: 3 lines
curl -X POST https://brains.subtitlesking.com/presign \
  -H "X-API-KEY: $KEY" -d '{"filename":"clip.mp4"}'
curl -X POST "$URL" -F "file=@clip.mp4"
curl -X POST https://brains.subtitlesking.com/queue -d '{"authToken":"$T"}'
Deepgram
# Deepgram: transcribe and you still need to render captions
from deepgram import DeepgramClient, PrerecordedOptions
import httpx

dg = DeepgramClient("your_key")

with open("clip.mp4", "rb") as audio:
    payload = {"buffer": audio}
    options = PrerecordedOptions(
        model="nova-2",
        smart_format=True,
        utterances=True,
    )
    res = dg.listen.rest.v("1").transcribe_file(payload, options)

# Now build SRT from utterances and burn with ffmpeg...

FAQ

Common questions.

Should I use Deepgram for live captions and Subtitles King for VOD?+

Yes — that's probably the right split. Deepgram is the right choice for sub-second streaming use cases. Subtitles King is the right choice for "I have a video file, I want subtitles burned in".

Does Subtitles King support custom vocabulary like Deepgram?+

Not currently. We use Whisper large which has very strong general-domain accuracy. If you need custom vocab, Deepgram or self-hosted Whisper with biasing are better fits today.

Try it. Three lines, free.