Step 1
Bring the source
Type an idea, use browser speech recognition, or upload supported audio for transient transcription.
AI content repurposing
Turn a podcast into social media posts you have actually read
Sign in, upload an episode file, and DraftLark transcribes it, puts the transcript in front of you to correct, and reshapes it into a post, a thread, a newsletter draft, and a five-slide carousel. The audio itself is passed through for transcription and never stored.
- No account needed to draft
- Every output editable
- Source snapshot kept with each version
What you actually get
The real path from an episode file to drafts you can edit.
- Upload the episode audioMP3, M4A, WAV, WebM, and OGG files up to 15 MB are accepted. The type and size are checked in the browser before anything uploads, so an unsupported file is rejected immediately rather than after a long transfer.
- Transient transcriptionThe file is sent to the configured Gemini service for transcription for that request only. DraftLark does not save the audio afterwards — only the transcript, and only if you save or generate.
- Correct the transcript firstThe transcript lands in the editable source field. Fix names, terms, and anything the model heard wrong before you generate, so the drafts inherit corrections rather than mistakes.
- Four drafts per episode, versionedEach generation is saved as a version with the exact transcript it used and a source fingerprint, so you can tell which episode text produced which set of posts.
How the workflow works
Three steps in one screen. Nothing here is a simulated demo — it is the studio you land in when you open DraftLark.
Step 2
Choose the formats
Generate bounded post, thread, newsletter, and static carousel drafts using the selected voice controls.
Step 3
Review and export
Compare drafts with the retained source snapshot, edit every output, and export text or static HTML only after human review.
Where DraftLark stops
Useful software starts with an accurate promise. These limits apply before you begin.
- The 15 MB per-file limit means long episodes need to be split or compressed first. Resumable uploads and background transcription of long recordings are not implemented.
- There are no speaker labels, timestamps, or chapter detection, and there is no podcast feed, RSS, or hosting-platform integration.
- Signing in with Google is required for upload and transcription; the signed-out editor is typed input only.
- DraftLark output is a draft, not fact-checked, original-by-guarantee, or publication-ready content.
- Uploaded audio is sent transiently for transcription and is not retained by DraftLark.
- The product does not claim scheduling, publishing, video generation, social-platform integrations, or a long-form blog output.
Podcast to Social Media Posts: common questions
Is my episode audio kept after transcription?
No. The audio is sent transiently to the model for that one transcription request and is not stored by DraftLark, in the browser or in Firebase. The transcript is stored only when you save a project or generate from it.
How long an episode can I upload?
Whatever fits in 15 MB for the file format you use. Longer episodes need to be compressed or split into segments before upload.
Do the posts quote the episode accurately?
The model is instructed to use only the transcript and to add nothing that is absent from it, and the transcript is stored beside each version. Automatic speech recognition still mishears words, so correct the transcript before generating and read the drafts before posting.