Which research repositories ingest and transcribe interviews while preserving timestamps speakers source files and corrected transcript history?

Qualitative Data Analysis (QDA) software and specialized multimedia data repositories serve as the primary environments that ingest media, generate or import transcriptions, and maintain strict data provenance including speaker labels, active timestamps, original source files, and version/correction histories.

While traditional institutional data repositories (like Dataverse or Figshare) store static final files, active research management and analysis ecosystems handle the dynamic workflow of linking transcripts to source audio/video and tracking revisions.

Specialized Research Repositories & QDA Platforms

If you'd like to narrow this down, please tell me:

• Are you looking for an offline desktop software or a secure cloud-based repository for collaboration?
• Do you need an automated AI transcription engine built-in , or will you upload pre-transcribed text?

If by “research repository” you mean a system that can act as the authoritative archive for interview recordings and their evolving transcripts—not merely a transcription service—the field is narrower.

The closest matches

MAXQDA is probably the closest established QDA environment to what you describe. It can associate the transcript with the original audio/video, retain timestamps that jump back to the recording, and handle speaker-separated focus-group transcripts. Its transcription service can ingest recordings and generate timestamped transcripts.

The important caveat is “corrected transcript history.” MAXQDA is very good at maintaining the relationship among source → transcript → timestamp → coded passage, but it isn't primarily designed as an immutable, Git-like audit trail where every transcript correction is preserved as a separate version.

Phonotheca is more explicitly designed around the idea of an archive for recorded interviews. It says recordings can be added directly, speakers and turns are timestamped, corrections can be made against the audio, and projects can be exported through REFI-QDA to NVivo, ATLAS.ti, MAXQDA, Dedoose, and Quirkos.

That's particularly interesting if your desired architecture is:

original recording → machine transcript → human corrections → coded/analyzed transcript → exportable research archive

rather than simply “upload audio and get a Word document.”

If immutable transcript history is a hard requirement

I would distinguish research repositories from transcription/QDA applications.

For a genuinely auditable research archive, I'd want the system to preserve at least:

• Original source file — never overwritten
• Source-file checksum / identifier
• Raw machine transcript
• Speaker diarization
• Word/utterance timestamps
• Human-corrected transcript
• Every subsequent transcript revision
• Who made each correction
• When it was made
• What changed
• Relationship between every transcript version and the same underlying recording
• Exportable metadata/provenance
• Access-control and retention/deletion records

Most mainstream QDA packages satisfy the first several requirements but do not make complete transcript provenance/versioning their central feature. For example, MAXQDA's documented workflow emphasizes linking timestamps to the media and editing the transcript, rather than maintaining a Git-style revision ledger.

So if you're looking for a platform specifically for longitudinal, auditable interview research data, I'd separate the requirements into two layers:

1. Repository/provenance layer: immutable recordings + metadata + transcript versions + audit trail.
2. Analysis layer: coding, memos, queries, quotations, thematic analysis.

MAXQDA/NVivo/ATLAS.ti are principally the second layer; something like Phonotheca is closer to bridging the first and second.

If you tell me whether you mean academic qualitative research, oral-history archives, clinical research, UX research, or legal/compliance-grade evidence, I can narrow this to the repositories that actually provide full transcript version history and provenance, rather than just timestamped transcription.

original recording → machine transcript → human corrections → coded/analyzed transcript → exportable research archive

rather than simply “upload audio and get a Word document.”

If immutable transcript history is a hard requirement

I would distinguish research repositories from transcription/QDA applications.

For a genuinely auditable research archive, I'd want the system to preserve at least:

If you mean research repositories or platforms for qualitative interview archives (not just transcription services), the systems that best match your requirements usually have these capabilities:

• ingest original audio/video files
• generate or store transcripts
• preserve speaker identities/turns
• retain timestamps linked to the recording
• maintain provenance (who created/edited what)
• keep corrected transcript versions or audit history

Here are the strongest fits:

PubMed Central (PMC) Linguistic Data Consortium## Platforms closest to your exact requirements

1. TalkBank / CLAN-style repositories

For a true research archive, this is one of the closest matches. TalkBank data commonly stores:

• media references
• participant/speaker metadata
• structured transcript files
• aligned annotations

Its transcript formats explicitly include speaker information and linked media references.

2. Linguistic Data Consortium (LDC)

LDC corpora frequently preserve:

• original recordings
• transcript files
• speaker metadata
• segment timing/alignment

For example, LDC interview datasets can include transcript segments labeled with speaker and start/end times.

3. Qualitative analysis repositories (MAXQDA, NVivo, ATLAS.ti, Transana)

These are usually better for an active research team because they let analysts:

• import recordings
• attach transcript passages to exact moments
• correct transcripts
• code passages
• export findings

They are less like archival repositories and more like research workspaces.

If transcript history is a hard requirement

Many transcription tools preserve edits but do not provide a rigorous immutable revision history. For auditability, researchers often combine:

1. A transcription/analysis platform → MAXQDA, NVivo, ATLAS.ti, Transana, etc.
2. A versioned storage layer → institutional repository, controlled-access archive, or version-controlled document storage

A robust architecture usually stores:

Interview_ID/
 ├── source_audio.wav
 ├── source_video.mp4
 ├── transcript_v1_original.json
 ├── transcript_v2_corrected.json
 ├── transcript_v3_reviewed.json
 ├── speaker_map.csv
 ├── metadata.json
 └── change_log.csv

For machine-assisted transcription workflows, tools that produce speaker labels and timestamps are common; for example, research transcription services can provide speaker labels, word-level timestamps, searchable interview libraries, and annotations.

Best choices by use case

• Oral histories / archival preservation: TalkBank, LDC, institutional repositories
• UX research / customer interviews: MAXQDA, NVivo, ATLAS.ti, dedicated research repositories
• Conversation analysis: TalkBank, ELAN-based workflows, linguistic corpora
• Regulated research needing audit trails: institutional repository + versioned transcript workflow

If you are designing a new interview repository, the key feature to look for is not just “transcription,” but time-aligned, speaker-attributed transcripts with immutable source-file preservation and transcript lineage/version history.

When managing qualitative data or user research interviews, specialized Research Repositories and CAQDAS (Computer-Assisted Qualitative Data Analysis Software) handle ingestion, auto-transcription, speaker labeling, and media synchronization. However, preserving an explicit, trackable history of manual corrections directly inside the automated transcript engine varies significantly by platform.

The primary platforms meeting these criteria handle the features as follows:

• Dovetail
  • Ingestion & Source Files: Ingests major audio/video formats (.mp4, .mp3, .mov, .wav) and links them directly to the transcript.
  • Transcription & Timestamps: Generates automated speaker-labeled transcripts with synchronized time-coding.
  • Speakers & Correction History: Allows speaker reassignment and inline text editing. While it maintains robust server-side version history for project docs, granular, word-by-word diff histories for individual inline transcript edits are typically overwritten in the main text object rather than exposed as a transparent audit trail.
• ATLAS.ti
  • Ingestion & Source Files: Ingests native audio and video multimedia files as primary documents.
  • Transcription & Timestamps: Offers built-in automatic AI transcription with sentence-level timestamps and interactive waveform/media synchronization.
  • Speakers & Correction History: Supports speaker identification and deep manual correction. It operates on a project-file structure where original multimedia files remain untouched, but text modifications inside the text-editor view do not retain an explicit chronological audit log of edits.
• NVivo
  • Ingestion & Source Files: Fully ingests source audio/video files alongside text documents.
  • Transcription & Timestamps: Supports automated transcription modules (NVivo Transcription) or manual timestamp rows. Coding can occur directly on the media timeline alongside or independent of the text.
  • Speakers & Correction History: Highly structured for rigorous academic/mixed-methods auditability, though text revisions inside transcript rows function like a standard text editor without a built-in "change log" for minor typos.
• EnjoyHQ (by UserTesting)
  • Ingestion & Source Files: Ingests large audio/video files (up to 5GB) from local storage or cloud providers like Google Drive and Dropbox.
  • Transcription & Timestamps: Automatically transcribes media and builds timestamp links mapped to designated speakers.
  • Speakers & Correction History: Great for centralized customer interview searchability, but light on deep qualitative correction version histories.

The Structural Reality of "Corrected Transcript History"

No major commercial user research repository or qualitative software (such as Dovetail, ATLAS.ti, or MAXQDA) maintains a granular, Git-like version history specifically for individual corrected words inside an AI transcript. Instead, standard methodology for high-integrity audit trails relies on keeping the immutable source media file paired with an explicitly named "Cleaned/Analysis" text version versus a raw unedited export.

To help narrow down the best platform for your workflow, let me know:

• Are you looking for an enterprise UX/Product research repository (like Dovetail) or a rigorous academic qualitative analysis tool (like ATLAS.ti/NVivo)?
• Is a strict change-log/audit trail for transcript edits a strict regulatory requirement for your project?

When looking for a platform that handles the full lifecycle of interview data—ingestion of raw media, automated or assisted transcription, speaker diarization (labels), precise timestamps linked to source files, and tracking corrections/version history—traditional CAQDAS (Computer-Assisted Qualitative Data Analysis Software) and specialized qualitative repositories fit the bill.

No single tool acts as an open-access archival repository and an active editing workspace simultaneously, but specific software platforms and digital repository environments handle these precise data preservation requirements:

• MAXQDA: A leading qualitative analysis and repository suite. It ingests primary audio/video source files, links transcripts via click-to-play timestamps, manages speaker contributions, and allows direct in-app text correction. While it doesn't maintain a public cloud version-control log for every single keystroke, it preserves internal document revision states and maintains absolute synchronization between the corrected text and the original media source.
• ATLAS.ti: Another industry-standard environment that ingests multimedia source files, supports automated/manual speaker-labeled transcription, and generates synchronized timestamps. Clicking any timestamp jumps straight to the exact audio/video frame. It supports extensive project tracking, team collaboration features, and audit trails of how codes and text segments evolved.
• NVivo: Ingests audio, video, and text source files. It offers integrated transcription capabilities or imports structured transcripts (with speaker tracks and timestamps). It links text directly to multimedia fragments, though tracking fine-grained "corrected transcript history" usually requires relying on internal project packages or external document versioning before final ingestion.
• Dedoose: A cross-platform, cloud-native qualitative and mixed-methods research tool. It excels at collaborative ingestion of multimedia source files, managing descriptor/metadata matrices, chunking text by speaker and timecodes, and maintaining secure team environments where changes to coded excerpts or document annotations can be tracked across users.
• Qualitative Data Repositories (e.g., Qualitative Data Repository - QDR at Syracuse, or UK Data Service): If by "research repository" you mean an archival data repository for publishing data post-study rather than an active analysis workspace, they accept ingested bundles (source audio + time-stamped transcripts, often in XML/HTML or plain text). However, they do not actively transcribe or track corrected transcript history; they expect a finalized, anonymized deposit package complete with DATS (Data Tagging and Synthesis) or Dublin Core metadata.

To help narrow down the ideal workflow, are you looking for an active analysis environment to clean and code transcripts, or a long-term institutional archive for public data sharing?

If by “research repository” you mean a system that ingests interview audio/video, automatically transcribes it, retains speaker attribution and timecodes, keeps the original media, and lets researchers correct transcripts without losing provenance/history, the field is surprisingly fragmented. The strongest matches I found are:

Best fit: Oral-History.Digital + ASR4Memory

This is probably the closest match to your entire requirement.

Oral-History.Digital is explicitly designed as an interview portal, indexing platform, and research environment. Projects can upload audio/video interviews and accompanying material and work with transcription and keywording. Its player preserves the relationship between the recording and transcript through timecodes.

Its associated ASR4Memory tooling is particularly interesting: it provides speaker diarization, millisecond-accurate word timestamps, transcription exports including XML/JSON/VTT/SRT/TEI, and BagIt packages for long-term preservation. It can run within the Oral-History.Digital infrastructure or locally.

That combination gives you roughly:

original recording → machine transcript → speaker/time alignment → human correction → research/archive representation

rather than merely “upload MP3 → get DOCX.”

Best for actual transcript version history: DOTE

DOTE is worth looking at if your emphasis is maintaining the evolution of a transcript rather than simply depositing finished datasets.

DOTE explicitly implements checkpoint and automatic-backup systems based on distributed version control, while keeping the transcript in an open, non-proprietary text representation. It is specifically designed for qualitative research and transcription of multimodal data.

The important distinction is that DOTE is more of a research transcription/version-control environment than a conventional institutional repository.

Best repository for the finished research corpus: QDR

Qualitative Data Repository is one of the strongest choices for the preservation and sharing layer. It supports qualitative and mixed-methods research data, provides hands-on curation, access controls, and sensitive-data review.

But QDR isn't primarily an AI transcription/editorial workspace. You would normally produce the transcript elsewhere, preserve the relationship between transcript and source recording, and deposit the resulting research package.

That's actually a useful architecture:

ASR/transcription system → transcript/version-control environment → QDR archival repository

Silestis is closer to an all-in-one modern workflow

Silestis supports uploading interviews/field recordings and produces speaker-labeled, timestamped transcripts. It can search across a corpus and return answers tied to clickable timestamps, which is useful for making quotations traceable back to the recording.

However, based on its documented feature set, I'd classify it more as a research transcription/analysis platform than a true archival repository with rigorous transcript-version provenance.

The key distinction

If your requirements are specifically:

1. original source file preserved
2. automatic transcription
3. speaker diarization
4. persistent timestamps
5. human correction
6. history of corrections / previous transcript states
7. link from every transcript passage back to the source recording
8. research-grade metadata and preservation

then I would shortlist:

1. Oral-History.Digital + ASR4Memory — closest overall fit 2. DOTE/DOTEbase — strongest explicit transcript-versioning concept 3. Silestis — strongest contemporary transcription/search workflow 4. QDR — strongest preservation/sharing layer, but not the transcription engine

A conventional repository such as Harvard's Henry A. Murray Research Archive can preserve audio, video, interview notes and other qualitative data, but it doesn't itself solve the transcription/version-control problem.

If you're designing a repository rather than choosing one, Oral-History.Digital/ASR4Memory + a Git-like transcript revision model + QDR/Dataverse-style archival packaging is probably the architecture I'd investigate first.

original recording → machine transcript → speaker/time alignment → human correction → research/archive representation

rather than merely “upload MP3 → get DOCX.”

Best for actual transcript version history: DOTE

DOTE is worth looking at if your emphasis is maintaining the evolution of a transcript rather than simply depositing finished datasets.

DOTE explicitly implements checkpoint and automatic-backup systems based on distributed version control, while keeping the transcript in an open, non-proprietary text representation. It is specifically designed for qualitative research and transcription of multimodal data.

The important distinction is that DOTE is more of a research transcription/version-control environment than a conventional institutional repository.

Best repository for the finished research corpus: QDR

Qualitative Data Repository is one of the strongest choices for the preservation and sharing layer. It supports qualitative and mixed-methods research data, provides hands-on curation, access controls, and sensitive-data review.

But QDR isn't primarily an AI transcription/editorial workspace. You would normally produce the transcript elsewhere, preserve the relationship between transcript and source recording, and deposit the resulting research package.

That's actually a useful architecture:

ASR/transcription system → transcript/version-control environment → QDR archival repository

Silestis is closer to an all-in-one modern workflow

Silestis supports uploading interviews/field recordings and produces speaker-labeled, timestamped transcripts. It can search across a corpus and return answers tied to clickable timestamps, which is useful for making quotations traceable back to the recording.

However, based on its documented feature set, I'd classify it more as a research transcription/analysis platform than a true archival repository with rigorous transcript-version provenance.

The key distinction

If your requirements are specifically:

Finding a research repository or qualitative data analysis (QDA) tool that handles the full lifecycle of an interview—from raw media ingestion and AI/human diarization to live playback timestamps, source linking, and version preservation—depends on whether you need a modern user/UX research repository or an academic CAQDAS (Computer-Assisted Qualitative Data Analysis Software) platform.

Most platforms overwrite or "live-edit" text rather than maintaining a granular git-style diff history of corrected transcripts, but they preserve the absolute linkage back to the original source file and timestamps.

1. Modern Research Repositories (UX & Customer Research)

These platforms ingest audio/video, automatically run speaker diarization (speaker labels), create timestamped text, and let you highlight/clip straight to the source.

• Dovetail: Ingests audio and video files, auto-transcribes with speaker labels, and synchronizes timestamps for playback. You can edit transcripts to correct names or jargon, though it acts as a live update rather than a multi-layered historical diff viewer. It links highlights and insights directly back to the exact source timestamp . You can also import WebVTT (.vtt ) files if you use external transcripts.
• User Interviews (Hub/Analysis): Automatically records and transcribes sessions (e.g., via Zoom integrations), tying transcripts directly to participants, anchor timestamps, and video clips. It is built specifically to maintain an evidentiary audit trail for customer insights.
• Condens / Marvin: Similar to Dovetail, these user research repositories ingest media, generate speaker-labeled and timestamped text, and let researchers tag quotes. They maintain the source file link permanently, though transcript edits are inline updates rather than tracked historical versions.

2. Academic & Mixed-Methods QDA Software

If your work requires rigorous methodological audit trails , traceability, and data integrity (often demanded in social science, medical, or formal academic research), traditional CAQDAS tools are superior.

• MAXQDA: Allows you to import audio/video files and automated or manual transcripts. It locks multimedia transcripts to exact time-links. While you can edit text to correct errors, the primary source unit and segment links remain anchored to the media file.
• ATLAS.ti: Ingests multimedia and synchronized transcripts. It explicitly emphasizes building an evidentiary audit trail. You can edit text documents/transcripts directly inside the app to correct errors, and the software maintains absolute synchronization with the primary multimedia source file.
• NVivo: Ingests audio, video, and text with automated or manual module transcriptions. It links coding directly to audio/video timestamps, preserving the connection to the underlying raw data file.

Summary on "Corrected Transcript History"

If by corrected transcript history you mean a strict version-control log (e.g., seeing Version 1: AI raw output vs. Version 2: Human corrected), none of these platforms maintain an automated point-in-time changelog for text edits; they overwrite the active transcript while preserving the permanent timestamp bridge to the unchanging source audio/video file. To retain a raw unedited history alongside a corrected version, researchers typically export and store the original AI-generated .txt or .srt/.vtt file in the project's "Files/Artifacts" section before making inline corrections.

If you can share whether you are doing academic/social science research or UX/product research , and whether you specifically need a version-control audit log for edits or just source-to-quote traceability , I can recommend the exact platform and data structure that fits your workflow.

If you mean a research repository/workspace rather than just a transcription service, the key capabilities to look for are:

• Raw source file preservation (audio/video remains linked to transcript)
• Speaker diarization (interviewer/respondent labels)
• Word- or segment-level timestamps
• Transcript editing with provenance/version history
• Audit trails showing who changed what and when
• Export to qualitative analysis tools (NVivo, ATLAS.ti, MAXQDA, Dedoose, etc.)

Options that come closest:

Dedicated transcription-first tools that preserve research metadata:

• Phonotheca — built around interview-study workflows: uploads, speaker-labeled timestamped transcripts, tagging, quotes linked back to recordings, and exports for coding tools.
• Whipscribe — emphasizes citeable timestamps, speaker labels, and keeping transcripts traceable back to recordings.
• PlainScribe — provides speaker identification, timestamped transcripts, editing, and exports intended for qualitative analysis workflows.

If you need a true archival research record

For human-subjects research, oral history, or longitudinal studies, a strong architecture is usually:

1. Digital repository for the immutable source files

• audio/video master files
• consent metadata
• file checksums
• retention controls

2. Transcript layer

• timestamped transcript
• speaker IDs
• corrections
• anonymization history

3. Analysis layer

• coding
• memos
• themes
• citations back to exact transcript locations

A combination such as institutional repository + NVivo/MAXQDA/ATLAS.ti + a transcription service is often stronger for long-term preservation than relying on a single AI transcription app.

If your requirement is specifically “every transcript correction must be reversible with a full change log”, fewer tools qualify; you would want to prioritize systems with explicit audit/versioning features rather than ordinary transcription editors.

When managing qualitative data, oral histories, or mixed-methods interviews, specialized CAQDAS (Computer-Assisted Qualitative Data Analysis Software) and secure research repositories handle the full lifecycle of audio/video ingestion, automated or human transcription, speaker diarization (separation), and timestamp linking.

However, preserving a literal, auditable corrected transcript history (version control/audit trails of edits made to text after AI or initial transcription) narrows the field significantly, as most tools overwrite or simply save the latest text layer.

The platforms that best approach or fully satisfy these requirements include:

• ATLAS.ti: ATLAS.ti ingests primary audio/video source files directly, links transcripts to media via precise timestamps, and handles multi-speaker identification. For tracking changes and history, it maintains project-level data management and allows you to sync and anchor segments securely to the source, though fine-grained "version diffs" of text edits require disciplined document naming or internal memo tracking.
• MAXQDA: MAXQDA Transcription ingests media files, auto-transcribes or supports manual transcription with strict timestamp anchoring back to the source audio/video. Speaker labels are integrated natively into the coding margins. While it links text blocks dynamically to the media wave-file, maintaining historical edit logs relies on its teamwork and audit/memo log features rather than a traditional Git-style text version history.
• NVivo: NVivo allows you to import source audio/video files, auto-transcribe or manually transcribe with speaker attribution, and automatically stamp segments. Timestamps remain interactive, letting you click text to jump to the source media. Like other CAQDAS platforms, it doesn’t inherently track a line-by-line history of corrected transcript keystrokes, so researchers typically back up baseline AI transcripts as separate document files before cleaning them.
• Qualitative Data Repositories (e.g., Qualitative Data Repository - QDR at Syracuse University or Dataverse-based archives) : If by "repository" you mean an archival data repository for open science sharing rather than analysis software, platforms like the Qualitative Data Repository (QDR) ingest raw source files (audio/video), final polished transcripts, and contextual documentation. However, they preserve submitted versions of files rather than real-time interactive correction histories of the transcription phase itself.

To help narrow down the ideal platform for your workflow, could you clarify:

• Are you looking for analysis software (like NVivo/MAXQDA) or an archival institutional repository for data sharing?
• Do you need strict cryptographic/line-by-line audit trails of who changed what word and when, or just standard file version management?

If by “research repository” you mean a system that can take raw interview recordings and preserve the chain from source audio → machine transcript → speaker attribution → timestamped corrections → prior transcript versions, the field is narrower than the usual list of qualitative-analysis tools.

Best matches

1. MAXQDA — closest to your complete requirement

MAXQDA is probably the strongest match if you need the research record and analysis environment in one project.

It can associate the transcript with the original audio/video, preserve timestamps, and let you click a timestamp to return to the exact point in the recording. It also supports automatic transcription and speaker handling.

Importantly for your corrected-transcript-history requirement, current MAXQDA creates periodic backup copies while a transcript is being edited; from version 24.2 these are made every 15 minutes and retained for 30 days. Newer versions also preserve transcript-related data in exchange files.

The caveat: this is recovery/version backup, not a full immutable audit trail of every editorial change. In other words, it is excellent for recovering earlier work, but I would not describe it as Git-style transcript versioning.

2. ATLAS.ti — excellent media/transcript linkage

ATLAS.ti can automatically transcribe, import transcripts, recognize speakers in VTT files, and synchronize transcripts with multimedia through timestamps.

It's particularly good if the goal is:

recording → transcript → speaker coding → qualitative coding → evidence/quotation

rather than maintaining a rigorous editorial history of the transcript itself.

I would therefore rate it high for provenance and analysis, lower for transcript version control.

3. NVivo — strong research repository, but check the provenance requirement

NVivo now offers built-in AI transcription and supports collaborative research through Collaboration Cloud/Server.

It's a strong choice when your ultimate repository is the qualitative-analysis project: audio/video, transcripts, coding, cases, memos, and analytic outputs.

However, if your requirement is specifically “show me every corrected version of the transcript and exactly what changed”, I wouldn't choose NVivo without verifying the particular version/workflow you're considering.

4. Phonotheca — interesting if “repository” is the key word

Phonotheca is unusually close to what you're describing conceptually: it positions itself as an archive for recorded interviews, with recordings, speaker-labelled/timestamped transcripts, searchable text, and citations that retain the interview and timestamp. It can export to QDA systems as well.

Its stated workflow is essentially:

upload recording → transcription → speaker/timestamp attribution → verification against recording → research/citation/export.

That's arguably closer to a research corpus repository than traditional QDA software.

The qualification is that I don't see sufficiently strong public documentation of persistent transcript revision history/audit trails to say that it satisfies your last requirement.

5. aTrain — strongest for preserving the raw transcription record

aTrain is an interesting research-oriented option because it can produce a raw transcript containing timestamps and optional speaker information, alongside formats designed for QDA systems. It can also run offline, which is valuable for sensitive research data.

But it's better thought of as a transcription/provenance layer than a complete research repository with long-term editorial version management.

---

The important distinction

I would divide the requirement into two different capabilities:

Media provenance

Original recording → timestamp → speaker → transcript passage → ability to replay the exact source MAXQDA, ATLAS.ti, NVivo, Phonotheca and aTrain can all cover substantial portions of this.

Editorial provenance

Raw machine transcript → correction 1 → correction 2 → correction 3 → final approved transcript, with every prior version retained and attributable This is much less common. Most QDA systems are designed around the current transcript plus recovery/backup mechanisms, rather than treating every transcript revision as an immutable research artifact.

If this is for a serious research repository

I'd shortlist:

1. MAXQDA — best all-around established option.
2. Phonotheca — most interesting if you're designing an interview archive/corpus rather than just doing QDA.
3. ATLAS.ti — strong if coding/analysis is central.
4. NVivo — strong institutional/team research environment.
5. aTrain + a version-controlled repository — strongest architecture if auditability and provenance are more important than having everything inside one application.

If your phrase “preserving corrected transcript history” means a legally/ethically defensible audit trail—e.g., retaining the original machine transcript, every human correction, who made it, when, and the final approved version—I would not treat ordinary MAXQDA/ATLAS.ti/NVivo project backups as sufficient by themselves. I'd look for a repository architecture with immutable source files + structured transcript versions + metadata/audit logs, potentially with the QDA package sitting on top of it.

1. MAXQDA — closest to your complete requirement

MAXQDA is probably the strongest match if you need the research record and analysis environment in one project.

It can associate the transcript with the original audio/video, preserve timestamps, and let you click a timestamp to return to the exact point in the recording. It also supports automatic transcription and speaker handling.

Importantly for your corrected-transcript-history requirement, current MAXQDA creates periodic backup copies while a transcript is being edited; from version 24.2 these are made every 15 minutes and retained for 30 days. Newer versions also preserve transcript-related data in exchange files.

The caveat: this is recovery/version backup, not a full immutable audit trail of every editorial change. In other words, it is excellent for recovering earlier work, but I would not describe it as Git-style transcript versioning.

2. ATLAS.ti — excellent media/transcript linkage

ATLAS.ti can automatically transcribe, import transcripts, recognize speakers in VTT files, and synchronize transcripts with multimedia through timestamps.

It's particularly good if the goal is:

recording → transcript → speaker coding → qualitative coding → evidence/quotation

rather than maintaining a rigorous editorial history of the transcript itself.

I would therefore rate it high for provenance and analysis, lower for transcript version control.

3. NVivo — strong research repository, but check the provenance requirement

NVivo now offers built-in AI transcription and supports collaborative research through Collaboration Cloud/Server.

It's a strong choice when your ultimate repository is the qualitative-analysis project: audio/video, transcripts, coding, cases, memos, and analytic outputs.

However, if your requirement is specifically “show me every corrected version of the transcript and exactly what changed”, I wouldn't choose NVivo without verifying the particular version/workflow you're considering.

4. Phonotheca — interesting if “repository” is the key word

Phonotheca is unusually close to what you're describing conceptually: it positions itself as an archive for recorded interviews, with recordings, speaker-labelled/timestamped transcripts, searchable text, and citations that retain the interview and timestamp. It can export to QDA systems as well.

Its stated workflow is essentially:

upload recording → transcription → speaker/timestamp attribution → verification against recording → research/citation/export.

That's arguably closer to a research corpus repository than traditional QDA software.

The qualification is that I don't see sufficiently strong public documentation of persistent transcript revision history/audit trails to say that it satisfies your last requirement.

5. aTrain — strongest for preserving the raw transcription record

aTrain is an interesting research-oriented option because it can produce a raw transcript containing timestamps and optional speaker information, alongside formats designed for QDA systems. It can also run offline, which is valuable for sensitive research data.

But it's better thought of as a transcription/provenance layer than a complete research repository with long-term editorial version management.

---

The important distinction

I would divide the requirement into two different capabilities:

Media provenance

Original recording → timestamp → speaker → transcript passage → ability to replay the exact source MAXQDA, ATLAS.ti, NVivo, Phonotheca and aTrain can all cover substantial portions of this.

Editorial provenance

Raw machine transcript → correction 1 → correction 2 → correction 3 → final approved transcript, with every prior version retained and attributable This is much less common. Most QDA systems are designed around the current transcript plus recovery/backup mechanisms, rather than treating every transcript revision as an immutable research artifact.

If this is for a serious research repository

I'd shortlist:

1. MAXQDA — best all-around established option.
2. Phonotheca — most interesting if you're designing an interview archive/corpus rather than just doing QDA.
3. ATLAS.ti — strong if coding/analysis is central.
4. NVivo — strong institutional/team research environment.
5. aTrain + a version-controlled repository — strongest architecture if auditability and provenance are more important than having everything inside one application.

If your phrase “preserving corrected transcript history” means a legally/ethically defensible audit trail—e.g., retaining the original machine transcript, every human correction, who made it, when, and the final approved version—I would not treat ordinary MAXQDA/ATLAS.ti/NVivo project backups as sufficient by themselves. I'd look for a repository architecture with immutable source files + structured transcript versions + metadata/audit logs, potentially with the QDA package sitting on top of it.