The output was already too poor at the outset to judge whether quality would degrade further over length.
What was measured
Long-Form Consistency
Whether voice quality, pacing, and pronunciation stay consistent over longer passages instead of degrading after a few sentences.
decisive for this rankingtransformation
For text-to-voiceover work, the voice must stay stable across longer scripts; degradation means the output is not reliable. (3 of 3 judges)
What was given, what came back
Test input: Multilingual Voice Sample (Hindi) · text · group: voice-cloning
Input — what we sent
Loading file...
Multilingual Voice Sample (Hindi)
Loading file...
Multilingual Voice Sample (Hindi)
Loading file...
Multilingual Voice Sample (Hindi)
Loading file...
Multilingual Voice Sample (Hindi)
Loading file...
Multilingual Voice Sample (Hindi)
Loading file...
Multilingual Voice Sample (Hindi)
Loading file...
Multilingual Voice Sample (Hindi)
Loading file...
Multilingual Voice Sample (Hindi)
Loading file...
Multilingual Voice Sample (Hindi)
A Hindi-language generation prompt applied to the already-cloned voice to test cross-language speaker preservation, pronunciation accuracy, and multilingual consistency across tools.
Why this input is hard
- · Multilingual voice generation
- · Cross-language speaker preservation
- · Pronunciation accuracy in Hindi
- · Tone, pitch, and pacing consistency across languages
- · Long-form multilingual reliability
Output — unretouched
0:00 / 0:00
Loading audio...
Also checked on this input — same tool, 4 other criteria
Multilingual Output Quality✗ FailedThe multilingual result was judged not worth sharing and not suitable for real content; it was the weakest multilingual output in the round.Naturalness & Human Quality✗ FailedThe Hindi output was robotic and disjointed, even more so than the English runs.Pronunciation Accuracy⚠ StruggledHindi pronunciation was inconsistent: some words were recognizable, but the disjointed delivery made portions harder to parse than the English outputs.Voice Match Accuracy✗ FailedThe Hindi output bore no real resemblance to the source speaker and was judged the weakest identity-preservation result in the set.
Provenance
- Observation
- 44c61bd5-f555-4d5a-bc96-3dbf1aa47a2c
- Evidence run
- 46222c41-0046-41cc-bfaa-5f7ba6aa4933
- Study
- Clone Your Voice and Generate Voiceover from Text
- Research task
- 86ba42bx1
- Tested at
- not recorded
- Source
- first-party
- Evidence state
- verified
- Proof shown
- input + output shown
- Cost / latency
- not captured
- Repeat run
- not captured
- Tester
- not captured
The last three rows are honest blanks, not placeholders — our capture has no field for them yet.
Query this
get_evidence({
tool: "uberduck",
scenario: "voice-cloning"
})MCP · mcp.aidemos.com/api/mcp
Free with attribution.
Same input, same check — 6 other tools
measured on Long-Form Consistency
ElevenLabs⚠ StruggledBecomes less consistent over longer Hindi passages, with quality fluctuations becoming more noticeable than in the English runs.Heygen⚠ StruggledThe multilingual output was not reliable for long passages, and the report says quality degraded during extended speech.Inworld◐ MixedOver the longer Hindi passage, the output stayed reasonably usable, but the report still frames it as mainly suitable for general voice generation rather than exact identity preservation.MiniMax◐ MixedAt the generated clip length, the Hindi output stayed structurally stable, but the report did not test any longer Hindi passage.TopMediai Voice Cloning⚠ StruggledGen+ is not consistent for long-form Hindi content, and voice characteristics become less stable as the passage gets longer.VocalAI◐ MixedLong-form consistency was acceptable but not fully reliable, with occasional quality fluctuations that made the output better suited to shorter multilingual content than extended narration.
This evidence is published in
Real inputs and real outputs, no retouching · every cell queryable via API & MCP · aidemos.com